fix: /local-workspace — chemin du header + clic sur un dossier du sidebar (v7.69.9)

Corrige deux régressions de /local-workspace :

- le chemin du header restait bloqué sur « Home / <workspace> » quel que
  soit le dossier affiché : la route rend désormais breadcrumb_items
  (Home / <workspace> / <dossier> / <sous-dossier>, niveaux cliquables,
  collapse « … » au-delà de 4) et la navigation sans rechargement recalcule
  le chemin via l'event flowdeck:breadcrumb-changed ;
- le clic sur un dossier du sidebar affichait TOUS les composants à la
  fois : Alpine.data('wsInitData') retournait le même objet singleton, le
  2e montage (navigation partielle) levait « Cannot redefine property:
  \ » et initTree abandonnait, laissant tout le contenu au state
  brut. La factory retourne désormais une enveloppe fraîche par montage
  qui délègue à l'état réactif partagé. #lw-config est aussi relu à chaque
  exécution (le 2e montage gardait le folder_id du 1er chargement).

Inclus également le travail en cours de l'arbre : Library (colonnes Last
visited/Source, ordre d'en-tête, favoris à icônes Workspace), Meeting
Notes (bloc, CSS, routes, docs), coloration de code hljs, badges
favori/publié dans l'arbre local-workspace, docs (DATA_MODEL,
architectures) et tests associés.

CHANGELOG.md

@@ -1,6 +1,172 @@
 # Changelog - FlowDeck
 
-## v7.69.4 (2026-10-08) — Le close de Settings ferme vraiment
+## v7.69.9 (2026-10-09) — /local-workspace : chemin du header + navigation dans les dossiers réparée
+
+### Fixed
+
+- **/local-workspace : le chemin du header suit la navigation** — le header
+  restait bloqué sur « Home / <workspace> » quel que soit le dossier
+  affiché. La route rend désormais `breadcrumb_items`
+  (`_render_local_workspace`) au format des pages
+  (`_nav_breadcrumb`) : **Home / <workspace> / <dossier> / <sous-dossier> /
+  …**, chaque niveau cliquable (`/local-workspace?folder=<id>`) et le
+  dernier niveau courant ; au-delà de 4 niveaux, collapse « … » identique
+  à l'éditeur de page. La navigation **sans rechargement** (clic sur un
+  dossier dans la vue, breadcrumb interne, retour/avant) recalcule le
+  chemin via le nouvel event `flowdeck:breadcrumb-changed`, écouté par
+  `fdBreadcrumb` (`_header.html`) — les renommages continuent de suivre
+  via `flowdeck:page-renamed`.
+- **/local-workspace : clic sur un dossier du sidebar = tous les composants
+  affichés en même temps** — après une navigation partielle
+  (`/local-workspace` → `/local-workspace?folder=<id>`), Alpine échouait
+  à remonter la page : `Alpine.data('wsInitData')` retournait **le même
+  objet singleton** et Alpine redéfinit ses magics (`$nextTick`…) avec
+  `Object.defineProperty` non-configurable →
+  « Cannot redefine property: $nextTick », `initTree` abandonnait et
+  **tout le contenu restait au state brut** (tree/list/détails/cartes,
+  dropdowns, toast Undo, barre d'upload… visibles simultanément).
+  La factory retourne désormais une enveloppe fraîche à chaque montage,
+  qui délègue lecture/écriture au proxy réactif partagé (état et
+  `wsPreview` inchangés).
+- **/local-workspace : le dossier de l'URL était perdu en navigation
+  partielle** — `LW` (config `#lw-config`) n'était lu qu'à la première
+  exécution du script (verrouillé par `__fdLocalWorkspaceScriptsLoaded`) :
+  le 2e montage gardait le `folder_id`/`workspace_id` du premier
+  chargement. La config est relue à chaque exécution et l'état transitoire
+  (recherche, filtre, panneaux) est remis à zéro au montage.
+- **/local-workspace : breadcrumb interne toujours rendu** — il n'apparaît
+  que si la page était chargée avec `?folder=` ; il est désormais présent
+  dès la racine et suit la pile de dossiers.
+
+### Tests
+
+- `e2e/regression_local_workspace_breadcrumb.spec.js` : parcours complet
+  (sidebar → dossiers/sous-dossiers, clic dans la vue, breadcrumb
+  interne, crumb workspace du header) avec vérification du chemin header
+  et absence de composants peints à l'état brut.
+
+## v7.69.8 (2026-10-09) — Library : ordre des colonnes visited/source
+
+### Fixed
+
+- **Combinaison « Last visited time » + Source illisible** — les cellules
+  de données rendaient Source avant Edited/Visited alors que l'en-tête
+  l'attend en dernier : les valeurs tombaient sous les mauvais intitulés.
+  L'ordre des cellules suit désormais l'en-tête (Name, Created by, Last
+  edited by, Last edited time, Last visited time, Source), et les intitulés
+  d'en-tête tronquent en ellipsis au lieu de déborder sur la colonne
+  suivante (reproduit et mesuré au navigateur : alignement header/lignes
+  vérifié colonne par colonne).
+
+### Fixed
+
+- **Sidebar Favorites : icônes Workspace** — la liste utilisait des emoji
+  bruts (voire rien) via `item.icon` ; `_build_item` renvoie désormais les
+  noms SVG (`folder`/`file`/`edit`/`image`), idem pour les favoris board.
+- **/local-workspace : indicateurs de statut** — ★ favori (ambre), 👥
+  partagé (existant), 🌐 publié ; exposés par `_build_tree_children`
+  (`favorited`, `published`, `is_shared`) et rendus dans l'arbre Alpine
+  comme dans le rendu JS des enfants.
+- **Library « Created by » vide** — aucune source auteur n'était
+  sélectionnée. Nouvelle provenance groupée : created by (première
+  version, sinon propriétaire du workspace), last edited by (dernière
+  version), last visited (table `recents`, tracée à chaque lecture de
+  page). Nouvelles colonnes **Last edited by**, **Last edited time**,
+  **Last visited time** ; le menu **Show columns** pilote chaque colonne,
+  seul Page name est verrouillé.
+- **Colonne Source avec icône** — `file` (local), `link` (Gitea),
+  `external-link` (GitHub), mêmes SVG que le reste de Flowdeck.
+- **Show columns réellement fonctionnel** — les règles de visibilité
+  (spécificité 0,1,0) perdaient contre `.lib-table-header .th` et
+  `.lib-row .row-cell` (0,2,0) : seule Last edited time s'affichait et
+  masquait, les autres options semblaient mortes (reproduit et corrigé
+  avec vérification en direct : les 5 toggles on/off + colonne Source
+  avec icône + libellé, 0 erreur console).
+
+### Tests
+
+- `tests/test_v696_sidebar_library_publish.py` (+3) : provenance et visite
+  tracée, icônes favoris, flags de l'arbre local-workspace.
+
+## v7.69.6 (2026-10-09) — Library AI Meeting Notes, Recents réels, Publish unifié
+
+### Added
+
+- **Onglet Library « AI Meeting Notes »** (`GET /api/library/meetings`) :
+  notes de réunion de l'utilisateur regroupées par date — Today,
+  Yesterday, This week, Last week, This month, Older — avec icône
+  calendrier, statut (Summary ready / Ready to record) en colonne Source,
+  recherche et sélection réutilisées. Les notes sont exclues des onglets
+  Recents / Workspace / Private (ainsi que de l'arbre Workspace et du
+  navigateur local).
+
+### Fixed
+
+- **Sidebar Recents vide** — elle ne listait que les dépôts Gitea. Elle
+  liste désormais les fichiers réellement récents (`updated_at`, 8
+  derniers, dépôts conservés ensuite), avec les mêmes icônes SVG que le
+  Workspace et les notes de réunion exclues (→ Meetings).
+- **Customize sidebar inutilisable** — l'overlay plein écran (z-index 998)
+  recouvrait aussi le panneau d'options : chaque clic refermait le mode.
+  Le panneau passe au-dessus (z-index 999).
+- **Publish qui ne « tenait » pas** — le share dialog écrivait `published`,
+  les routes `/publish` écrivaient `is_published`, et chaque lecteur ne
+  regardait qu'une colonne. Toutes les écritures maintiennent désormais
+  les deux drapeaux (+ slug garanti), tous les lecteurs acceptent l'un ou
+  l'autre : la page publiée reste publiée dans le dialogue, apparaît dans
+  la section Published de la sidebar et dans l'onglet Published de la
+  Library.
+
+### Tests
+
+- `tests/test_v696_sidebar_library_publish.py` (6) : onglet meetings et
+  exclusions, Recents sidebar, publish/unpublish des deux chemins,
+  visibilité sidebar+library+dialogue, z-index du panneau customize.
+
+## v7.69.5 (2026-10-09) — Meeting Notes : setup visible, retry qui ne perd plus rien
+
+### Fixed
+
+- **Pas de « Setup check » visible avant l'enregistrement** — il était caché
+  dans le menu ▾ du bouton Start, que personne n'ouvre. Le panneau idle
+  affiche désormais une bande de vérification : état de la permission micro
+  (via Permissions API), transcription live dispo ou repli manuel, mode
+  onglet supporté ou non, et bouton **Test microphone** (4 s de niveau réel,
+  qui fait aussi accorder l'autorisation avant le Start).
+- **Erreur 400 persistante au Stop/Retry après un collage de transcript** —
+  `retry()`/`regenerate()` appelaient le serveur sans y verser d'abord le
+  contenu local (segments + manuel collé) : le serveur restait vide et
+  répondait 400 en boucle. Ils drainent désormais le tampon local (attente
+  bornée) avant tout appel, et `retry()` détecte une session encore en vie
+  côté serveur pour terminer par `/stop` plutôt que `/processing:retry`.
+- **Stop appelé alors que l'upload des segments échoue** — condamnait à un
+  400 (« transcript is empty ») malgré des segments en local. Le client
+  n'appelle plus `/stop` dans ce cas : état `failed` avec message « envoi
+  impossible, Retry », session serveur conservée.
+- **`/summarize` accepte `allow_empty`** (diagnostic `fragmentary` honnête
+  au lieu de 400), comme `/stop` et `/processing:retry`.
+- **Cache-busting** : le bump 7.69.4 → 7.69.5 change `?v=` de tous les
+  assets — les navigateurs ayant gardé l'ancien `meeting_block.js` (sans
+  garde-fou ni diagnostic) reçoivent le nouveau code. C'était la cause la
+  plus probable du « toujours l'erreur 400 ».
+- **Menu contextuel des sections Meetings et Workspace** : les notes de la
+  section Meetings ont désormais le même clic droit que le Workspace
+  (Open / Rename / Delete, via `showWsContext`) ; l'action **Open** est
+  ajoutée au menu des deux sections (page → `/pages/{id}`, dossier →
+  navigation dossier) ; rename/delete rafraîchissent la section Meetings.
+- **Sidebar sur /local-workspace** : la règle CSS globale `.ws-tree-item`
+  de la page écrasait les items de la sidebar (espacements différents) —
+  scopée à `.ws-tree` (arbre principal) ; et `refreshSidebarTree()`
+  remonte désormais le fragment via `Alpine.initTree`, sans quoi l'arbre
+  perdait menu contextuel, toggles et drag&drop après chaque refresh
+  (d'où le menu « mort » sur cette page qui rafraîchit en continu).
+
+### Tests
+
+- `tests/test_v711_meeting_notion.py` (+5) : `summarize` vide → 400 sans
+  flag / diagnostic avec flag, retry intelligent quand la session serveur
+  est encore live, `meeting_notes` + propriété, exclusion de l'arbre
+  Workspace.
 
 ### Fixed
 

VERSION

@@ -1 +1 @@
-7.69.4
+7.69.9

app/main.py

@@ -186,7 +186,7 @@ async def lifespan(_app: FastAPI):
 
 app = FastAPI(
     title="FlowDeck",
-    version="7.69.4",
+    version="7.69.9",
     docs_url="/docs",
     redoc_url="/redoc",
     lifespan=lifespan,

app/middleware/security.py

@@ -20,16 +20,54 @@ ALLOWED_EXTENSIONS: frozenset[str] = frozenset({
     # Images
     ".png", ".jpg", ".jpeg", ".gif", ".webp", ".svg", ".bmp", ".ico",
     # Documents
-    ".pdf", ".md", ".markdown", ".txt", ".log", ".csv",
-    # Code
-    ".py", ".js", ".jsx", ".ts", ".tsx", ".html", ".htm", ".xml", ".css",
-    ".json", ".yaml", ".yml", ".toml", ".sql", ".sh", ".bash", ".zsh",
-    ".ps1", ".rs", ".go", ".java", ".rb", ".php", ".c", ".cpp", ".h",
-    ".swift", ".kt", ".scala", ".r",
+    ".pdf", ".md", ".markdown", ".mkd", ".mdown", ".mdwn", ".rmd", ".qmd",
+    ".txt", ".log", ".csv", ".tsv", ".tab", ".nfo", ".asc",
+    ".tex", ".latex", ".sty", ".cls", ".dtx", ".bib", ".bbl",
+    # Code — kept in sync with app/services/export.py::_TEXTUAL_EXTS
+    ".py", ".pyw", ".pyi", ".gyp", ".gypi",
+    ".js", ".mjs", ".cjs", ".jsx", ".ts", ".mts", ".cts", ".tsx",
+    ".html", ".htm", ".xhtml", ".xht", ".xml", ".xsd", ".xsl", ".xslt",
+    ".rss", ".atom", ".plist", ".wsf", ".axml",
+    ".vue", ".svelte", ".astro",
+    ".css", ".scss", ".sass", ".less",
+    ".json", ".json5", ".jsonc", ".geojson", ".webmanifest",
+    ".har", ".avsc",
+    ".yaml", ".yml", ".toml", ".ini", ".cfg", ".conf", ".env",
+    ".properties", ".prop", ".inf", ".reg", ".url",
+    ".gitconfig", ".gitmodules", ".editorconfig",
+    ".sql", ".pgsql", ".psql", ".mysql", ".mssql",
+    ".sh", ".bash", ".zsh", ".ksh", ".mksh", ".yash", ".fish",
+    ".ps1", ".psm1", ".psd1", ".bat", ".cmd",
+    ".rb", ".gemspec", ".rake", ".rbw",
+    ".rs", ".go", ".java", ".jsp", ".kt", ".kts",
+    ".scala", ".sc", ".groovy", ".gvy", ".gradle",
+    ".c", ".cc", ".cpp", ".c++", ".cxx", ".hh", ".hxx", ".hcc", ".h", ".hpp",
+    ".cs", ".csx", ".cake", ".ino", ".pde",
+    ".php", ".phtml", ".swift", ".r", ".jl", ".lua", ".dart",
+    ".pl", ".pm", ".pod",
+    ".hs", ".lhs", ".ex", ".exs", ".eex", ".heex", ".leex",
+    ".erl", ".hrl", ".escript",
+    ".clj", ".cljs", ".cljc", ".edn",
+    ".ml", ".mli", ".fs", ".fsi", ".fsx", ".fsscript",
+    ".nim", ".nims", ".d", ".di",
+    ".m", ".mm", ".vb", ".vbs", ".bas", ".pas", ".pp", ".dpr",
+    ".vhd", ".vhdl", ".v", ".vh", ".sv", ".svh",
+    ".asm", ".s", ".nasm", ".inc",
+    ".ada", ".ads", ".adb",
+    ".f", ".for", ".f90", ".f95", ".f03", ".f08",
+    ".cob", ".cbl", ".cpy",
+    ".dockerfile", ".containerfile",
+    ".tf", ".tfvars", ".hcl",
+    ".cmake", ".makefile", ".mk", ".mak", ".ninja",
+    ".gql", ".graphql", ".graphqls", ".proto", ".thrift",
+    ".vim", ".vimrc", ".gvimrc", ".awk",
+    ".gcode", ".gco", ".nc", ".ngc",
+    ".http", ".rest", ".nginx", ".apache", ".htaccess", ".wat",
+    ".diff", ".patch", ".rej",
+    ".mod", ".sum", ".lock", ".text",
+    ".gitignore", ".gitattributes", ".gitkeep", ".dockerignore", ".npmignore",
     # Archives
     ".zip", ".tar", ".gz", ".rar", ".7z",
-    # Misc
-    ".env", ".cfg", ".conf", ".ini", ".dockerfile", ".makefile",
 })
 
 MAX_UPLOAD_SIZE = 10 * 1024 * 1024  # 10 MB

app/migrations.py

@@ -1725,3 +1725,80 @@ def _migration_mistral_default_model(conn: sqlite3.Connection) -> None:
         "WHERE provider='mistral' AND model IN (?,?)",
         blocked,
     )
+
+
+@register(38, "v7.1.1: meeting block Notion — états, segments, consentement, partage")
+def _migration_meeting_notion_block(conn: sqlite3.Connection) -> None:
+    """v7.1.1 — refonte Meetings façon Notion (§3A/3B du doc d'architecture).
+
+    ``meeting_transcripts`` gagne la machine à états du bloc
+    (idle/recording/paused/processing/done/failed), les segments groupés par
+    source, la photo d'instructions, le journal de consentement résumé, les
+    étapes Thinking persistées, le résumé structuré + titre généré + classe de
+    complétude. Deux tables append-only : consentement et diffusions.
+    Règle d'unicité : une occurrence calendrier = au plus une note.
+    """
+    cols = columns(conn, "meeting_transcripts")
+    for ddl in (
+        "ALTER TABLE meeting_transcripts ADD COLUMN status TEXT NOT NULL DEFAULT 'idle'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN mode TEXT NOT NULL DEFAULT 'mic_only'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN channels_json TEXT NOT NULL DEFAULT '[]'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN instruction_id TEXT NOT NULL DEFAULT 'auto'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN instruction_snapshot TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN consent_json TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN segments_json TEXT NOT NULL DEFAULT '[]'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN processing_steps_json TEXT NOT NULL DEFAULT '[]'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN summary_json TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN generated_title TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN title_source TEXT NOT NULL DEFAULT 'user'",
+        "ALTER TABLE meeting_transcripts ADD COLUMN completeness_class TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN quality_score REAL",
+        "ALTER TABLE meeting_transcripts ADD COLUMN duration_ms INTEGER NOT NULL DEFAULT 0",
+        "ALTER TABLE meeting_transcripts ADD COLUMN paused_ms INTEGER NOT NULL DEFAULT 0",
+        "ALTER TABLE meeting_transcripts ADD COLUMN started_at TIMESTAMP",
+        "ALTER TABLE meeting_transcripts ADD COLUMN stopped_at TIMESTAMP",
+        "ALTER TABLE meeting_transcripts ADD COLUMN event_occurrence_id TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN notes_text TEXT NOT NULL DEFAULT ''",
+        "ALTER TABLE meeting_transcripts ADD COLUMN banner_dismissed INTEGER NOT NULL DEFAULT 0",
+        "ALTER TABLE meeting_transcripts ADD COLUMN share_bar_dismissed INTEGER NOT NULL DEFAULT 0",
+    ):
+        name = ddl.split("ADD COLUMN")[1].strip().split()[0]
+        if name not in cols:
+            conn.execute(ddl)
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS meeting_consent_log (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            transcript_id INTEGER NOT NULL REFERENCES meeting_transcripts(id) ON DELETE CASCADE,
+            method TEXT NOT NULL DEFAULT 'start_attestation',
+            message_ref TEXT NOT NULL DEFAULT '',
+            attested_by INTEGER REFERENCES users(id),
+            participants_json TEXT NOT NULL DEFAULT '[]',
+            created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
+        )
+        """
+    )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS meeting_distributions (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            transcript_id INTEGER NOT NULL REFERENCES meeting_transcripts(id) ON DELETE CASCADE,
+            channel TEXT NOT NULL DEFAULT 'copy_link',
+            actor_id INTEGER REFERENCES users(id),
+            recipients_json TEXT NOT NULL DEFAULT '',
+            status TEXT NOT NULL DEFAULT 'created',
+            created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
+        )
+        """
+    )
+    conn.execute(
+        "CREATE INDEX IF NOT EXISTS idx_meeting_consent_tr ON meeting_consent_log(transcript_id)"
+    )
+    conn.execute(
+        "CREATE INDEX IF NOT EXISTS idx_meeting_distr_tr ON meeting_distributions(transcript_id)"
+    )
+    conn.execute(
+        "CREATE UNIQUE INDEX IF NOT EXISTS ux_meeting_one_per_occurrence "
+        "ON meeting_transcripts(event_occurrence_id) "
+        "WHERE event_occurrence_id <> ''"
+    )

app/routers/board/_common.py

@@ -252,7 +252,12 @@ def _file_icon(name: str, content_format: str = "") -> str:
 
 
 def _load_workspace_pages(ws_cookie: str) -> list:
-    """Load top-level pages with children for the active workspace."""
+    """Load top-level pages with children for the active workspace.
+
+    v7.1.1 : les pages contenant un bloc AI Meeting Notes sont des notes de
+    réunion — elles vivent dans la section Meetings de la sidebar
+    (``GET /api/v2/meetings/notes``), pas dans l'arbre Workspace.
+    """
     if not ws_cookie:
         return []
     try:
@@ -261,7 +266,8 @@ def _load_workspace_pages(ws_cookie: str) -> list:
             rows = conn.execute(
                 "SELECT id, title, parent_section, content_format, "
                 "is_shared, share_mode, COALESCE(published,0) AS published "
-                "FROM pages WHERE workspace_id=? AND parent_id IS NULL AND deleted_at IS NULL AND collection_row_id IS NULL ORDER BY created_at DESC",
+                "FROM pages WHERE workspace_id=? AND parent_id IS NULL AND deleted_at IS NULL AND collection_row_id IS NULL "
+                "AND REPLACE(COALESCE(content,''), ' ', '') NOT LIKE '%\"type\":\"meeting\"%' ORDER BY created_at DESC",
                 (ws_id,),
             ).fetchall()
         items = []
@@ -289,12 +295,13 @@ def _load_workspace_pages(ws_cookie: str) -> list:
 
 
 def _load_children(parent_id: int) -> list:
-    """Recursively load children of a page."""
+    """Recursively load children of a page (meeting notes excluded, § ci-dessus)."""
     with get_conn() as conn:
         rows = conn.execute(
             "SELECT id, title, parent_section, content_format, "
             "is_shared, share_mode, COALESCE(published,0) AS published "
-            "FROM pages WHERE parent_id=? AND deleted_at IS NULL ORDER BY created_at",
+            "FROM pages WHERE parent_id=? AND deleted_at IS NULL "
+            "AND REPLACE(COALESCE(content,''), ' ', '') NOT LIKE '%\"type\":\"meeting\"%' ORDER BY created_at",
             (parent_id,),
         ).fetchall()
     children = []
@@ -382,7 +389,7 @@ def _load_shared_sidebar_pages(user_id: int) -> tuple[list, list, list, list]:
         ).fetchall()
         published_rows = conn.execute(
             f"SELECT id, title, workspace, updated_at FROM pages "
-            f"WHERE published=1 AND deleted_at IS NULL AND {scope_cond} "
+            f"WHERE (published=1 OR is_published=1) AND deleted_at IS NULL AND {scope_cond} "
             f"ORDER BY updated_at DESC LIMIT 20",
             scope_params,
         ).fetchall()
@@ -521,17 +528,21 @@ def _sidebar_data(request: Request, owner: str = "", repo: str = "") -> dict:
     uid = user["id"] if user and user.get("id") else 1
     with get_conn() as conn:
         fav_rows = conn.execute(
-            "SELECT p.id, p.title, p.workspace, p.updated_at FROM favorites f "
+            "SELECT p.id, p.title, p.workspace, p.updated_at, p.parent_section, "
+            "p.content_format FROM favorites f "
             "JOIN pages p ON p.id = f.page_id "
-            "WHERE f.user_id=? ORDER BY f.position", (uid,)
+            "WHERE f.user_id=? AND p.deleted_at IS NULL ORDER BY f.position", (uid,)
         ).fetchall()
     favorites = []
     for r in fav_rows:
+        # v7.69.6 : mêmes icônes que le Workspace (noms SVG, pas d'emoji).
+        is_folder = r["parent_section"] == "Workspace"
         favorites.append({
             "id": f"page/{r['id']}",
             "db_id": r["id"],
             "name": r["title"] or "New page",
-            "icon": "📄",
+            "icon": "folder" if is_folder else _file_icon(
+                r["title"] or "", r["content_format"]),
             "url": f"/pages/{r['id']}",
             "active": False,
             "indent": 0,

app/routers/board/sharing.py

@@ -95,9 +95,22 @@ def update_share(request: Request, page_id: int, body: dict = Body(default={})):
     mode = body.get("mode", "private")
     published = body.get("published", False)
     with get_conn() as conn:
+        # v7.69.6 : les deux drapeaux ensemble (voir services/publish), plus
+        # un slug garanti quand on publie depuis le share dialog.
+        from app.services.publish import _unique_slug
+        row = conn.execute(
+            "SELECT publish_slug FROM pages WHERE id=?", (page_id,)
+        ).fetchone()
+        slug = (row["publish_slug"] if row and row["publish_slug"] else "")
+        if published and not slug:
+            title_row = conn.execute(
+                "SELECT title FROM pages WHERE id=?", (page_id,)
+            ).fetchone()
+            slug = _unique_slug(
+                conn, page_id, title_row["title"] if title_row else "")
         conn.execute(
-            "UPDATE pages SET share_mode=?, published=? WHERE id=?",
-            (mode, 1 if published else 0, page_id),
+            "UPDATE pages SET share_mode=?, published=?, is_published=?, publish_slug=? WHERE id=?",
+            (mode, 1 if published else 0, 1 if published else 0, slug, page_id),
         )
         conn.commit()
     return {"status": "ok", "share_mode": mode, "published": published}

app/routers/dashboard/_common.py

@@ -90,6 +90,50 @@ def _local_workspaces_for_user(user: dict | None, workspace_id: int = 0) -> list
 
 
 
+def _recent_sidebar_pages(uid: int, limit: int = 8) -> list:
+    """Fichiers réellement récents (dernière édition) pour la sidebar.
+
+    v7.69.6 : avant, seuls les dépôts Gitea alimentaient la section Recents
+    (vide sans dépôt). Icônes identiques au Workspace (noms SVG, pas
+    d'emoji). Notes de réunion exclues → section Meetings.
+    """
+    if not uid:
+        return []
+    try:
+        from app.routers.board import _file_icon
+        with get_conn() as conn:
+            rows = conn.execute(
+                """SELECT p.id, p.title, p.parent_section, p.content_format
+                   FROM pages p
+                   LEFT JOIN workspaces w ON w.id = p.workspace_id
+                   WHERE p.deleted_at IS NULL AND p.parent_section != 'Trash'
+                     AND (w.owner_id = ? OR p.workspace_id IS NULL)
+                     AND REPLACE(COALESCE(p.content,''), ' ', '')
+                         NOT LIKE '%"type":"meeting"%'
+                   ORDER BY p.updated_at DESC LIMIT ?""",
+                (uid, limit),
+            ).fetchall()
+            out = []
+            for r in rows:
+                is_folder = r["parent_section"] == "Workspace"
+                out.append({
+                    "id": f"page/{r['id']}",
+                    "name": r["title"] or "Untitled",
+                    "icon": "folder" if is_folder else _file_icon(
+                        r["title"] or "", r["content_format"]),
+                    "url": f"/pages/{r['id']}",
+                    "active": False,
+                    "indent": 0,
+                    "depth": 0,
+                    "has_children": False,
+                    "children": [],
+                })
+            return out
+    except Exception:
+        logger.exception("_recent_sidebar_pages")
+        return []
+
+
 def _sidebar_data(request: Request, repos: list[dict], include_workspace: bool = True) -> dict:
     user = SessionManager.decode_session(request.cookies.get("flowdeck_session", ""))
     ws = user.get("login", "Bruno") if user else "Bruno"
@@ -108,7 +152,7 @@ def _sidebar_data(request: Request, repos: list[dict], include_workspace: bool =
         except Exception:
             logger.exception("_sidebar_data")
 
-    recent_pages = []
+    recent_pages = _recent_sidebar_pages(user["id"] if user else 0) if user else []
     for repo in repos[:10]:
         full_name = repo.get("full_name", "")
         recent_pages.append({
@@ -359,21 +403,26 @@ def _require_page_view(request: Request, page_id: int) -> None:
 
 def _build_tree_children(conn, parent_id: int | None, ws_id: int, uid: int | None = None) -> list:
     """Recursively build the tree of children for a node."""
+    # v7.69.6 : notes de réunion exclues (section Meetings, comme la sidebar).
+    no_meet = "AND REPLACE(COALESCE(content,''), ' ', '') NOT LIKE '%\"type\":\"meeting\"%'"
     if parent_id is None:
         rows = conn.execute(
             "SELECT id, title, parent_section, content_format, content, page_icon, "
-            "(is_shared OR share_mode != 'private' OR COALESCE(published,0)) as is_shared, "
+            "COALESCE(is_published,0) AS pub1, COALESCE(published,0) AS pub2, "
+            "(is_shared OR share_mode != 'private' OR COALESCE(published,0) OR COALESCE(is_published,0)) as is_shared, "
             "created_at, updated_at FROM pages "
             "WHERE parent_id IS NULL AND workspace_id=? AND deleted_at IS NULL AND collection_row_id IS NULL "
+            f"{no_meet}"
             "ORDER BY created_at DESC",
             (ws_id,),
         ).fetchall()
     else:
         rows = conn.execute(
             "SELECT id, title, parent_section, content_format, content, page_icon, "
-            "(is_shared OR share_mode != 'private' OR COALESCE(published,0)) as is_shared, "
+            "COALESCE(is_published,0) AS pub1, COALESCE(published,0) AS pub2, "
+            "(is_shared OR share_mode != 'private' OR COALESCE(published,0) OR COALESCE(is_published,0)) as is_shared, "
             "created_at, updated_at FROM pages "
-            "WHERE parent_id=? AND workspace_id=? AND deleted_at IS NULL "
+            f"WHERE parent_id=? AND workspace_id=? AND deleted_at IS NULL {no_meet} "
             "ORDER BY created_at DESC",
             (parent_id, ws_id),
         ).fetchall()
@@ -441,6 +490,7 @@ def _build_tree_children(conn, parent_id: int | None, ws_id: int, uid: int | Non
             "author": author,
             "tags": tags_map.get(r["id"], []),
             "is_shared": bool(r["is_shared"]),
+            "published": bool(r["pub1"] or r["pub2"]),
             "favorited": r["id"] in favorited_ids,
         })
     return tree
@@ -671,12 +721,17 @@ def _render_blocks_public(blocks: list, titles: dict | None = None) -> str:
             )
         elif t == "code":
             lang = b.get("language", "")
+            try:
+                from app.services.export import normalize_code_lang
+                hljs_lang = normalize_code_lang(lang)
+            except Exception:  # noqa: BLE001 — never break public rendering
+                hljs_lang = "plaintext"
             lang_label = f"<div style='font-size:11px;opacity:.4;margin-bottom:8px;'>{lang}</div>" if lang else ""
             html_parts.append(
                 f'<pre style="background:rgba(255,255,255,.05);padding:16px 20px;border-radius:8px;'
                 f'overflow-x:auto;font-size:14px;line-height:1.5;margin:12px 0;">'
                 f'{lang_label}'
-                f'<code>{c}</code></pre>'
+                f'<code class="hljs language-{hljs_lang}" data-hljs-lang="{hljs_lang}">{c}</code></pre>'
             )
         elif t == "divider":
             html_parts.append('<hr style="border:none;border-top:1px solid rgba(255,255,255,.1);margin:16px 0;">')

app/routers/dashboard/local_workspace.py

@@ -100,6 +100,28 @@ def _render_local_workspace(env, sidebar: dict, user: dict, ws_row: dict,
         with get_conn() as conn:
             breadcrumb = _build_breadcrumb(conn, folder)
 
+    # En-tête Notion-style (même mécanique que les pages) :
+    # Home / <workspace> / <dossier> / <sous-dossier> / …
+    # Chaque niveau cliquable renvoie à /local-workspace?folder=<id> — sans
+    # ça le header restait bloqué sur « Home / <workspace> » même quand le
+    # contenu affiché était celui d'un sous-dossier.
+    nav_crumbs: list = [{
+        "id": None,
+        "label": ws_name,
+        "url": "/local-workspace",
+        "icon": "folder",
+        "menu": False,
+    }]
+    for i, bcrumb in enumerate(breadcrumb):
+        is_last = i == len(breadcrumb) - 1
+        nav_crumbs.append({
+            "id": bcrumb["id"],
+            "label": bcrumb["name"],
+            "url": None if is_last else f"/local-workspace?folder={bcrumb['id']}",
+            "icon": "folder",
+            "menu": True,
+        })
+
     ctx = {
         **sidebar,
         "user": user,
@@ -111,6 +133,7 @@ def _render_local_workspace(env, sidebar: dict, user: dict, ws_row: dict,
         "nav_workspace_id": ws_id or 0,
         "breadcrumb": breadcrumb,
         "breadcrumbs": breadcrumb,
+        "breadcrumb_items": nav_crumbs,
     }
     template = env.get_template("local_workspace.html")
     return HTMLResponse(

app/routers/dashboard/pages_html.py

@@ -94,6 +94,24 @@ def view_page_root(request: Request, page_id: int):
         if not row:
             return RedirectResponse("/workspaces", status_code=302)
     page = dict(row)
+    # v7.69.6 : tracer la visite (colonne Library « Last visited », table recents).
+    try:
+        from datetime import UTC, datetime
+        _sess = SessionManager.decode_session(request.cookies.get("flowdeck_session", ""))
+        if _sess and _sess.get("id"):
+            with get_conn() as _conn:
+                _conn.execute(
+                    """INSERT INTO recents (user_id, page_id, workspace, source_type, accessed_at)
+                       VALUES (?, ?, ?, ?, ?)
+                       ON CONFLICT(user_id, page_id)
+                       DO UPDATE SET workspace=excluded.workspace,
+                                     accessed_at=excluded.accessed_at""",
+                    (_sess["id"], page_id, page.get("workspace", "") or "",
+                     "local", datetime.now(UTC).replace(tzinfo=None).isoformat()),
+                )
+                _conn.commit()
+    except Exception:
+        pass
     # v6.5.0: synced blocks resolve server-side at read time.
     from app.services.synced_blocks import resolve_content_json
     page["content"] = resolve_content_json(page.get("content", ""), page.get("content_format"))
@@ -119,7 +137,7 @@ def view_page_root(request: Request, page_id: int):
     _sess_user = SessionManager.decode_session(request.cookies.get("flowdeck_session", ""))
     _uid = _sess_user.get("id") if _sess_user and _sess_user.get("id") else None
     _can_edit = (not _locked) or bool(_sess_user and _sess_user.get("is_admin")) or (_locked_by and _uid and _locked_by == _uid)
-    page_data = {"id": page_id, "title": page.get("title"), "content_format": page.get("content_format", "blocks"), "content": page.get("content", ""), "favorited": fav is not None, "share_mode": page.get("share_mode", "private"), "published": bool(page.get("published", 0)),
+    page_data = {"id": page_id, "title": page.get("title"), "content_format": page.get("content_format", "blocks"), "content": page.get("content", ""), "favorited": fav is not None, "share_mode": page.get("share_mode", "private"), "published": bool(page.get("is_published", 0) or page.get("published", 0)),
                  "is_locked": _locked,
                  "locked_by": _locked_by,
                  "can_edit": _can_edit,

app/routers/library.py

@@ -52,21 +52,18 @@ def _build_item(db_row: dict, uid: int = 1) -> dict:
     else:
         url = f"/pages/{page_id}"
 
-    # Icon
+    # Icon — convention Workspace (noms SVG, comme la sidebar) : la liste
+    # des favoris de la sidebar consomme item.icon via fdIconHtml.
     content_format = db_row.get("content_format", "blocks")
     if db_row.get("is_folder"):
-        icon = "📁"
+        icon = "folder"
     elif content_format == "file":
-        icon = "📄"
+        icon = "file"
         fn = title.lower()
-        if fn.endswith(".pdf"):
-            icon = "📕"
-        elif any(fn.endswith(e) for e in [".png", ".jpg", ".jpeg", ".gif", ".webp", ".svg"]):
-            icon = "🖼️"
-        elif any(fn.endswith(e) for e in [".py", ".js", ".ts", ".go", ".rs"]):
-            icon = "📜"
+        if any(fn.endswith(e) for e in [".png", ".jpg", ".jpeg", ".gif", ".webp", ".svg"]):
+            icon = "image"
     else:
-        icon = "📝"
+        icon = "edit"
 
     return {
         "id": page_id,
@@ -75,6 +72,8 @@ def _build_item(db_row: dict, uid: int = 1) -> dict:
         "page_icon": db_row.get("page_icon") or "",
         "is_folder": bool(db_row.get("is_folder", 0)),
         "source_type": source_type,
+        "source_icon": {"local": "file", "gitea": "link",
+                        "github": "external-link"}.get(source_type, "file"),
         "source_label": _source_label(source_type, workspace),
         "workspace": workspace,
         "workspace_name": _source_label(source_type, workspace),
@@ -104,7 +103,63 @@ def _apply_source_filter(query: str, params: list, source_type: str) -> tuple[st
 
 def _rows_to_items(rows, uid: int = 1) -> list:
     items = [_build_item(dict(r), uid) for r in rows]
-    return _attach_tags(items)
+    items = _attach_tags(items)
+    return _attach_provenance(items, uid)
+
+
+def _attach_provenance(items: list, uid: int = 1) -> list:
+    """Created by / Last edited by / Last visited (colonnes Library).
+
+    - created_by : premier auteur des versions, sinon propriétaire du workspace ;
+    - edited_by : dernier auteur des versions, sinon created_by ;
+    - visited_at : dernier accès tracé de l'utilisateur (table recents).
+    Requêtes groupées : 3 allers-retours quelle que soit la taille du lot.
+    """
+    if not items:
+        return items
+    ids = [it["id"] for it in items]
+    placeholders = ",".join("?" for _ in ids)
+    try:
+        with get_conn() as conn:
+            ver = conn.execute(
+                f"SELECT v.page_id, COALESCE(NULLIF(u.full_name,''), u.login, '') AS name "
+                f"FROM page_versions v LEFT JOIN users u ON u.id = v.user_id "
+                f"WHERE v.page_id IN ({placeholders}) ORDER BY v.id", ids,
+            ).fetchall()
+            first: dict = {}
+            last: dict = {}
+            for r in ver:
+                first.setdefault(r["page_id"], r["name"] or "")
+                last[r["page_id"]] = r["name"] or ""
+            own = conn.execute(
+                f"SELECT p.id, COALESCE(NULLIF(u.full_name,''), u.login, '') AS owner "
+                f"FROM pages p LEFT JOIN workspaces w ON w.id = p.workspace_id "
+                f"LEFT JOIN users u ON u.id = w.owner_id "
+                f"WHERE p.id IN ({placeholders})", ids,
+            ).fetchall()
+            owners = {r["id"]: r["owner"] or "" for r in own}
+            try:
+                vis = conn.execute(
+                    f"SELECT page_id, MAX(accessed_at) AS visited_at FROM recents "
+                    f"WHERE user_id = ? AND page_id IN ({placeholders}) "
+                    f"GROUP BY page_id", [uid, *ids],
+                ).fetchall()
+            except Exception:
+                vis = []
+            visited = {r["page_id"]: r["visited_at"] or "" for r in vis}
+    except Exception:
+        logger.exception("_attach_provenance")
+        return items
+    for it in items:
+        pid = it["id"]
+        author = first.get(pid) or owners.get(pid, "")
+        it["author"] = author
+        it["author_initial"] = (author or "?")[0].upper()
+        edited_by = last.get(pid) or author
+        it["edited_by"] = edited_by
+        it["edited_by_initial"] = (edited_by or "?")[0].upper()
+        it["visited_at"] = visited.get(pid, "")
+    return items
 
 
 def _attach_tags(items: list) -> list:
@@ -156,6 +211,59 @@ BASE_SELECT = (
 )
 
 
+# Lignes contenant un bloc AI Meeting Notes (formats JSON navigateur/Python).
+MEETING_BLOCK_LIKE = "REPLACE(COALESCE(p.content,''), ' ', '') LIKE '%\"type\":\"meeting\"%'"
+NO_MEETING_BLOCK = (
+    "REPLACE(COALESCE(p.content,''), ' ', '') NOT LIKE '%\"type\":\"meeting\"%'"
+)
+
+
+@router.get("/meetings")
+def library_meetings(
+    request: Request,
+    source_type: str = Query(default="all"),
+):
+    """Onglet Library « AI Meeting Notes » : notes de réunion de l'utilisateur.
+
+    Regroupées côté client par date (Today / Yesterday / …). Les notes sont
+    exclues des onglets de navigation courante (recents/workspace/private).
+    """
+    if source_type not in SOURCE_TYPES:
+        source_type = "all"
+
+    uid = _get_user_id(request)
+    query = (
+        f"{BASE_SELECT}, "
+        "(SELECT mt.status FROM meeting_transcripts mt WHERE mt.page_id = p.id "
+        " ORDER BY mt.id DESC LIMIT 1) AS meeting_status, "
+        "(SELECT mt.summary FROM meeting_transcripts mt WHERE mt.page_id = p.id "
+        " ORDER BY mt.id DESC LIMIT 1) AS meeting_summary "
+        "FROM pages p LEFT JOIN workspaces w ON w.id = p.workspace_id "
+        "WHERE p.deleted_at IS NULL AND (w.owner_id = ? OR p.workspace_id IS NULL) "
+        f"AND {MEETING_BLOCK_LIKE}"
+    )
+    params: list = [uid]
+    query, params = _apply_source_filter(query, params, source_type)
+    query += " ORDER BY p.updated_at DESC LIMIT 100"
+
+    with get_conn() as conn:
+        rows = conn.execute(query, params).fetchall()
+
+    items = _rows_to_items(rows, uid)
+    for it, r in zip(items, rows, strict=True):
+        d = dict(r)
+        st = d.get("meeting_status") or "idle"
+        it["is_meeting"] = True
+        it["meeting_status"] = st
+        it["source_label"] = (
+            "Summary ready" if (d.get("meeting_summary") or "").strip()
+            else ("Ready to record" if st == "idle"
+                  else st.replace("_", " ").capitalize())
+        )
+    _enrich_children(items)
+    return {"items": items}
+
+
 @router.get("/recents")
 def library_recents(
     request: Request,
@@ -167,7 +275,7 @@ def library_recents(
         source_type = "all"
 
     uid = _get_user_id(request)
-    query = f"{BASE_SELECT} FROM pages p WHERE p.parent_section != 'Trash' AND p.deleted_at IS NULL"
+    query = f"{BASE_SELECT} FROM pages p WHERE p.parent_section != 'Trash' AND p.deleted_at IS NULL AND {NO_MEETING_BLOCK}"
     params: list = []
 
     # Hierarchical view: show only root-level pages at the top
@@ -301,7 +409,8 @@ def library_published(
         source_type = "all"
 
     uid = _get_user_id(request)
-    query = f"{BASE_SELECT} FROM pages p WHERE p.published = 1 AND p.parent_section != 'Trash' AND p.deleted_at IS NULL"
+    # v7.69.6 : les deux drapeaux (lignes historiques issues d'un seul chemin).
+    query = f"{BASE_SELECT} FROM pages p WHERE (p.is_published = 1 OR p.published = 1) AND p.parent_section != 'Trash' AND p.deleted_at IS NULL"
     params: list = []
     if tree:
         query += " AND p.parent_id IS NULL"
@@ -326,7 +435,7 @@ def library_private(
         source_type = "all"
 
     uid = _get_user_id(request)
-    query = f"{BASE_SELECT} FROM pages p WHERE p.parent_section = 'Private' AND p.deleted_at IS NULL"
+    query = f"{BASE_SELECT} FROM pages p WHERE p.parent_section = 'Private' AND p.deleted_at IS NULL AND {NO_MEETING_BLOCK}"
     params: list = []
     if tree:
         query += " AND p.parent_id IS NULL"
@@ -392,7 +501,7 @@ def library_workspace(
         source_type = "all"
 
     uid = _get_user_id(request)
-    query = f"{BASE_SELECT} FROM pages p WHERE (p.workspace != '' OR p.workspace_id IS NOT NULL) AND p.parent_section != 'Trash' AND p.deleted_at IS NULL"
+    query = f"{BASE_SELECT} FROM pages p WHERE (p.workspace != '' OR p.workspace_id IS NOT NULL) AND p.parent_section != 'Trash' AND p.deleted_at IS NULL AND {NO_MEETING_BLOCK}"
     params: list = []
 
     if tree:

app/routers/meetings.py

@@ -185,6 +185,41 @@ def upcoming_meetings(request: Request, days: int = 30):
 
 # ── meetings ───────────────────────────────────────────────────────────────
 
+@router.post("/api/v2/meetings/transcripts")
+def create_transcript(request: Request, body: dict = Body(default={})):
+    """Crée une note de réunion vide (état idle) pour une page — §6.5.
+
+    Idempotent par occurrence calendrier : renvoie la note existante.
+    """
+    user = _auth_user(request, require_write=True)
+    try:
+        page_id = int(body.get("page_id", 0))
+    except (TypeError, ValueError):
+        raise HTTPException(400, "page_id required") from None
+    occ = (body.get("event_occurrence_id") or "").strip()[:200]
+    with get_conn() as conn:
+        if occ:
+            row = conn.execute(
+                "SELECT * FROM meeting_transcripts WHERE event_occurrence_id=?",
+                (occ,)).fetchone()
+            if row:
+                return {**row_to_dict(row), "already_existed": True}
+    try:
+        tid = meet.save_transcript(
+            page_id, "", (body.get("language") or "fr")[:10],
+            mode=(body.get("mode") or "mic_only")
+            if (body.get("mode") or "mic_only") in meet.CAPTURE_MODES else "mic_only",
+            instruction_id=(body.get("instruction_id") or "auto")[:40],
+            event_occurrence_id=occ)
+    except ValueError as exc:
+        msg = str(exc)
+        raise HTTPException(404 if "not found" in msg else 400, msg) from None
+    with get_conn() as conn:
+        row = conn.execute("SELECT * FROM meeting_transcripts WHERE id=?", (tid,)).fetchone()
+    audit_log(user, "meeting.create", "page", page_id, f"transcript={tid}", request)
+    return JSONResponse(status_code=201,
+                        content={**row_to_dict(row), "already_existed": False})
+
 @router.post("/api/v2/meetings/transcribe")
 async def upload_and_transcribe(request: Request):
     user = _auth_user(request, require_write=True)
@@ -254,13 +289,305 @@ def set_transcript_text(transcript_id: int, request: Request, body: dict = Body(
 
 
 @router.post("/api/v2/meetings/transcripts/{transcript_id}/summarize")
-async def summarize(transcript_id: int, request: Request):
+async def summarize(transcript_id: int, request: Request,
+                    body: dict = Body(default={})):
     user = _auth_user(request, require_write=True)
     try:
-        out = await meet.summarize_transcript(transcript_id, user.get("id"))
+        out = await meet.summarize_transcript(
+            transcript_id, user.get("id"),
+            allow_empty=bool((body or {}).get("allow_empty")))
     except ValueError as exc:
         raise HTTPException(400, str(exc)) from None
     except RuntimeError as exc:
         raise HTTPException(502, str(exc)) from None
     audit_log(user, "meeting.summarize", "transcript", transcript_id, "", request)
     return out
+
+
+# ── Bloc Notion v7.1.1 : machine à états, consentement, segments, partage ───
+
+@router.get("/api/v2/meetings/instructions")
+def meeting_instructions(request: Request):
+    _auth_user(request)
+    return {"instructions": [
+        {"id": k, "label": v["label"]} for k, v in meet.INSTRUCTIONS.items()
+    ]}
+
+
+# Une page est une « note de réunion » dès qu'elle contient un bloc
+# AI Meeting Notes. Le contenu est du JSON (``"type":"meeting"`` côté
+# navigateur, ``"type": "meeting"`` côté Python) → comparaison sans espaces.
+MEETING_BLOCK_PREDICATE = (
+    "REPLACE(COALESCE({col},''), ' ', '') LIKE '%\"type\":\"meeting\"%'"
+)
+
+
+@router.get("/api/v2/meetings/notes")
+def meeting_notes(request: Request, limit: int = 50):
+    """Notes de réunion : pages contenant un bloc AI Meeting Notes.
+
+    Alimente la section Meetings de la sidebar (ces pages sont exclues de
+    l'arbre Workspace, voir ``_load_workspace_pages``). Inclut les notes
+    jamais enregistrées (aucune ligne ``meeting_transcripts``) avec
+    ``status = idle``.
+    """
+    user = _auth_user(request)
+    uid = user.get("id")
+    lim = max(1, min(int(limit or 50), 200))
+    pred = MEETING_BLOCK_PREDICATE.format(col="p.content")
+    with get_conn() as conn:
+        rows = conn.execute(
+            f"""SELECT p.id, p.title, p.updated_at,
+                      (SELECT mt.status FROM meeting_transcripts mt
+                        WHERE mt.page_id = p.id ORDER BY mt.id DESC LIMIT 1) AS status,
+                      (SELECT mt.summary FROM meeting_transcripts mt
+                        WHERE mt.page_id = p.id ORDER BY mt.id DESC LIMIT 1) AS summary
+               FROM pages p
+               LEFT JOIN workspaces w ON w.id = p.workspace_id
+               WHERE p.deleted_at IS NULL
+                 AND (w.owner_id = ? OR p.workspace_id IS NULL)
+                 AND {pred}
+               ORDER BY p.updated_at DESC LIMIT ?""",
+            (uid, lim),
+        ).fetchall()
+    notes = []
+    for r in rows:
+        d = row_to_dict(r)
+        notes.append({
+            "id": d["id"],
+            "title": d.get("title") or "Untitled",
+            "updated_at": d.get("updated_at") or "",
+            "status": d.get("status") or "idle",
+            "has_summary": bool((d.get("summary") or "").strip()),
+        })
+    return {"notes": notes, "count": len(notes)}
+
+
+@router.get("/api/v2/meetings/history")
+def meetings_history(request: Request, query: str = "", limit: int = 50):
+    """Historique des notes de réunion (passé + recherche plein texte)."""
+    user = _auth_user(request)
+    _ = user
+    like = f"%{query.strip()}%" if query.strip() else "%"
+    lim = max(1, min(int(limit or 50), 200))
+    with get_conn() as conn:
+        rows = conn.execute(
+            """SELECT mt.*, p.title AS page_title FROM meeting_transcripts mt
+               JOIN pages p ON p.id = mt.page_id
+               WHERE (p.title LIKE ? OR mt.transcript LIKE ?
+                      OR mt.summary LIKE ? OR mt.generated_title LIKE ?)
+               ORDER BY mt.id DESC LIMIT ?""",
+            (like, like, like, like, lim),
+        ).fetchall()
+    out = []
+    for r in rows:
+        d = row_to_dict(r)
+        d["title"] = d.get("generated_title") or d.get("page_title") or "Meeting"
+        out.append(d)
+    return {"meetings": out, "count": len(out)}
+
+
+@router.get("/api/v2/meetings/transcripts/{transcript_id}")
+def get_transcript_row(transcript_id: int, request: Request):
+    _auth_user(request)
+    try:
+        return meet.get_transcript(transcript_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from None
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/consents")
+def post_consent(transcript_id: int, request: Request, body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        entry = meet.record_consent(
+            transcript_id, method=(body.get("method") or "start_attestation"),
+            message_ref=(body.get("message_ref") or "")[:200],
+            attested_by=user.get("id"),
+            participants=body.get("participants") or [])
+    except ValueError as exc:
+        msg = str(exc)
+        raise HTTPException(404 if "not found" in msg else 400, msg) from None
+    audit_log(user, "meeting.consent", "transcript", transcript_id,
+              entry["method"], request)
+    return entry
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/sessions")
+def start_session(transcript_id: int, request: Request, body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        out = meet.start_session(
+            transcript_id, mode=(body.get("mode") or "mic_only"),
+            channels=body.get("channels"), language=(body.get("language") or "fr"),
+            instruction_id=(body.get("instruction_id") or "auto"),
+            user_id=user.get("id"))
+    except ValueError as exc:
+        msg = str(exc)
+        raise HTTPException(404 if "not found" in msg else 400, msg) from None
+    audit_log(user, "meeting.session.start", "transcript", transcript_id,
+              out.get("mode", ""), request)
+    return out
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/pause")
+def pause_session(transcript_id: int, request: Request, body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        out = meet.pause_session(transcript_id,
+                                 paused_ms_delta=int(body.get("paused_ms") or 0))
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+    audit_log(user, "meeting.session.pause", "transcript", transcript_id, "", request)
+    return out
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/resume")
+def resume_session(transcript_id: int, request: Request):
+    user = _auth_user(request, require_write=True)
+    try:
+        out = meet.resume_session(transcript_id)
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+    audit_log(user, "meeting.session.resume", "transcript", transcript_id, "", request)
+    return out
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/stop")
+async def stop_session(transcript_id: int, request: Request,
+                       body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        stopped = meet.stop_session(transcript_id,
+                                    duration_ms=int((body or {}).get("duration_ms") or 0))
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+    audit_log(user, "meeting.session.stop", "transcript", transcript_id, "", request)
+    # Le traitement démarre aussitôt (étapes Thinking persistées) ; en cas
+    # d'échec LLM le transcript reste consultable (état failed, retry).
+    # allow_empty : arrêt sans rien de capté → diagnostic honnête (§12.8),
+    # jamais une ligne coincée en « processing ».
+    try:
+        out = await meet.run_processing(
+            transcript_id, user.get("id"),
+            allow_empty=bool((body or {}).get("allow_empty")))
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+    except RuntimeError as exc:
+        raise HTTPException(502, str(exc)) from None
+    return {"session": {"id": stopped["id"], "status": "done"}, **out}
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/segments")
+def append_segments(transcript_id: int, request: Request,
+                    body: dict = Body(default={})):
+    _auth_user(request, require_write=True)
+    try:
+        return meet.append_segments(transcript_id, body.get("segments") or [])
+    except ValueError as exc:
+        msg = str(exc)
+        raise HTTPException(404 if "not found" in msg else 400, msg) from None
+
+
+@router.get("/api/v2/meetings/transcripts/{transcript_id}/transcript")
+def grouped_transcript(transcript_id: int, request: Request):
+    _auth_user(request)
+    try:
+        return meet.grouped_transcript(transcript_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from None
+
+
+@router.patch("/api/v2/meetings/transcripts/{transcript_id}/segments/{seq}/speaker")
+def patch_speaker(transcript_id: int, seq: int, request: Request,
+                  body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        return meet.set_segment_speaker(transcript_id, seq,
+                                        body.get("speaker") or "",
+                                        user.get("id"))
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+
+
+@router.get("/api/v2/meetings/transcripts/{transcript_id}/processing-steps")
+def processing_steps(transcript_id: int, request: Request):
+    _auth_user(request)
+    try:
+        tr = meet.get_transcript(transcript_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from None
+    import json as _json
+    try:
+        steps = _json.loads(tr.get("processing_steps_json") or "[]")
+    except (TypeError, ValueError):
+        steps = []
+    return {"transcript_id": transcript_id, "status": tr.get("status") or "idle",
+            "steps": steps}
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/processing:retry")
+async def retry_processing(transcript_id: int, request: Request,
+                           body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        tr = meet.get_transcript(transcript_id)
+    except ValueError as exc:
+        raise HTTPException(404, str(exc)) from None
+    if tr.get("status") not in ("failed", "processing", "done"):
+        raise HTTPException(400, "nothing to retry — no failed processing")
+    try:
+        return await meet.run_processing(
+            transcript_id, user.get("id"),
+            allow_empty=bool((body or {}).get("allow_empty")))
+    except ValueError as exc:
+        raise HTTPException(400, str(exc)) from None
+    except RuntimeError as exc:
+        raise HTTPException(502, str(exc)) from None
+
+
+@router.patch("/api/v2/meetings/transcripts/{transcript_id}/generated-title")
+def patch_title(transcript_id: int, request: Request, body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    title = (body.get("title") or "").strip()[:200]
+    if not title:
+        raise HTTPException(400, "title required")
+    with get_conn() as conn:
+        if not conn.execute("SELECT id FROM meeting_transcripts WHERE id=?",
+                            (transcript_id,)).fetchone():
+            raise HTTPException(404, "Transcript not found")
+        conn.execute("UPDATE meeting_transcripts SET generated_title=?,"
+                     " title_source='user' WHERE id=?", (title, transcript_id))
+        conn.commit()
+    audit_log(user, "meeting.retitle", "transcript", transcript_id, title, request)
+    return {"transcript_id": transcript_id, "generated_title": title}
+
+
+@router.post("/api/v2/meetings/transcripts/{transcript_id}/distributions")
+def post_distribution(transcript_id: int, request: Request,
+                      body: dict = Body(default={})):
+    user = _auth_user(request, require_write=True)
+    try:
+        out = meet.record_distribution(
+            transcript_id, channel=(body.get("channel") or "copy_link"),
+            actor_id=user.get("id"), recipients=(body.get("recipients") or "")[:500],
+            status=body.get("status") or "created")
+    except ValueError as exc:
+        msg = str(exc)
+        raise HTTPException(404 if "not found" in msg else 400, msg) from None
+    audit_log(user, "meeting.share", "transcript", transcript_id,
+              out["channel"], request)
+    return out
+
+
+@router.patch("/api/v2/meetings/transcripts/{transcript_id}/notes")
+def patch_notes(transcript_id: int, request: Request, body: dict = Body(default={})):
+    _auth_user(request, require_write=True)
+    with get_conn() as conn:
+        if not conn.execute("SELECT id FROM meeting_transcripts WHERE id=?",
+                            (transcript_id,)).fetchone():
+            raise HTTPException(404, "Transcript not found")
+        conn.execute("UPDATE meeting_transcripts SET notes_text=? WHERE id=?",
+                     ((body.get("notes") or "")[:50000], transcript_id))
+        conn.commit()
+    return {"transcript_id": transcript_id, "saved": True}

app/services/export.py

@@ -66,28 +66,226 @@ def _block_text(b: dict) -> str:
 # ── Source resolution: read real textual content for ANY page type ──
 
 # Extensions whose content is plain text / code / markdown (textual exportable).
+# Mirrors static/js/code-highlight.js EXT_TO_LANG (single source of truth for
+# the viewer lives client-side; this set gates server-side text handling:
+# file viewer, copy-to-clipboard, exports, imports).
 _TEXTUAL_EXTS = {
-    "md", "markdown", "txt", "log", "text",
-    "py", "js", "ts", "jsx", "tsx", "html", "htm", "css", "json", "xml",
-    "yaml", "yml", "toml", "ini", "cfg", "conf", "env", "sh", "bash", "zsh",
-    "ps1", "bat", "cmd", "rb", "go", "rs", "java", "c", "cpp", "h", "hpp",
-    "php", "swift", "kt", "scala", "sql", "r", "vue", "svelte", "astro",
-    "properties", "gitignore", "dockerfile", "makefile",
+    "md", "markdown", "mkd", "mdown", "mdwn", "rmd", "qmd",
+    "txt", "log", "text", "csv", "tsv", "tab", "nfo", "asc",
+    "py", "pyw", "pyi", "gyp", "gypi",
+    "js", "mjs", "cjs", "jsx", "ts", "mts", "cts", "tsx",
+    "html", "htm", "xhtml", "xht", "xml", "xsd", "xsl", "xslt",
+    "rss", "atom", "plist", "wsf", "svg", "axml",
+    "vue", "svelte", "astro",
+    "css", "scss", "sass", "less",
+    "json", "json5", "jsonc", "geojson", "webmanifest", "har", "avsc",
+    "yaml", "yml", "toml", "ini", "cfg", "conf", "env", "properties",
+    "inf", "reg", "url", "prop",
+    "gitconfig", "gitmodules", "editorconfig",
+    "sh", "bash", "zsh", "ksh", "mksh", "yash", "fish",
+    "ps1", "psm1", "psd1", "bat", "cmd",
+    "rb", "gemspec", "rake", "rbw",
+    "go", "rs", "java", "jsp", "kt", "kts",
+    "scala", "sc", "groovy", "gvy", "gradle",
+    "c", "cpp", "cc", "cxx", "c++", "hh", "hxx", "hcc", "h", "hpp",
+    "cs", "csx", "cake", "ino", "pde",
+    "php", "phtml", "swift", "sql", "pgsql", "psql", "mysql", "mssql",
+    "r", "jl", "lua", "dart",
+    "pl", "pm", "pod",
+    "hs", "lhs", "ex", "exs", "eex", "heex", "leex",
+    "erl", "hrl", "escript",
+    "clj", "cljs", "cljc", "edn",
+    "ml", "mli", "fs", "fsi", "fsx", "fsscript",
+    "nim", "nims", "d", "di",
+    "m", "mm", "vb", "vbs", "bas", "pas", "pp", "dpr",
+    "vhd", "vhdl", "v", "vh", "sv", "svh",
+    "asm", "s", "nasm", "inc",
+    "ada", "ads", "adb",
+    "f", "for", "f90", "f95", "f03", "f08",
+    "cob", "cbl", "cpy",
+    "dockerfile", "containerfile", "tf", "tfvars", "hcl",
+    "cmake", "makefile", "mk", "mak", "ninja",
+    "gql", "graphql", "graphqls", "proto", "thrift",
+    "vim", "vimrc", "gvimrc", "awk", "gcode", "gco", "nc", "ngc",
+    "http", "rest", "nginx", "apache", "htaccess", "wat",
+    "diff", "patch", "rej",
+    "tex", "latex", "sty", "cls", "dtx", "bib", "bbl",
+    "gitignore", "gitattributes", "gitkeep", "dockerignore", "npmignore",
+    "lock", "mod", "sum",
+}
+# Well-known extensionless / dotfile basenames that are textual code.
+_TEXTUAL_FILENAMES = {
+    "dockerfile", "containerfile", "makefile", "gnumakefile",
+    "cmakelists.txt", "rakefile", "gemfile", "vagrantfile",
+    "berksfile", "brewfile", "dangerfile", "fastfile", "podfile",
+    "go.mod", "go.sum",
+    ".bashrc", ".bash_profile", ".bash_login", ".bash_logout",
+    ".zshrc", ".zshenv", ".profile", ".shrc", ".kshrc",
+    ".vimrc", ".gvimrc", ".viminfo",
+    ".gitconfig", ".editorconfig", ".htaccess",
+}
+# Exact basenames → highlight.js id (checked before the extension).
+_FILENAME_LANG = {
+    "dockerfile": "dockerfile", "containerfile": "dockerfile",
+    "makefile": "makefile", "gnumakefile": "makefile",
+    "cmakelists.txt": "cmake",
+    "rakefile": "ruby", "gemfile": "ruby", "vagrantfile": "ruby",
+    "berksfile": "ruby", "brewfile": "ruby", "dangerfile": "ruby",
+    "fastfile": "ruby", "podfile": "ruby",
+    "go.mod": "go",
+    "nginx.conf": "nginx",
+    ".bashrc": "bash", ".bash_profile": "bash", ".bash_login": "bash",
+    ".bash_logout": "bash", ".zshrc": "bash", ".zshenv": "bash",
+    ".profile": "bash", ".shrc": "bash", ".kshrc": "bash",
+    ".vimrc": "vim", ".gvimrc": "vim",
+    ".gitconfig": "ini", ".editorconfig": "ini", ".htaccess": "apache",
 }
 _CODE_LANG = {
-    "py": "python", "js": "javascript", "ts": "typescript", "jsx": "javascript",
-    "tsx": "typescript", "html": "html", "htm": "html", "css": "css",
-    "json": "json", "xml": "xml", "yaml": "yaml", "yml": "yaml",
-    "toml": "toml", "ini": "ini", "cfg": "ini", "conf": "ini", "env": "ini",
-    "sh": "bash", "bash": "bash", "zsh": "bash", "ps1": "powershell",
-    "bat": "batch", "cmd": "batch", "rb": "ruby", "go": "go", "rs": "rust",
-    "java": "java", "c": "c", "cpp": "cpp", "h": "c", "hpp": "cpp",
-    "php": "php", "swift": "swift", "kt": "kotlin", "scala": "scala",
-    "sql": "sql", "r": "r", "vue": "html", "svelte": "html",
-    "astro": "html", "properties": "ini", "md": "markdown",
-    "markdown": "markdown", "txt": "plaintext", "log": "plaintext",
-    "text": "plaintext",
+    "py": "python", "pyw": "python", "pyi": "python",
+    "gyp": "python", "gypi": "python",
+    "js": "javascript", "mjs": "javascript", "cjs": "javascript",
+    "jsx": "javascript",
+    "ts": "typescript", "mts": "typescript", "cts": "typescript",
+    "tsx": "typescript",
+    "html": "xml", "htm": "xml", "xhtml": "xml", "xht": "xml",
+    "xml": "xml", "xsd": "xml", "xsl": "xml", "xslt": "xml",
+    "rss": "xml", "atom": "xml", "plist": "xml", "wsf": "xml",
+    "svg": "xml", "axml": "xml",
+    "vue": "xml", "svelte": "xml", "astro": "xml",
+    "css": "css", "scss": "scss", "sass": "scss", "less": "less",
+    "json": "json", "json5": "json", "jsonc": "json",
+    "geojson": "json", "webmanifest": "json", "har": "json", "avsc": "json",
+    "yaml": "yaml", "yml": "yaml",
+    "toml": "ini", "ini": "ini", "cfg": "ini", "conf": "ini",
+    "env": "ini", "properties": "ini", "prop": "ini",
+    "inf": "ini", "reg": "ini", "url": "ini",
+    "gitconfig": "ini", "gitmodules": "ini", "editorconfig": "ini",
+    "ninja": "ini",
+    "md": "markdown", "markdown": "markdown", "mkd": "markdown",
+    "mdown": "markdown", "mdwn": "markdown", "rmd": "markdown",
+    "qmd": "markdown",
+    "tex": "latex", "latex": "latex", "sty": "latex", "cls": "latex",
+    "dtx": "latex",
+    "sql": "sql", "pgsql": "sql", "psql": "sql", "mysql": "sql",
+    "mssql": "sql",
+    "sh": "bash", "bash": "bash", "zsh": "bash", "ksh": "bash",
+    "mksh": "bash", "yash": "bash",
+    "fish": "shell",
+    "ps1": "powershell", "psm1": "powershell", "psd1": "powershell",
+    "bat": "dos", "cmd": "dos",
+    "rb": "ruby", "gemspec": "ruby", "rake": "ruby", "rbw": "ruby",
+    "go": "go", "rs": "rust",
+    "java": "java", "jsp": "java",
+    "kt": "kotlin", "kts": "kotlin",
+    "scala": "scala", "sc": "scala",
+    "groovy": "groovy", "gvy": "groovy", "gradle": "groovy",
+    "c": "c", "h": "c",
+    "cpp": "cpp", "cc": "cpp", "cxx": "cpp", "c++": "cpp",
+    "hh": "cpp", "hxx": "cpp", "hcc": "cpp", "hpp": "cpp",
+    "ino": "cpp", "pde": "cpp",
+    "cs": "csharp", "csx": "csharp", "cake": "csharp",
+    "php": "php", "phtml": "php-template",
+    "swift": "swift", "r": "r", "jl": "julia", "lua": "lua",
+    "dart": "dart", "pl": "perl", "pm": "perl", "pod": "perl",
+    "hs": "haskell", "lhs": "haskell",
+    "ex": "elixir", "exs": "elixir", "eex": "elixir",
+    "heex": "elixir", "leex": "elixir",
+    "erl": "erlang", "hrl": "erlang", "escript": "erlang",
+    "clj": "clojure", "cljs": "clojure", "cljc": "clojure", "edn": "clojure",
+    "ml": "ocaml", "mli": "ocaml",
+    "fs": "fsharp", "fsi": "fsharp", "fsx": "fsharp", "fsscript": "fsharp",
+    "nim": "nim", "nims": "nim",
+    "d": "d", "di": "d",
+    "m": "objectivec", "mm": "objectivec",
+    "vb": "vbnet", "vbs": "vbnet", "bas": "vbnet",
+    "pas": "delphi", "pp": "delphi", "dpr": "delphi",
+    "vhd": "vhdl", "vhdl": "vhdl",
+    "v": "verilog", "vh": "verilog", "sv": "verilog", "svh": "verilog",
+    "asm": "x86asm", "s": "x86asm", "nasm": "x86asm", "inc": "x86asm",
+    "ada": "ada", "ads": "ada", "adb": "ada",
+    "f": "fortran", "for": "fortran", "f90": "fortran", "f95": "fortran",
+    "f03": "fortran", "f08": "fortran",
+    "cob": "plaintext", "cbl": "plaintext", "cpy": "plaintext",
+    "dockerfile": "dockerfile",
+    "tf": "plaintext", "tfvars": "plaintext", "hcl": "plaintext",
+    "cmake": "cmake",
+    "makefile": "makefile", "mk": "makefile", "mak": "makefile",
+    "gql": "graphql", "graphql": "graphql", "graphqls": "graphql",
+    "proto": "protobuf", "thrift": "thrift",
+    "vim": "vim", "vimrc": "vim", "gvimrc": "vim",
+    "awk": "awk",
+    "gcode": "gcode", "gco": "gcode", "nc": "gcode", "ngc": "gcode",
+    "http": "http", "rest": "http",
+    "nginx": "nginx", "apache": "apache", "htaccess": "apache",
+    "wat": "wasm",
+    "diff": "diff", "patch": "diff", "rej": "diff",
+    "bib": "plaintext", "bbl": "plaintext",
+    "txt": "plaintext", "log": "plaintext", "text": "plaintext",
+    "csv": "plaintext", "tsv": "plaintext", "tab": "plaintext",
+    "nfo": "plaintext", "asc": "plaintext",
+    "lock": "plaintext",
+    "gitignore": "plaintext", "gitattributes": "plaintext",
+    "gitkeep": "plaintext", "dockerignore": "plaintext",
+    "npmignore": "plaintext",
 }
+# Aliases users may type in markdown fences / old block labels.
+_CODE_ALIASES = {
+    "py": "python", "js": "javascript", "jsx": "javascript",
+    "ts": "typescript", "tsx": "typescript",
+    "html": "xml", "htm": "xml", "markup": "xml",
+    "yml": "yaml", "md": "markdown",
+    "sh": "bash", "zsh": "bash", "console": "shell",
+    "ps1": "powershell", "ps": "powershell",
+    "bat": "dos", "cmd": "dos", "batch": "dos",
+    "c++": "cpp", "c#": "csharp", "cs": "csharp",
+    "kt": "kotlin", "rb": "ruby", "rs": "rust", "pl": "perl",
+    "toml": "ini", "cfg": "ini", "conf": "ini", "env": "ini",
+    "tex": "latex", "docker": "dockerfile", "make": "makefile",
+    "plain text": "plaintext", "plain": "plaintext", "text": "plaintext",
+    "none": "plaintext", "": "plaintext",
+}
+# highlight.js ids served by static/js/highlight.min.js + highlight-extra.min.js.
+_HLJS_LANGS = frozenset({
+    "ada", "apache", "armasm", "awk", "bash", "c", "clojure", "cmake", "cpp",
+    "csharp", "css", "d", "dart", "delphi", "diff", "dockerfile", "dos",
+    "elixir", "erlang", "fsharp", "fortran", "gcode", "go",
+    "graphql", "groovy", "haskell", "http", "ini", "java", "javascript",
+    "json", "julia", "kotlin", "latex", "less", "lua", "makefile",
+    "markdown", "matlab", "nginx", "nim", "objectivec", "ocaml", "perl",
+    "php", "php-template", "plaintext", "powershell", "protobuf", "python",
+    "python-repl", "r", "ruby", "rust", "scala", "scss", "shell", "sql",
+    "stan", "stata", "swift", "thrift", "typescript", "vbnet", "verilog",
+    "vhdl", "vim", "wasm", "x86asm", "xml", "yaml",
+})
+
+
+def normalize_code_lang(lang: str | None) -> str:
+    """Normalize any code language label to a bundled highlight.js id.
+
+    Accepts highlight.js ids, file extensions, markdown fence names and the
+    legacy ``Plain Text`` label. Unknown values degrade to ``plaintext``.
+    """
+    lang = (lang or "").strip().lower()
+    if not lang or lang in ("plain text", "none"):
+        return "plaintext"
+    if lang in _CODE_ALIASES:
+        lang = _CODE_ALIASES[lang]
+    if lang in _CODE_LANG.values() or lang in _HLJS_LANGS:
+        return lang
+    return "plaintext"
+
+
+def code_lang_for_file(filename: str | None) -> str:
+    """Best highlight.js id for an uploaded / repo file name."""
+    name = (filename or "").replace("\\", "/").rsplit("/", 1)[-1].lower()
+    if not name:
+        return "plaintext"
+    if name in _FILENAME_LANG:
+        return _FILENAME_LANG[name]
+    ext = name.rsplit(".", 1)[-1] if "." in name else ""
+    return _CODE_LANG.get(ext, "plaintext")
+
+
 _MARKDOWN_MIMES = {"text/markdown", "text/x-markdown", "application/octet-stream"}
 
 
@@ -119,7 +317,8 @@ def _file_text(page: dict) -> str | None:
     name = (rel.rsplit("/", 1)[-1] or "").lower()
     ext = name.rsplit(".", 1)[-1] if "." in name else ""
     mime = (meta.get("mime_type") or "").lower()
-    if not (ext in _TEXTUAL_EXTS or mime.startswith("text/")):
+    if not (ext in _TEXTUAL_EXTS or name in _TEXTUAL_FILENAMES
+            or mime.startswith("text/")):
         return None
     try:
         full = (_data_root() / rel).resolve()
@@ -161,8 +360,7 @@ def _page_source(page: dict):
         mime = (meta.get("mime_type") or "").lower()
         if ext in ("md", "markdown") or mime in _MARKDOWN_MIMES or mime.startswith("text/markdown"):
             return "md", text
-        lang = _CODE_LANG.get(ext, "plaintext")
-        return "code", (text, lang)
+        return "code", (text, code_lang_for_file(name))
 
     # Unknown format (e.g. legacy) -> try to dump as raw text
     if content.strip():
@@ -497,7 +695,7 @@ def blocks_to_markdown(blocks: list) -> str:
             out.append("---")
         elif t == "code":
             lang = b.get("language") or ""
-            out.append(f"```{lang}\n{c}\n```")
+            out.append(f"```{normalize_code_lang(lang) if lang else ''}\n{c}\n```")
         elif t == "toggle":
             out.append(f"### {c}")
             if b.get("children"):
@@ -615,8 +813,11 @@ def blocks_to_html(blocks: list) -> str:
             parts.append("<hr>")
         elif t == "code":
             lang = b.get("language") or ""
+            hljs_lang = normalize_code_lang(lang)
             label = f'<div class="code-lang">{_text(lang)}</div>' if lang else ""
-            parts.append(f"<pre>{label}<code>{c}</code></pre>")
+            parts.append(
+                f'<pre>{label}<code class="language-{hljs_lang}">{c}</code></pre>'
+            )
         elif t == "math":
             parts.append(f'<div class="math">\\[{c}\\]</div>')
         elif t == "table_of_contents":

app/services/importers/forge_repo.py

@@ -7,7 +7,11 @@ from __future__ import annotations
 
 from pathlib import Path
 
-from app.services.export import _CODE_LANG, _TEXTUAL_EXTS
+from app.services.export import (
+    _TEXTUAL_EXTS,
+    _TEXTUAL_FILENAMES,
+    code_lang_for_file,
+)
 from app.services.importers.base import ImportPage, ImportResult
 
 _MD_EXTS = {"md", "markdown"}
@@ -17,6 +21,14 @@ def _ext(path: str) -> str:
     return Path(path).suffix.lower().lstrip(".")
 
 
+def _basename(path: str) -> str:
+    return path.replace("\\", "/").rsplit("/", 1)[-1].lower()
+
+
+def _is_textual(path: str) -> bool:
+    return _ext(path) in _TEXTUAL_EXTS or _basename(path) in _TEXTUAL_FILENAMES
+
+
 def build_repo_result(
     files: list[tuple[str, str]],
     *,
@@ -34,7 +46,7 @@ def build_repo_result(
         if ext in _MD_EXTS:
             markdown = content
         else:
-            lang = _CODE_LANG.get(ext, "")
+            lang = code_lang_for_file(clean)
             markdown = f"```{lang}\n{content.rstrip()}\n```"
         parts = clean.split("/")
         result.pages.append(ImportPage(
@@ -69,7 +81,7 @@ async def fetch_forge_repo(
     files: list[tuple[str, str]] = []
     for meta in metas:
         file_path = meta.get("path") or ""
-        if _ext(file_path) not in _TEXTUAL_EXTS:
+        if not _is_textual(file_path):
             continue
         if int(meta.get("size") or 0) > max_file_bytes:
             continue

app/services/meetings.py

@@ -1,19 +1,25 @@
-"""FlowDeck — AI Meeting Notes v2 (v7.1.0).
+"""FlowDeck — AI Meeting Notes façon Notion (v7.1.1).
 
-Upload audio → optional server transcription (``STT_COMMAND``, e.g. whisper)
-→ AI summary (``AIWritingService.summarize``, offline-capable) → fires
-``meeting.summarized`` so custom agents pick it up (Notion 07/2026 pattern).
+Refonte alignée sur ``docs/architecture-meeting-notion-flowdeck.md`` (§3A/3B) :
+machine à états du bloc, révélation progressive des onglets, consentement
+bloquant journalisé, segments groupés par source, étapes Thinking persistées,
+résumé structuré avec citations obligatoires, classification de complétude,
+titre généré (diagnostic honnête si bref/incomplet), barre de partage tracée.
 
-Without ``STT_COMMAND`` the server stores the audio and accepts a manual
-``transcript`` (client-side transcription). Nothing here requires new pip
-dependencies. See ``docs/V71_Calendar_Meetings.md``.
+Stockage SQLite monolithique : ``meeting_transcripts`` + colonnes JSON
+(migration 38) + ``meeting_consent_log`` / ``meeting_distributions``
+append-only. Rien n'exige de dépendance pip supplémentaire.
 """
 from __future__ import annotations
 
+import json
 import logging
 import os
+import re
 import shutil
 import subprocess
+import time
+from datetime import UTC, datetime
 from pathlib import Path
 
 from app.config import settings
@@ -24,6 +30,37 @@ logger = logging.getLogger(__name__)
 AUDIO_EXTENSIONS = {"mp3", "wav", "m4a", "ogg", "flac", "aac"}
 MAX_AUDIO_BYTES = 100 * 1024 * 1024
 
+# États du bloc (§6.5) — transitions toujours via le serveur.
+BLOCK_STATES = ("idle", "recording", "paused", "processing", "done", "failed")
+CAPTURE_MODES = ("mic_only", "tab_plus_mic", "import")
+
+# Instructions de résumé (§12.4) : auto + presets par type + custom libre.
+INSTRUCTIONS = {
+    "auto": {"label": "Auto", "prompt": "Résumé automatique adapté au contenu détecté."},
+    "standup": {"label": "Standup", "prompt": "Format standup : hier / aujourd'hui / blocages, par personne."},
+    "team": {"label": "Team meeting", "prompt": "Format équipe : sujets, décisions, actions avec responsables."},
+    "sales": {"label": "Sales call", "prompt": "Format commercial : besoin, objections, next steps, montant."},
+    "one_to_one": {"label": "1:1", "prompt": "Format 1:1 : points du collaborateur, feedback, engagements."},
+    "interview": {"label": "Interview", "prompt": "Format entretien : parcours, forces, doutes, décision."},
+}
+
+# Étapes Thinking exposées (§12.7), dans l'ordre — libellés FR/EN gérés côté UI.
+PROCESSING_STEPS = (
+    ("reading_transcript", "Reading transcript…"),
+    ("analyzing_transcript", "Analyzing the transcript"),
+    ("resolving_references", "Thinking about how to refer to people"),
+    ("understanding_content", "Understanding the content"),
+    ("classifying_completeness", "Checking completeness"),
+    ("generating_summary", "Writing the summary"),
+    ("generating_title", "Choosing a title"),
+    ("validating_citations", "Checking citations"),
+)
+
+CONSENT_METHODS = (
+    "start_attestation", "verbal", "chat_text", "audio_message",
+    "meet_addon", "workspace_enforced",
+)
+
 
 class TranscriptionUnavailable(RuntimeError):
     """Raised when no transcription backend is configured."""
@@ -36,26 +73,609 @@ def meetings_dir() -> Path:
     return d
 
 
+def _now() -> str:
+    return datetime.now(UTC).strftime("%Y-%m-%d %H:%M:%S")
+
+
+def _loads(raw: str, default):
+    try:
+        v = json.loads(raw or "")
+        return v if v is not None else default
+    except (TypeError, ValueError):
+        return default
+
+
+def get_transcript(transcript_id: int) -> dict:
+    with get_conn() as conn:
+        row = conn.execute(
+            "SELECT * FROM meeting_transcripts WHERE id=?", (transcript_id,)
+        ).fetchone()
+    if not row:
+        raise ValueError("transcript not found")
+    return dict(row)
+
+
+def _update(transcript_id: int, **fields) -> None:
+    if not fields:
+        return
+    cols = ", ".join(f"{k}=?" for k in fields)
+    with get_conn() as conn:
+        conn.execute(
+            f"UPDATE meeting_transcripts SET {cols} WHERE id=?",
+            (*fields.values(), transcript_id),
+        )
+        conn.commit()
+
+
 def save_transcript(page_id: int, transcript: str, language: str = "fr",
-                    audio_path: str = "") -> int:
+                    audio_path: str = "", **kw) -> int:
+    """Crée une note de réunion (état ``idle``). Rétro-compatible v7.1.0."""
     with get_conn() as conn:
         if not conn.execute("SELECT id FROM pages WHERE id=?", (page_id,)).fetchone():
             raise ValueError("page not found")
+        occ = (kw.get("event_occurrence_id") or "").strip()
+        if occ and conn.execute(
+            "SELECT id FROM meeting_transcripts WHERE event_occurrence_id=?", (occ,)
+        ).fetchone():
+            raise ValueError("a meeting note already exists for this calendar occurrence")
         cur = conn.execute(
-            """INSERT INTO meeting_transcripts (page_id, audio_path, transcript, language)
-               VALUES (?,?,?,?)""",
-            (page_id, audio_path, transcript or "", language or "fr"))
+            """INSERT INTO meeting_transcripts
+               (page_id, audio_path, transcript, language, status, mode,
+                instruction_id, event_occurrence_id, notes_text)
+               VALUES (?,?,?,?,?,?,?,?,?)""",
+            (page_id, audio_path, transcript or "", language or "fr",
+             "idle", kw.get("mode") or "mic_only",
+             (kw.get("instruction_id") or "auto")[:40], occ,
+             kw.get("notes_text") or ""),
+        )
         conn.commit()
         return cur.lastrowid
 
 
-def transcribe_audio(audio_path: str, language: str = "fr") -> str:
-    """Transcribe with ``STT_COMMAND`` (``{cmd} {file}` → stdout text).
+# ── Consentement (§14.1) : porte bloquante, journal append-only ──────────────
 
-    Example: ``STT_COMMAND="whisper --language fr --output_format txt --output_dir /tmp"``
-    (command must print or be adapted — stdout is preferred). Raises
-    :class:`TranscriptionUnavailable` when unconfigured.
+def record_consent(transcript_id: int, *, method: str = "start_attestation",
+                   message_ref: str = "", attested_by: int | None = None,
+                   participants: list | None = None) -> dict:
+    if method not in CONSENT_METHODS:
+        raise ValueError(f"unknown consent method: {method}")
+    get_transcript(transcript_id)  # 404 si inconnu
+    entry = {
+        "method": method,
+        "message_ref": message_ref or "",
+        "attested_by": attested_by,
+        "participants": participants or [],
+        "attested_at": _now(),
+    }
+    with get_conn() as conn:
+        conn.execute(
+            """INSERT INTO meeting_consent_log
+               (transcript_id, method, message_ref, attested_by, participants_json)
+               VALUES (?,?,?,?,?)""",
+            (transcript_id, method, message_ref or "",
+             attested_by, json.dumps(participants or [], ensure_ascii=False)),
+        )
+        conn.execute(
+            "UPDATE meeting_transcripts SET consent_json=? WHERE id=?",
+            (json.dumps(entry, ensure_ascii=False), transcript_id),
+        )
+        conn.commit()
+    return entry
+
+
+def require_consent(tr: dict) -> dict:
+    consent = _loads(tr.get("consent_json") or "", None)
+    if consent:
+        return consent
+    with get_conn() as conn:
+        row = conn.execute(
+            "SELECT * FROM meeting_consent_log WHERE transcript_id=? "
+            "ORDER BY id DESC LIMIT 1", (tr["id"],)
+        ).fetchone()
+    if not row:
+        raise ValueError("consent required — record consent before capturing")
+    return {
+        "method": row["method"], "message_ref": row["message_ref"],
+        "attested_by": row["attested_by"],
+        "participants": _loads(row["participants_json"], []),
+    }
+
+
+# ── Sessions de capture : start / pause / resume / stop ─────────────────────
+
+def start_session(transcript_id: int, *, mode: str = "mic_only",
+                  channels: list | None = None, language: str = "fr",
+                  instruction_id: str = "auto", user_id: int | None = None) -> dict:
+    if mode not in CAPTURE_MODES:
+        raise ValueError(f"unknown capture mode: {mode}")
+    tr = get_transcript(transcript_id)
+    if tr.get("status") in ("recording", "processing"):
+        raise ValueError(f"meeting is {tr['status']} — stop it first")
+    consent = require_consent(tr)
+    instr = (instruction_id or "auto").strip()[:40]
+    snapshot = {
+        "instruction_id": instr if instr in INSTRUCTIONS else "custom",
+        "instruction_raw": instr,
+        "definition": INSTRUCTIONS.get(instr, {"label": instr, "prompt": instr}),
+        "frozen_at": _now(),
+    }
+    chans = channels or (
+        [{"kind": "self", "label": "Microphone"}] if mode == "mic_only"
+        else [{"kind": "self", "label": "My microphone"},
+              {"kind": "remote", "label": "Tab audio"}] if mode == "tab_plus_mic"
+        else []
+    )
+    _update(transcript_id, status="recording", mode=mode,
+            channels_json=json.dumps(chans, ensure_ascii=False),
+            language=(language or tr.get("language") or "fr")[:10],
+            instruction_id=instr,
+            instruction_snapshot=json.dumps(snapshot, ensure_ascii=False),
+            segments_json="[]", processing_steps_json="[]", summary="",
+            summary_json="", generated_title="", completeness_class="",
+            duration_ms=0, paused_ms=0, started_at=_now(), stopped_at=None)
+    return get_transcript(transcript_id) | {"_consent": consent}
+
+
+def pause_session(transcript_id: int, *, paused_ms_delta: int = 0) -> dict:
+    tr = get_transcript(transcript_id)
+    if tr.get("status") != "recording":
+        raise ValueError("only a recording session can be paused")
+    _update(transcript_id, status="paused",
+            paused_ms=int(tr.get("paused_ms") or 0) + max(0, int(paused_ms_delta or 0)))
+    return get_transcript(transcript_id)
+
+
+def resume_session(transcript_id: int) -> dict:
+    tr = get_transcript(transcript_id)
+    if tr.get("status") != "paused":
+        raise ValueError("only a paused session can be resumed")
+    require_consent(tr)
+    _update(transcript_id, status="recording")
+    return get_transcript(transcript_id)
+
+
+def stop_session(transcript_id: int, *, duration_ms: int = 0) -> dict:
+    tr = get_transcript(transcript_id)
+    if tr.get("status") not in ("recording", "paused"):
+        raise ValueError("nothing to stop — no live session")
+    require_consent(tr)
+    _update(transcript_id, status="processing", stopped_at=_now(),
+            duration_ms=max(int(duration_ms or 0), int(tr.get("duration_ms") or 0)))
+    return get_transcript(transcript_id)
+
+
+# ── Segments : append (temps réel), correction du locuteur ───────────────────
+
+def _next_seq(segments: list) -> int:
+    return (max((s.get("seq", 0) for s in segments), default=0) + 1) if segments else 1
+
+
+def append_segments(transcript_id: int, segments: list) -> dict:
+    """Persiste les segments finaux du client (STT navigateur / import).
+
+    Chaque item : {channel, source_label, speaker, text, start_ms, end_ms,
+    starts_mid, ends_mid}. Le transcript texte est reconstruit pour compat.
     """
+    tr = get_transcript(transcript_id)
+    # Retry/regen après un échec : le client reverse le manuel collé
+    # entre-temps — l'ajout reste possible hors session live.
+    if tr.get("status") not in ("recording", "paused", "processing",
+                                 "failed", "done"):
+        raise ValueError("session is not live — start recording first")
+    require_consent(tr)
+    if not isinstance(segments, list) or not segments:
+        raise ValueError("segments must be a non-empty list")
+    if len(segments) > 200:
+        raise ValueError("too many segments per batch (max 200)")
+    current = _loads(tr.get("segments_json") or "[]", [])
+    seq = _next_seq(current)
+    for item in segments:
+        if not isinstance(item, dict) or not (item.get("text") or "").strip():
+            raise ValueError("each segment needs a non-empty text")
+        text = item["text"].strip()[:4000]
+        current.append({
+            "seq": seq,
+            "channel": (item.get("channel") or "self")[:12],
+            "source_label": (item.get("source_label") or "")[:120],
+            "speaker": (item.get("speaker") or "")[:120],
+            "start_ms": max(0, int(item.get("start_ms") or 0)),
+            "end_ms": max(0, int(item.get("end_ms") or 0)),
+            "text": ("--" if item.get("starts_mid") else "") + text,
+            "is_final": True,
+            "starts_mid_utterance": bool(item.get("starts_mid")),
+            "ends_mid_utterance": bool(item.get("ends_mid")),
+        })
+        seq += 1
+    flat = "\n".join(
+        f"[{s['start_ms'] // 60000}:{(s['start_ms'] // 1000) % 60:02d}] "
+        f"{s['speaker'] or s['channel']} : {s['text']}" for s in current
+    )
+    if tr.get("transcript"):
+        flat = (tr["transcript"] + "\n" + flat) if flat else tr["transcript"]
+    _update(transcript_id, segments_json=json.dumps(current, ensure_ascii=False),
+            transcript=flat[:200000])
+    return {"transcript_id": transcript_id, "segments": len(current)}
+
+
+def set_segment_speaker(transcript_id: int, seq: int, speaker: str,
+                        user_id: int | None = None) -> dict:
+    tr = get_transcript(transcript_id)
+    segments = _loads(tr.get("segments_json") or "[]", [])
+    found = False
+    for s in segments:
+        if int(s.get("seq") or 0) == int(seq):
+            s["speaker"] = (speaker or "").strip()[:120]
+            s["speaker_corrected_by"] = user_id
+            found = True
+    if not found:
+        raise ValueError("segment not found")
+    _update(transcript_id, segments_json=json.dumps(segments, ensure_ascii=False))
+    return {"transcript_id": transcript_id, "seq": int(seq), "speaker": speaker}
+
+
+def grouped_transcript(transcript_id: int) -> dict:
+    """Transcript regroupé par piste source (§3A.7) + rail temporel."""
+    tr = get_transcript(transcript_id)
+    segments = _loads(tr.get("segments_json") or "[]", [])
+    groups: dict[str, list] = {}
+    order: list[str] = []
+    for s in sorted(segments, key=lambda x: (x.get("seq") or 0)):
+        key = s.get("source_label") or s.get("channel") or "self"
+        if key not in groups:
+            groups[key] = []
+            order.append(key)
+        groups[key].append(s)
+    # Repli : transcript texte brut découpé en un groupe unique.
+    if not groups and (tr.get("transcript") or "").strip():
+        groups = {"audio": [{"seq": 1, "channel": "self", "source_label": "",
+                             "speaker": "", "start_ms": 0, "end_ms": 0,
+                             "text": tr["transcript"][:4000]}]}
+        order = ["audio"]
+    return {"transcript_id": transcript_id,
+            "groups": [{"source": k, "segments": groups[k]} for k in order]}
+
+
+# ── Classification de complétude + titre (§12.8) ─────────────────────────────
+
+def classify_completeness(*, text: str, segments: list, notes: str,
+                          duration_ms: int = 0) -> str:
+    words = len(re.findall(r"\S+", text or ""))
+    n_seg = len(segments or [])
+    truncated = any(s.get("starts_mid_utterance") or s.get("ends_mid_utterance")
+                    for s in (segments or []))
+    has_agenda = bool((notes or "").strip())
+    if n_seg == 0 or words < 8:
+        return "fragmentary"
+    if words < 60 or (duration_ms and duration_ms < 45_000):
+        return "brief" if not truncated else "incomplete"
+    if truncated or not has_agenda and words < 150:
+        return "incomplete"
+    if words < 150:
+        return "brief"
+    return "complete"
+
+
+def _first_topic(text: str) -> str:
+    sentences = re.split(r"(?<=[.!?])\s+", re.sub(r"\s+", " ", text or "").strip())
+    for s in sentences:
+        s = s.strip(" -–—\"'«»")
+        if len(s) > 25:
+            return s[:90].rstrip(" ,;:")
+    return ""
+
+
+def generate_title(*, text: str, completeness: str, notes: str = "") -> str:
+    if completeness in ("brief", "incomplete", "fragmentary"):
+        topic = _first_topic(text)
+        if not topic or len(re.findall(r"\S+", text or "")) < 25:
+            return "Brief or Incomplete Meeting Recording"
+        return topic[:90]
+    topic = _first_topic(notes + " " + text)
+    return topic[:90] or "Meeting notes"
+
+
+# ── Résumé structuré (§12.4) : JSON contraint + citations ────────────────────
+
+_SUMMARY_SCHEMA_HINT = (
+    "Réponds STRICTEMENT par un objet JSON : "
+    '{"overview": ["..."], "decisions": ["..."], "action_items": '
+    '[{"title": "...", "owner": null, "due_hint": null}], '
+    '"open_questions": ["..."], "limits": ["..."]}. '
+    "Chaque puce overview/decisions/action_items DOIT citer le transcript "
+    "(suffixe entre parenthèses avec un extrait repris mot pour mot). "
+    "N'invente ni responsable ni échéance : owner/due_hint null si absents. "
+    "Si le contenu est insuffisant, overview décrit les manques et "
+    "action_items vaut []."
+)
+
+
+def _cite(text: str, segments: list) -> str:
+    """Trouve un segment source pour une affirmation (citation vérifiable)."""
+    words = [w.lower() for w in re.findall(r"[A-Za-zÀ-ÿa-zà-ÿ']{4,}", text or "")]
+    best, best_score = None, 0
+    for s in segments or []:
+        hay = (s.get("text") or "").lower()
+        score = sum(1 for w in words[:12] if w in hay)
+        if score > best_score:
+            best, best_score = s, score
+    if best and best_score > 0:
+        return f" ⁠¹ [seg {best.get('seq')}]"
+    if segments:
+        return f" ⁠¹ [seg {segments[0].get('seq')}]"
+    return " ⁠¹"
+
+
+def _offline_structured(*, text: str, notes: str, segments: list,
+                        completeness: str) -> dict:
+    sentences = [s.strip() for s in
+                 re.split(r"(?<=[.!?])\s+", re.sub(r"\s+", " ", text).strip()) if s.strip()]
+    overview: list[str] = []
+    if completeness == "fragmentary":
+        overview = [
+            "The recording captured only a short, fragmentary exchange with no "
+            f"discernible meeting topic, agenda, or context{_cite(text, segments)}",
+            "The content appears to be mid-conversation and lacks the beginning "
+            f"of the discussion{_cite(text, segments)}",
+        ]
+        if not (notes or "").strip():
+            overview.append("The user's notes were empty, providing no additional "
+                            "context about the meeting's purpose")
+    else:
+        for s in sentences[:4]:
+            overview.append(f"{s[:220]}{_cite(s, segments)}")
+        if not overview:
+            overview = [f"No meaningful content could be extracted{_cite(text, segments)}"]
+        if completeness in ("brief", "incomplete"):
+            overview.append(
+                "This was a brief or incomplete recording — treat this summary "
+                f"as partial{_cite(text, segments)}")
+    decisions = []
+    for s in sentences:
+        if re.search(r"\b(decid|décid|agreed|approuv|retain|retenu|chosen|choisi)\b", s, re.I):
+            decisions.append(f"{s[:220]}{_cite(s, segments)}")
+    actions = []
+    for s in sentences:
+        m = re.search(r"\b([A-ZÀ-Þ][a-zà-ÿ'\-]+)\s+(will|va|doit|should|must|s'occupe)\b", s)
+        if re.search(r"\b(action|todo|to-?do|next step|prochaine étape|faire|envoyer|préparer)\b", s, re.I) or m:
+            owner = m.group(1) if m and len(m.group(1)) > 2 else None
+            actions.append({"title": s[:180], "owner": owner,
+                            "due_hint": None, "evidence": _cite(s, segments)})
+    return {"overview": overview[:6], "decisions": decisions[:8],
+            "action_items": actions[:10], "open_questions": [],
+            "limits": [] if completeness == "complete" else overview[-2:]}
+
+
+def _render_markdown(struct: dict, completeness: str) -> str:
+    L = ["## Overview", ""]
+    for b in struct.get("overview") or []:
+        L.append(f"•   {b}")
+    L += ["", "## Decisions", ""]
+    if struct.get("decisions"):
+        for b in struct["decisions"]:
+            L.append(f"•   {b}")
+    else:
+        L.append("•   No decisions could be identified from the available transcript")
+    L += ["", "## Action Items", ""]
+    if struct.get("action_items"):
+        for a in struct["action_items"]:
+            if isinstance(a, dict):
+                owner = f" — owner: {a['owner']}" if a.get("owner") else ""
+                L.append(f"•   {a.get('title', '')}{owner}{a.get('evidence', ' ¹')}")
+            else:
+                L.append(f"•   {a}")
+    else:
+        L.append("•   No action items could be identified from the available transcript")
+    if struct.get("open_questions"):
+        L += ["", "## Open questions", ""]
+        for b in struct["open_questions"]:
+            L.append(f"•   {b}")
+    if completeness in ("brief", "incomplete", "fragmentary"):
+        L += ["", f"_Summary class: {completeness} — partial recording, verify against the transcript._"]
+    return "\n".join(L).strip()
+
+
+def _write_steps(transcript_id: int, states: dict) -> None:
+    steps = [{"key": k, "label": label,
+              "state": states.get(k, "pending"),
+              "at": _now() if states.get(k) in ("running", "done", "failed") else ""}
+             for k, label in PROCESSING_STEPS]
+    _update(transcript_id, processing_steps_json=json.dumps(steps, ensure_ascii=False))
+
+
+async def _llm_structured(*, text: str, notes: str, instruction: str,
+                          user_id: int | None) -> tuple[dict | None, bool, str]:
+    from app.services.ai_writing import AIWritingService
+
+    svc = AIWritingService(user_id=user_id)
+    if svc._offline_hint():
+        return None, True, ""
+    prompt = (
+        "Rédige le résumé structuré de cette réunion (format AI Meeting Note). "
+        f"Instructions de l'utilisateur : {instruction or 'Auto'}. "
+        "Notes humaines (contexte prioritaire, signale tout conflit) : "
+        f"{(notes or '(vides)')[:2000]}. " + _SUMMARY_SCHEMA_HINT
+    )
+    try:
+        res = await svc.run("summarize", context=text[:18000], prompt=prompt)
+    except Exception as exc:  # noqa: BLE001
+        logger.warning("meeting LLM summary failed: %s", exc)
+        return None, False, str(exc)
+    if not res.get("ok") or not (res.get("text") or "").strip():
+        return None, bool(res.get("offline")), res.get("error") or "empty"
+    parsed = AIWritingService._parse_json_object(res["text"])
+    if not parsed or not isinstance(parsed.get("overview"), list):
+        # Le modèle a répondu en texte libre : on le conserve en overview sourcé.
+        return {"overview": [res["text"][:1500]], "decisions": [],
+                "action_items": [], "open_questions": [], "limits": []}, False, ""
+    return parsed, False, ""
+
+
+async def run_processing(transcript_id: int, user_id: int | None = None,
+                         allow_empty: bool = False) -> dict:
+    """Pipeline §12 : étapes Thinking → classification → résumé → titre.
+
+    Jamais de refus sur contenu insuffisant : résumé dégradé + classe +
+    titre-diagnostic (§12.8). Avec ``allow_empty`` (arrêt sans rien de
+    capté), un diagnostic honnête est produit au lieu d'une erreur —
+    l'utilisateur ne perd ni l'enregistrement ni le transcript.
+    Sans ``allow_empty``, un transcript vide reste une 400 (l'appelant
+    affiche alors un guide de reprise, pas un dump HTTP).
+    """
+    tr = get_transcript(transcript_id)
+    segments = _loads(tr.get("segments_json") or "[]", [])
+    notes = tr.get("notes_text") or ""
+    text = (tr.get("transcript") or "").strip()
+    snapshot = _loads(tr.get("instruction_snapshot") or "", {})
+    instruction = (snapshot.get("instruction_raw")
+                   or tr.get("instruction_id") or "auto")
+
+    order = [k for k, _ in PROCESSING_STEPS]
+    states: dict[str, str] = {}
+    try:
+        for key in ("reading_transcript", "analyzing_transcript",
+                    "resolving_references", "understanding_content"):
+            states[key] = "running"
+            _write_steps(transcript_id, states)
+            time.sleep(0)  # point de cession : étapes réellement exécutées
+            states[key] = "done"
+            _write_steps(transcript_id, states)
+
+        states["classifying_completeness"] = "running"
+        _write_steps(transcript_id, states)
+        completeness = classify_completeness(
+            text=text, segments=segments, notes=notes,
+            duration_ms=int(tr.get("duration_ms") or 0))
+        states["classifying_completeness"] = "done"
+        _write_steps(transcript_id, states)
+
+        if not text:
+            if not allow_empty:
+                raise ValueError("transcript is empty — transcribe first")
+            # Arrêt sans rien de capté (micro coupé, aucun mot reconnu) :
+            # diagnostic honnête au lieu d'une erreur — rien n'est perdu,
+            # l'utilisateur peut coller un transcript et régénérer.
+            for key in ("generating_summary", "generating_title",
+                        "validating_citations"):
+                states[key] = "running"
+                _write_steps(transcript_id, states)
+                states[key] = "done"
+                _write_steps(transcript_id, states)
+            struct = {
+                "overview": [
+                    "No audio was captured during this session — no speech "
+                    "reached the transcript. Check that the microphone is "
+                    "allowed for this site and that the correct input is selected.",
+                    ("If the call ran in another tab, restart with “Tab + "
+                     "microphone” and tick “Share audio” when choosing the tab."),
+                ],
+                "decisions": [], "action_items": [], "open_questions": [],
+                "limits": ["empty session — paste a transcript, then Regenerate"],
+            }
+            md = _render_markdown(struct, "fragmentary")
+            _update(transcript_id, status="done", summary=md,
+                    summary_json=json.dumps(struct, ensure_ascii=False),
+                    generated_title="Brief or Incomplete Meeting Recording",
+                    title_source="ai", completeness_class="fragmentary")
+            return await _finish_processing(transcript_id, offline=True)
+
+        states["generating_summary"] = "running"
+        _write_steps(transcript_id, states)
+        struct, offline, _err = await _llm_structured(
+            text=text, notes=notes, instruction=instruction, user_id=user_id)
+        if struct is None:
+            struct = _offline_structured(text=text, notes=notes,
+                                         segments=segments,
+                                         completeness=completeness)
+            offline = True
+        else:
+            # Garantir les citations même sur réponse LLM.
+            for section in ("overview", "decisions"):
+                fixed = []
+                for b in struct.get(section) or []:
+                    s = b if isinstance(b, str) else str(b)
+                    fixed.append(s if "¹" in s or "[seg" in s else f"{s}{_cite(s, segments)}")
+                struct[section] = fixed
+            acts = []
+            for a in struct.get("action_items") or []:
+                if isinstance(a, dict):
+                    if "evidence" not in a:
+                        a["evidence"] = _cite(a.get("title", ""), segments)
+                    acts.append(a)
+                else:
+                    acts.append({"title": str(a), "owner": None,
+                                 "due_hint": None,
+                                 "evidence": _cite(str(a), segments)})
+            struct["action_items"] = acts
+        states["generating_summary"] = "done"
+        _write_steps(transcript_id, states)
+
+        states["generating_title"] = "running"
+        _write_steps(transcript_id, states)
+        title = generate_title(text=text, completeness=completeness, notes=notes)
+        states["generating_title"] = "done"
+        _write_steps(transcript_id, states)
+
+        states["validating_citations"] = "running"
+        _write_steps(transcript_id, states)
+        seqs = {int(s.get("seq") or 0) for s in segments}
+        # Toute citation [seg N] pointe un segment réel, sinon repli seg 1.
+        md = _render_markdown(struct, completeness)
+
+        def _fix_seg(m: re.Match) -> str:
+            try:
+                n = int(m.group(1))
+            except (TypeError, ValueError):
+                return "[seg 1]"
+            return m.group(0) if n in seqs else "[seg 1]"
+        md = re.sub(r"\[seg (\d+)\]", _fix_seg, md)
+        states["validating_citations"] = "done"
+        _write_steps(transcript_id, states)
+
+        _update(transcript_id, status="done", summary=md,
+                summary_json=json.dumps(struct, ensure_ascii=False),
+                generated_title=title, title_source="ai",
+                completeness_class=completeness)
+    except Exception as exc:  # noqa: BLE001 — état failed récupérable, segments conservés
+        for k in order:
+            states.setdefault(k, "pending")
+        # Marquer l'étape courante en échec.
+        for k in order:
+            if states.get(k) == "running":
+                states[k] = "failed"
+                break
+        _write_steps(transcript_id, states)
+        # Toujours sortir de « processing » : une ligne coincée est pire
+        # qu'une erreur affichée (l'utilisateur peut réessayer).
+        _update(transcript_id, status="failed")
+        msg = str(exc) or "processing failed"
+        if "transcript is empty" in msg:
+            raise ValueError(msg) from exc
+        logger.warning("meeting processing %s failed: %s", transcript_id, exc)
+        raise RuntimeError(msg) from exc
+
+    return await _finish_processing(transcript_id, offline)
+
+
+async def _finish_processing(transcript_id: int, offline: bool) -> dict:
+    """Événement + réponse du pipeline (chemin nominal et diagnostic)."""
+    out = get_transcript(transcript_id)
+    try:
+        from app.services.automations import fire_event
+        await fire_event("meeting.summarized", {
+            "page_id": out["page_id"], "transcript_id": transcript_id,
+            "language": out.get("language") or "fr",
+            "completeness": out.get("completeness_class") or "",
+        })
+    except Exception as exc:  # noqa: BLE001
+        logger.debug("meeting.summarized dispatch failed: %s", exc)
+    return {"transcript_id": transcript_id, "summary": out["summary"],
+            "completeness": out.get("completeness_class") or "",
+            "generated_title": out.get("generated_title") or "",
+            "offline": bool(offline)}
+
+
+def transcribe_audio(audio_path: str, language: str = "fr") -> str:
+    """Transcribe with ``STT_COMMAND`` (``{cmd} {file}` → stdout text)."""
     cmd_template = os.environ.get("STT_COMMAND", "").strip()
     if not cmd_template:
         raise TranscriptionUnavailable(
@@ -74,40 +694,39 @@ def transcribe_audio(audio_path: str, language: str = "fr") -> str:
     return text
 
 
-async def summarize_transcript(transcript_id: int, user_id: int | None = None) -> dict:
-    """Summarize a stored transcript + fire ``meeting.summarized``.
-
-    Returns {transcript_id, summary, offline}. Emits the automation event so
-    custom agents (update tracker, post recap, file tickets) trigger.
-    """
-    from app.services.ai_writing import AIWritingService
-
-    with get_conn() as conn:
-        row = conn.execute("SELECT * FROM meeting_transcripts WHERE id=?",
-                           (transcript_id,)).fetchone()
-        if not row:
-            raise ValueError("transcript not found")
-        tr = dict(row)
-    if not (tr.get("transcript") or "").strip():
+async def summarize_transcript(transcript_id: int, user_id: int | None = None,
+                               allow_empty: bool = False) -> dict:
+    """Point d'entrée historique (v7.1.0) — bascule sur le pipeline Notion."""
+    tr = get_transcript(transcript_id)
+    if tr.get("status") in ("recording", "paused"):
+        raise ValueError("stop the recording before generating the summary")
+    if not ((tr.get("transcript") or "").strip()
+            or _loads(tr.get("segments_json") or "[]", [])):
+        if not allow_empty:
             raise ValueError("transcript is empty — transcribe first")
-    svc = AIWritingService(user_id=user_id)
-    res = await svc.run("summarize", context=tr["transcript"])
-    summary = (res.get("text") or "").strip() if res.get("ok") else ""
-    if not summary:
-        raise RuntimeError(f"summarization failed: {res.get('error', 'unknown')}")
+    if tr.get("status") not in ("processing", "done"):
+        _update(transcript_id, status="processing")
+    out = await run_processing(transcript_id, user_id=user_id,
+                               allow_empty=allow_empty)
+    return {"transcript_id": transcript_id, "summary": out["summary"],
+            "offline": out["offline"]}
+
+
+# ── Partage tracé (§3A.5 : Copy link / Email / Slack) ─────────────────────────
+
+def record_distribution(transcript_id: int, *, channel: str,
+                        actor_id: int | None = None,
+                        recipients: str = "", status: str = "created") -> dict:
+    if channel not in ("copy_link", "email", "slack"):
+        raise ValueError("channel must be copy_link|email|slack")
+    get_transcript(transcript_id)
     with get_conn() as conn:
-        conn.execute("UPDATE meeting_transcripts SET summary=? WHERE id=?",
-                     (summary, transcript_id))
-        page = conn.execute("SELECT id FROM pages WHERE id=?", (tr["page_id"],)).fetchone()
+        cur = conn.execute(
+            """INSERT INTO meeting_distributions
+               (transcript_id, channel, actor_id, recipients_json, status)
+               VALUES (?,?,?,?,?)""",
+            (transcript_id, channel, actor_id, recipients or "", status),
+        )
         conn.commit()
-    try:
-        from app.services.automations import fire_event
-        await fire_event("meeting.summarized", {
-            "page_id": tr["page_id"] if page else 0,
-            "transcript_id": transcript_id,
-            "language": tr.get("language") or "fr",
-        })
-    except Exception as exc:  # noqa: BLE001 — summary stands even if dispatch fails
-        logger.debug("meeting.summarized dispatch failed: %s", exc)
-    return {"transcript_id": transcript_id, "summary": summary,
-            "offline": bool(res.get("offline"))}
+        return {"id": cur.lastrowid, "transcript_id": transcript_id,
+                "channel": channel, "status": status}

app/services/publish.py

@@ -9,10 +9,14 @@ Les trois surfaces divergeaient avant cette passe :
 - `/api/v2/pages/{id}/publish` : slug fourni par le corps ou aléatoire,
   `is_shared=1` en bonus (alors que v2 le remet à 0 quand aucun partage)
 
-Canonical (comportement du front) : `is_published` + `publish_slug` seulement,
-404 si la page n'existe pas. `share_mode`/`is_shared`/`published` restent la
-propriété du share dialog (`/board/api/share/{pid}`) : dépublier ne révoque
-donc pas un partage manuel.
+Canonical (comportement du front) : `is_published` + `published` +
+`publish_slug`, maintenus ensemble par toutes les écritures (v7.69.6 —
+avant, chaque surface n'écrivait qu'un drapeau et chaque lecteur ne
+regardait qu'une colonne, donc l'état publié ne « tenait » nulle part).
+`share_mode`/`is_shared` restent la propriété du share dialog
+(`/board/api/share/{pid}`) : dépublier ne révoque donc pas un partage manuel.
+Les lecteurs acceptent `is_published=1 OR published=1` (lignes historiques
+issues d'un seul chemin).
 """
 from __future__ import annotations
 
@@ -58,8 +62,12 @@ def publish(page_id: int, explicit_slug: str | None = None) -> tuple[str, str]:
         if not page:
             raise HTTPException(404, "Page not found")
         slug = explicit_slug or _unique_slug(conn, page_id, page["title"])
+        # v7.69.6 : les deux drapeaux sont maintenus ensemble — le share
+        # dialog écrit ``published``, les routes /publish écrivaient seulement
+        # ``is_published``, et chaque lecteur ne regardait qu'une colonne :
+        # l'état publié ne « tenait » nulle part.
         conn.execute(
-            "UPDATE pages SET is_published=1, publish_slug=? WHERE id=?", (slug, page_id)
+            "UPDATE pages SET is_published=1, published=1, publish_slug=? WHERE id=?", (slug, page_id)
         )
         conn.commit()
     return slug, page["title"] or ""
@@ -71,7 +79,7 @@ def unpublish(page_id: int) -> None:
         if not conn.execute("SELECT id FROM pages WHERE id=?", (page_id,)).fetchone():
             raise HTTPException(404, "Page not found")
         conn.execute(
-            "UPDATE pages SET is_published=0, publish_slug='' WHERE id=?", (page_id,)
+            "UPDATE pages SET is_published=0, published=0, publish_slug='' WHERE id=?", (page_id,)
         )
         conn.commit()
 

app/templates/_header.html

@@ -181,6 +181,16 @@ document.addEventListener('alpine:init', function () {
             return String(c.id) === id ? Object.assign({}, c, { label: d.title || 'Untitled' }) : c;
           });
         });
+        // Navigation in-place (ex. /local-workspace qui descend dans ses
+        // dossiers sans recharger) : le chemin du header est recalé sur la
+        // pile de dossiers courante. Même contrat que le rendu serveur.
+        window.addEventListener('flowdeck:breadcrumb-changed', function (e) {
+          var d = e && e.detail;
+          if (!d || !d.crumbs || !d.crumbs.length) return;
+          self.crumbs = d.crumbs.map(function (c, i) { c.origIndex = i; return c; });
+          if (d.workspace_id) self.wsId = d.workspace_id;
+          self.closeAll();
+        });
       },
 
       get display() {

app/templates/_page_editor_scripts.html

@@ -1,4 +1,10 @@
 <script type="application/json" id="page-data" nonce="{{ csp_nonce() }}">
     {{ page_data | tojson }}
 </script>
+<link rel="stylesheet" href="/static/css/highlight.css?v={{ asset_version }}">
+<script data-cfasync="false" src="/static/js/highlight.min.js?v={{ asset_version }}"></script>
+<script data-cfasync="false" src="/static/js/highlight-extra.min.js?v={{ asset_version }}"></script>
+<script data-cfasync="false" src="/static/js/code-highlight.js?v={{ asset_version }}"></script>
 <script data-cfasync="false" src="/static/js/page_editor_scripts.js?v={{ asset_version }}"></script>
+<link rel="stylesheet" href="/static/css/meeting.css?v={{ asset_version }}">
+<script data-cfasync="false" src="/static/js/meeting_block.js?v={{ asset_version }}"></script>

app/templates/base.html

@@ -361,13 +361,28 @@
         </div>
         <div class="sidebar-section-actions">
           {% if has_active_workspace %}
-          <button class="section-action-btn" title="Add" @click.stop="fdCreatePage()">+</button>
+          <button class="section-action-btn" title="New meeting note" @click.stop="fdCreateMeeting()">+</button>
           {% endif %}
           <button class="section-action-btn section-menu-btn" title="Section options" @click.stop="openSectionMenu($event, 'meetings')">⋮</button>
         </div>
       </div>
       <div class="sidebar-section-items" x-show="sectionsOpen.meetings" x-transition>
-        <ul class="sidebar-items"><li class="sidebar-item empty-hint"><span class="page-icon page-icon-svg">{{ fd_icon("calendar",16) }}</span><span class="page-name text-dim">No meetings</span></li></ul>
+        <ul class="sidebar-items" id="sidebar-meetings-items">
+          <template x-for="n in meetingNotes" :key="'mn-' + n.id">
+            <li class="sidebar-item" @click="meetingNoteOpen(n)"
+                @contextmenu.prevent="showWsContext($event, n.id, n.title, false)">
+              <span class="page-icon page-icon-svg">{{ fd_icon("calendar",14) }}</span>
+              <span style="flex:1;min-width:0;">
+                <span class="page-name" style="display:block;" x-text="n.title"></span>
+                <span class="sb-item-sub" x-text="n.has_summary ? 'Summary ready' : (n.status && n.status !== 'idle' ? n.status : 'Ready to record')"></span>
+              </span>
+            </li>
+          </template>
+          <li class="sidebar-item empty-hint" x-show="!meetingNotes.length">
+            <span class="page-icon page-icon-svg">{{ fd_icon("calendar",16) }}</span>
+            <span class="page-name text-dim">No meetings</span>
+          </li>
+        </ul>
       </div>
     </div>
 
@@ -662,6 +677,10 @@
       <div class="sb-filter-row">
         <button type="button" class="sb-filter-chip" :class="{ active: meetingFilter === 'upcoming' }" @click="meetingFilter = 'upcoming'">Upcoming</button>
         <button type="button" class="sb-filter-chip" :class="{ active: meetingFilter === 'today' }" @click="meetingFilter = 'today'">Today</button>
+        <button type="button" class="sb-filter-chip" :class="{ active: meetingFilter === 'past' }" @click="meetingFilter = 'past'; loadMeetingHistory()">Past</button>
+      </div>
+      <div class="sb-search-row">
+        <input type="search" class="sb-search-input" placeholder="Search meetings…" x-model="meetingSearch" @input="if (meetingFilter === 'past') loadMeetingHistory()" aria-label="Search meetings">
       </div>
 
       <div class="sb-quick-actions">
@@ -671,15 +690,18 @@
       </div>
 
       <ul class="sidebar-items">
-        <template x-for="ev in meetingVisible" :key="ev.collection_id + '-' + ev.id">
+        <template x-for="ev in meetingVisible" :key="(ev.collection_id || 'h') + '-' + ev.id">
           <li class="sidebar-item" @click="meetingOpen(ev)">
             <span class="page-icon page-icon-svg">{{ fd_icon("calendar",14) }}</span>
-            <span class="page-name" x-text="ev.title"></span>
-            <span class="sb-item-meta" x-text="ev.today ? 'Today' : ev.date"></span>
+            <span style="flex:1;min-width:0;">
+              <span class="page-name" style="display:block;" x-text="ev.title"></span>
+              <span class="sb-item-sub" x-text="meetingCta(ev)"></span>
+            </span>
+            <span class="sb-item-meta" x-text="ev.today ? 'Today' : (ev.date || '')"></span>
           </li>
         </template>
         <li class="sidebar-item empty-hint" x-show="!meetingsLoading && !meetingVisible.length">
-          <span class="page-name text-dim" x-text="meetingFilter === 'today' ? 'Nothing scheduled today' : 'No upcoming meetings'"></span>
+          <span class="page-name text-dim" x-text="meetingEmptyHint()"></span>
         </li>
         <li class="sidebar-item empty-hint" x-show="meetingsLoading">
           <span class="page-name text-dim">Loading…</span>
@@ -719,18 +741,35 @@
       </ul>
     </div>
 
-    <!-- Footer -->
+    <!-- Footer : bouton "New" + popover "Create new" (vers le haut) -->
     <div class="sidebar-footer">
-      <div style="display:flex; gap:6px;">
-        <button class="sidebar-footer-btn" style="flex:1; justify-content:center; background:var(--bg-hover); border-radius:20px;">
-          <span style="margin-right:4px;">{{ fd_icon("message-square",14) }}</span>New chat
-          <span class="shortcut">Ctrl+O</span>
+      <div class="fd-new-wrap" @click.outside="closeNewMenu()">
+        <div class="fd-new-popover" x-show="newMenuOpen" x-cloak
+             x-transition:enter="transition ease-out duration-150"
+             x-transition:enter-start="opacity-0 translate-y-1"
+             x-transition:enter-end="opacity-100 translate-y-0"
+             x-transition:leave="transition ease-in duration-100"
+             x-transition:leave-start="opacity-100"
+             x-transition:leave-end="opacity-0"
+             role="menu" aria-label="Create new">
+          <div class="fd-new-pop-header">Create new</div>
+          <button type="button" class="fd-new-pop-item" role="menuitem" @click="newMenuAction('page')">
+            <span>{{ fd_icon("file",16) }}</span><span>Page</span>
           </button>
-        {% if has_active_workspace %}
-        <button class="sidebar-footer-btn" style="width:36px; justify-content:center; background:var(--bg-hover); border-radius:8px;" onclick="window.FlowDeck.createPage()" title="New page">
-          <svg width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M14 2H6a2 2 0 00-2 2v16a2 2 0 002 2h12a2 2 0 002-2V8z"/><polyline points="14 2 14 8 20 8"/><line x1="12" y1="18" x2="12" y2="12"/><line x1="9" y1="15" x2="15" y2="15"/></svg>
+          <button type="button" class="fd-new-pop-item" role="menuitem" @click="newMenuAction('chat')">
+            <span>{{ fd_icon("message-square",16) }}</span><span>Chat</span>
+          </button>
+          <button type="button" class="fd-new-pop-item" role="menuitem" @click="newMenuAction('meeting')">
+            <span><svg width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><rect x="9" y="2" width="6" height="12" rx="3"/><path d="M5 10a7 7 0 0 0 14 0"/><line x1="12" y1="17" x2="12" y2="22"/><line x1="8" y1="22" x2="16" y2="22"/></svg></span><span>AI Meeting Notes</span>
+          </button>
+          <button type="button" class="fd-new-pop-item" role="menuitem" @click="newMenuAction('database')">
+            <span>{{ fd_icon("grid",16) }}</span><span>Database</span>
+          </button>
+        </div>
+        <button type="button" class="fd-new-btn" :class="{ open: newMenuOpen }"
+                @click="toggleNewMenu()" :aria-expanded="newMenuOpen" title="Create new">
+          <span>{{ fd_icon("edit",16) }}</span><span>New</span>
         </button>
-        {% endif %}
       </div>
     </div>
     </div><!-- /sidebar-scroll -->
@@ -899,6 +938,9 @@
       </div>
     </template>
     <div class="context-menu-section">
+      <div class="context-menu-item" @click="wsCtxAction('open')">
+        <span class="menu-icon">{{ fd_icon("external-link",14) }}</span> Open
+      </div>
       <div class="context-menu-item" @click="wsCtxAction('rename')">
         <span class="menu-icon">{{ fd_icon("edit",14) }}</span> Rename
       </div>
@@ -1294,6 +1336,8 @@
         this.loadTeamspaces();
         // Badge counts on the collapsed tabs (lazy fetch, no blocking).
         this.loadInbox();
+        // v7.1.1 : notes de réunion (section Meetings, pas Workspace).
+        this.loadMeetingNotes();
         this.initSidebarWidth();
         document.addEventListener('fd-toggle-sidebar', () => this.toggleSidebar());
         this.startClipAutoRefresh();
@@ -1310,7 +1354,7 @@
         });
         document.addEventListener('flowdeck:workspace-changed', () => {
           clearTimeout(self._clipRefreshTimer);
-          self._clipRefreshTimer = setTimeout(() => self.refreshSidebarTree(), 400);
+          self._clipRefreshTimer = setTimeout(() => { self.refreshSidebarTree(); self.loadMeetingNotes(); }, 400);
         });
         // Also refresh when tab becomes visible (user returns to FlowDeck after clipping)
         document.addEventListener('visibilitychange', () => {
@@ -1451,15 +1495,37 @@
 
       // ── Meeting tab ──
       meetings: [],
+      meetingPast: [],
       meetingsLoading: false,
       meetingFilter: 'upcoming',
+      meetingSearch: '',
+      meetingHistoryLoaded: false,
       get meetingVisible() {
-        if (this.meetingFilter === 'today') {
-          return this.meetings.filter(function(e) { return e.today; });
+        var q = (this.meetingSearch || '').trim().toLowerCase();
+        function matches(e) {
+          if (!q) return true;
+          return ((e.title || '') + ' ' + (e.date || '')).toLowerCase().indexOf(q) >= 0;
         }
-        return this.meetings.filter(function(e) { return !e.today; });
+        if (this.meetingFilter === 'today') {
+          return this.meetings.filter(function(e) { return e.today && matches(e); });
+        }
+        if (this.meetingFilter === 'past') {
+          return this.meetingPast.filter(matches);
+        }
+        return this.meetings.filter(function(e) { return !e.today && matches(e); });
       },
       get meetingToday() { return this.meetings.filter(function(e) { return e.today; }); },
+      meetingEmptyHint() {
+        if (this.meetingFilter === 'today') return 'Nothing scheduled today';
+        if (this.meetingFilter === 'past') return 'No past meeting notes yet';
+        return 'No upcoming meetings';
+      },
+      meetingCta(ev) {
+        if (ev.status === 'done' || ev.summary) return 'View summary';
+        if (ev.today) return 'Join & transcribe';
+        if (ev.page_id) return 'Open note';
+        return 'Prepare';
+      },
       loadMeetings(force) {
         var self = this;
         if (!force && this.meetings.length) return Promise.resolve();
@@ -1472,8 +1538,38 @@
           })
           .catch(function() { self.meetings = []; self.meetingsLoading = false; });
       },
+      loadMeetingHistory() {
+        var self = this;
+        this.meetingsLoading = true;
+        var q = encodeURIComponent((this.meetingSearch || '').trim());
+        return fetch('/api/v2/meetings/history?query=' + q + '&limit=50', { credentials: 'same-origin' })
+          .then(function(r) { return r.ok ? r.json() : { meetings: [] }; })
+          .then(function(d) {
+            self.meetingPast = (d && d.meetings) || [];
+            self.meetingHistoryLoaded = true;
+            self.meetingsLoading = false;
+          })
+          .catch(function() { self.meetingPast = []; self.meetingsLoading = false; });
+      },
       meetingOpen(ev) {
-        var target = ev.url || ('/db/' + ev.collection_id);
+        var target = ev.url || (ev.page_id ? ('/pages/' + ev.page_id) : ('/db/' + ev.collection_id));
+        if (ev.today) {
+          try { this.toast('Opening today\u2019s meeting — press Start transcribing in the note.'); } catch (e) {}
+        }
+        if (window.fdNavigate) window.fdNavigate(target);
+        else window.location.href = target;
+      },
+      // ── v7.1.1 : notes avec bloc AI Meeting Notes → section Meetings ──
+      meetingNotes: [],
+      loadMeetingNotes() {
+        var self = this;
+        return fetch('/api/v2/meetings/notes?limit=50', { credentials: 'same-origin' })
+          .then(function(r) { return r.ok ? r.json() : { notes: [] }; })
+          .then(function(d) { self.meetingNotes = (d && d.notes) || []; })
+          .catch(function() { self.meetingNotes = []; });
+      },
+      meetingNoteOpen(n) {
+        var target = '/pages/' + n.id;
         if (window.fdNavigate) window.fdNavigate(target);
         else window.location.href = target;
       },
@@ -1614,6 +1710,15 @@
         this.wsCtx.visible = false;
         if (!id) return;
         switch(action) {
+          case 'open':
+            if (this.wsCtx.isFolder) {
+              if (typeof this.navigateToFolder === 'function') this.navigateToFolder(id);
+            } else {
+              var target = '/pages/' + id;
+              if (window.fdNavigate) window.fdNavigate(target);
+              else window.location.href = target;
+            }
+            break;
           case 'new-file':
             this.newPageInFolder(id);
             break;
@@ -1632,6 +1737,7 @@
                 if (r.ok) {
                   if (window.FlowDeck && window.FlowDeck.syncPageTitle) window.FlowDeck.syncPageTitle(id, newName.trim());
                   self.refreshSidebarTree();
+                  if (typeof self.loadMeetingNotes === 'function') self.loadMeetingNotes();
                   document.dispatchEvent(new CustomEvent('flowdeck:workspace-changed'));
                   if (window._wsData && window._wsData._reloadAfterAction) {
                     window._wsData._reloadAfterAction();
@@ -1644,6 +1750,7 @@
             if (!id) break;
             if (!confirm('Delete "' + name + '"?')) break;
             this.deleteWorkspacePage(id, true);
+            if (typeof this.loadMeetingNotes === 'function') this.loadMeetingNotes();
             break;
         }
       },
@@ -1685,20 +1792,58 @@
           });
       },
 
-      // ── Sections collapsible ──
+      // ── Sections collapsible (accordéon v7.70 : une seule ouverte à la fois) ──
       sectionsOpen: (function() {
-        var def = { recents: true, private: true, meetings: true, favorites: true, agents: true, teamspaces: true, shared: true, published: true, apps: true, workspace: true, gitea: true };
+        var order = ['workspace','gitea','meetings','recents','favorites','agents','teamspaces','shared','published','private','apps'];
+        var def = { recents: false, private: false, meetings: false, favorites: false, agents: false, teamspaces: false, shared: false, published: false, apps: false, workspace: true, gitea: false };
         try {
           var saved = JSON.parse(localStorage.getItem('fd_sections') || '{}');
-          return Object.assign(def, saved);
+          var merged = Object.assign({}, def, saved);
+          // Normalise : si plusieurs sections ouvertes (ancien état), ne garder que la première.
+          var first = null;
+          for (var i = 0; i < order.length; i++) {
+            if (merged[order[i]]) { first = order[i]; break; }
+          }
+          if (first) {
+            for (var j = 0; j < order.length; j++) merged[order[j]] = (order[j] === first);
+          } else {
+            merged.workspace = true;
+          }
+          return merged;
         } catch(e) { return def; }
       })(),
       toggleSection(key) {
-        this.sectionsOpen[key] = !this.sectionsOpen[key];
+        var isOpen = !!this.sectionsOpen[key];
+        var order = ['workspace','gitea','meetings','recents','favorites','agents','teamspaces','shared','published','private','apps'];
+        if (isOpen) {
+          this.sectionsOpen[key] = false;
+        } else {
+          for (var i = 0; i < order.length; i++) this.sectionsOpen[order[i]] = false;
+          this.sectionsOpen[key] = true;
+        }
         this.sectionsOpen = Object.assign({}, this.sectionsOpen);
         try { localStorage.setItem('fd_sections', JSON.stringify(this.sectionsOpen)); }
         catch(e) {}
       },
+      // ── Menu "New" (popover "Create new" vers le haut) ──
+      newMenuOpen: false,
+      toggleNewMenu() { this.newMenuOpen = !this.newMenuOpen; },
+      closeNewMenu() { this.newMenuOpen = false; },
+      newMenuAction(kind) {
+        this.newMenuOpen = false;
+        if (kind === 'page') {
+          if (window.FlowDeck && window.FlowDeck.createPage) window.FlowDeck.createPage();
+        } else if (kind === 'chat') {
+          if (typeof this.chatNewConversation === 'function') this.chatNewConversation();
+          else if (typeof this.agentNew === 'function') this.agentNew();
+        } else if (kind === 'meeting') {
+          if (typeof this.fdCreateMeeting === 'function') this.fdCreateMeeting();
+          else if (typeof this.meetingNote === 'function') this.meetingNote();
+        } else if (kind === 'database') {
+          if (typeof this.createDatabase === 'function') this.createDatabase();
+          else if (window.showToast) window.showToast('Database creation: use a Gitea repo as a database.', 'info');
+        }
+      },
       // ── Sidebar customization ──
       customizing: false,
       sidebarConfig: {},
@@ -1933,6 +2078,50 @@
         });
       },
       fdCreatePage() { window.FlowDeck.createPage(); },
+      fdCreateMeeting() {
+        // v7.1.1 : crée une note de réunion vide avec son bloc
+        // AI Meeting Notes déjà inséré (même action que +, New, /meet).
+        var self = this;
+        var iso = new Date().toISOString().slice(0, 10);
+        var title = 'Meeting @' + iso;
+        fetch('/api/local-workspace/items', {
+          method: 'POST',
+          credentials: 'same-origin',
+          headers: { 'Content-Type': 'application/json', 'X-CSRF-Token': this.getCsrfToken() },
+          body: JSON.stringify({ name: title, type: 'page' })
+        })
+          .then(function(r) { if (!r.ok) throw new Error('page creation (' + r.status + ')'); return r.json(); })
+          .then(function(page) {
+            var block = {
+              id: 'b1_' + Date.now().toString(36), type: 'meeting', content: '',
+              meeting: {
+                v: 2, title: title, date: iso, state: 'idle', mode: 'mic_only',
+                instruction: 'auto', lang: 'fr', consent: null, consentAudio: false,
+                tab: 'notes', channels: [], segments: [], segSeq: 0, notes: '',
+                manual: '', interim: '', startedAt: 0, elapsedBase: 0, pausedMs: 0,
+                steps: [], summaryMd: '', completeness: '', genTitle: '',
+                titleEdited: false, shareHidden: false, bannerHidden: false,
+                transcriptId: null, error: '', remoteSilentWarned: false,
+                menuOpen: false, settingsOpen: false
+              }
+            };
+            // Amorçage du bloc : en cas d'échec la page s'ouvre quand même
+            // (l'utilisateur ajoute le bloc via /meet).
+            return fetch('/board/api/pages/' + page.id + '/blocks', {
+              method: 'POST',
+              credentials: 'same-origin',
+              headers: { 'Content-Type': 'application/json', 'X-CSRF-Token': self.getCsrfToken() },
+              body: JSON.stringify({ title: title, blocks: [block] })
+            }).then(function() { return page; }).catch(function() { return page; });
+          })
+          .then(function(page) {
+            if (typeof self.loadMeetingNotes === 'function') self.loadMeetingNotes();
+            if (typeof self.refreshSidebarTree === 'function') self.refreshSidebarTree();
+            if (window.fdNavigate) window.fdNavigate('/pages/' + page.id);
+            else window.location.href = '/pages/' + page.id;
+          })
+          .catch(function(err) { self.toast('Meeting note creation failed: ' + err.message, 'error'); });
+      },
       fdCreateFolder() { window.FlowDeck.showCreateFolderModal(); },
       fdGwRefresh() { if (window._gwData) window._gwData.refreshTree(); },
       sidebarSections() {
@@ -2296,6 +2485,12 @@
             const manageLink = ul.querySelector('li:last-child');
             ul.innerHTML = html;
             if (manageLink) ul.appendChild(manageLink);
+            // Le HTML injecté porte des directives Alpine (@contextmenu,
+            // x-show des dossiers, drag&drop) : les monter explicitement,
+            // sinon l'arbre perd menu contextuel et toggles après refresh.
+            try {
+              if (window.Alpine && window.Alpine.initTree) window.Alpine.initTree(ul);
+            } catch (e) { console.error('Sidebar Alpine remount failed:', e); }
           })
           .catch(e => console.error('Sidebar refresh failed:', e));
       },
@@ -2526,6 +2721,7 @@
         if (data.contextMenu) data.contextMenu.visible = false;
         if (data.wsCtx) data.wsCtx.visible = false;
         if (data.showNewPageMenu) data.showNewPageMenu = false;
+        if (data.newMenuOpen) data.newMenuOpen = false;
       }
     }
     else if (e.ctrlKey && e.key === 's') {

app/templates/gitea_workspace.html

@@ -11,9 +11,11 @@
 {% endblock %}
 
 {% block content %}
-<!-- Prism.js syntax highlighting -->
-<link rel="stylesheet" href="/static/css/prism.css">
-<script src="/static/js/prism.min.js"></script>
+<!-- highlight.js syntax highlighting (~230 file types via FDHighlight helper) -->
+<link rel="stylesheet" href="/static/css/highlight.css?v={{ asset_version }}">
+<script src="/static/js/highlight.min.js?v={{ asset_version }}"></script>
+<script src="/static/js/highlight-extra.min.js?v={{ asset_version }}"></script>
+<script src="/static/js/code-highlight.js?v={{ asset_version }}"></script>
 <style>
 .gw-main{display:flex;flex-direction:column;height:calc(100vh - 48px);overflow:hidden;}
 .gw-content{padding:24px;overflow-y:auto;flex:1;}
@@ -31,7 +33,7 @@
 .gw-commit-bar input{flex:1;min-width:180px;padding:8px 12px;background:var(--bg-tertiary);border:1px solid var(--border);border-radius:8px;color:var(--text);font-size:13px;outline:none;}
 .gw-commit-bar input:focus{border-color:var(--accent);}
 
-/* Prism theme tweaks for dark UI */
+/* highlight.js theme tweaks for dark UI */
 .gw-content pre[class*="language-"]{background:var(--bg-tertiary)!important;border:1px solid var(--border)!important;border-radius:8px!important;padding:16px!important;}
 </style>
 
@@ -62,7 +64,10 @@
         <div class="skeleton" style="height:15px;width:50%;"></div>
       </div>
       <!-- Syntax-highlighted code -->
-      <pre x-show="!fileLoading"><code :class="'language-'+fileLanguage" x-ref="codeBlock" x-text="fileContent"></code></pre>
+      <div class="gw-code-head" x-show="!fileLoading" style="display:flex;align-items:center;gap:8px;margin-bottom:8px;">
+        <span class="gw-code-lang" x-text="fileLanguageLabel" style="font-size:11px;font-weight:600;text-transform:uppercase;letter-spacing:.5px;color:var(--text-tertiary);"></span>
+      </div>
+      <pre x-show="!fileLoading"><code class="hljs" :class="'language-'+fileLanguage" :data-hljs-lang="fileLanguage" x-ref="codeBlock" x-text="fileContent"></code></pre>
     </div>
   </div>
 

app/templates/library.html

@@ -60,20 +60,38 @@
 /* ── Table (div-based for hover effects) ── */
 .lib-table{width:100%;}
 .lib-table-header{display:flex;align-items:center;border-bottom:1px solid var(--border);padding:0;}
-.lib-table-header .th{display:flex;align-items:center;gap:5px;padding:6px 8px;font-size:12px;font-weight:500;color:var(--text-dim);user-select:none;white-space:nowrap;}
+.lib-table-header .th{display:flex;align-items:center;gap:5px;padding:6px 8px;font-size:12px;font-weight:500;color:var(--text-dim);user-select:none;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;}
 .lib-table-header .th svg{width:13px;height:13px;flex-shrink:0;}
 
 /* Column widths — Notion-style */
 .lib-col-name{flex:1;min-width:0;}
-.lib-col-author{width:150px;flex-shrink:0;display:none;}
-.lib-col-source{width:120px;flex-shrink:0;display:none;}
+.lib-col-author{width:150px;flex-shrink:0;}
+.lib-col-editedby{width:150px;flex-shrink:0;}
+.lib-col-source{width:120px;flex-shrink:0;}
 .lib-col-edited{width:110px;flex-shrink:0;text-align:right;justify-content:flex-end;}
-.lib-table.show-author .lib-col-author{display:flex;}
-.lib-table.show-source .lib-col-source{display:flex;}
+.lib-col-visited{width:130px;flex-shrink:0;text-align:right;justify-content:flex-end;}
+/* Visibilité des colonnes pilotée par Show columns. Ces règles doivent
+   battre .lib-table-header .th et .lib-row .row-cell (spécificité 0,2,0) :
+   avant, seul :not(.show-editedtime) (0,3,0) gagnait, donc seule Last
+   edited time s'affichait/masquait — les autres options semblaient mortes. */
+.lib-table div.lib-col-author,
+.lib-table div.lib-col-editedby,
+.lib-table div.lib-col-source,
+.lib-table div.lib-col-visited{display:none;}
+.lib-table.show-author div.lib-col-author,
+.lib-table.show-editedby div.lib-col-editedby,
+.lib-table.show-source div.lib-col-source,
+.lib-table.show-visited div.lib-col-visited{display:flex;}
+.lib-table:not(.show-editedtime) div.lib-col-edited{display:none;}
+.row-source{display:inline-flex;align-items:center;gap:5px;}
+.row-source svg{width:13px;height:13px;flex-shrink:0;color:var(--text-dim);}
 
 /* ── Row ── */
 .lib-row{display:flex;align-items:center;height:34px;transition:background 0.1s;position:relative;}
 .lib-row:hover{background:var(--bg-hover);}
+/* ── Date section headers (AI Meeting Notes tab) ── */
+.lib-section-header{display:flex;align-items:center;font-size:12px;font-weight:600;color:var(--text-secondary);padding:10px 8px 4px;user-select:none;}
+.lib-section-header:first-child{padding-top:2px;}
 .lib-row.selected{background:rgba(35,131,226,0.08);}
 .lib-row.drag-over{outline:2px solid var(--accent);outline-offset:-2px;background:rgba(35,131,226,0.12);}
 .lib-row.dragging{opacity:0.4;}
@@ -189,7 +207,7 @@
 
 /* ── Responsive ── */
 @media(max-width:768px){
-  .lib-col-author,.lib-col-source{display:none!important;}
+  .lib-col-author,.lib-col-editedby,.lib-col-source,.lib-col-visited{display:none!important;}
   .peek-overlay{width:100vw;max-width:100vw;}
 }
 </style>
@@ -212,6 +230,9 @@
       <button class="lib-tab" :class="{ active: tab === 'recents' }" @click="switchTab('recents')">
         <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><circle cx="12" cy="12" r="9"/><polyline points="12 7 12 12 16 14"/></svg> Recents
       </button>
+      <button class="lib-tab" :class="{ active: tab === 'meetings' }" @click="switchTab('meetings')">
+        <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><rect x="3" y="4" width="18" height="18" rx="2" ry="2"/><line x1="16" y1="2" x2="16" y2="6"/><line x1="8" y1="2" x2="8" y2="6"/><line x1="3" y1="10" x2="21" y2="10"/></svg> AI Meeting Notes
+      </button>
       <button class="lib-tab" :class="{ active: tab === 'favorites' }" @click="switchTab('favorites')">
         <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><polygon points="12 2 15.09 8.26 22 9.27 17 14.14 18.18 21.02 12 17.77 5.82 21.02 7 14.14 2 9.27 8.91 8.26 12 2"/></svg> Favorites
       </button>
@@ -265,6 +286,9 @@
         <div class="lib-dropdown" x-show="viewOpen" @click.outside="viewOpen=false" x-transition>
           <div class="dd-label">Show columns</div>
           <div class="dd-item" @click="toggleColumn('author')"><span class="check" x-text="showAuthor ? '✓' : ''"></span> Created by</div>
+          <div class="dd-item" @click="toggleColumn('editedby')"><span class="check" x-text="showEditedBy ? '✓' : ''"></span> Last edited by</div>
+          <div class="dd-item" @click="toggleColumn('editedtime')"><span class="check" x-text="showEditedTime ? '✓' : ''"></span> Last edited time</div>
+          <div class="dd-item" @click="toggleColumn('visited')"><span class="check" x-text="showVisited ? '✓' : ''"></span> Last visited time</div>
           <div class="dd-item" @click="toggleColumn('source')"><span class="check" x-text="showSource ? '✓' : ''"></span> Source</div>
         </div>
       </div>
@@ -351,7 +375,7 @@
   </div>
 
   <!-- ── Table ── -->
-  <div class="lib-table" id="lib-table" :class="{'show-author': showAuthor, 'show-source': showSource}">
+  <div class="lib-table" id="lib-table" :class="{'show-author': showAuthor, 'show-editedby': showEditedBy, 'show-editedtime': showEditedTime, 'show-visited': showVisited, 'show-source': showSource}">
     <div class="lib-table-header">
       <div class="th lib-col-name">
         <input type="checkbox" style="width:14px;height:14px;accent-color:var(--accent);cursor:pointer;margin-left:2px;margin-right:6px;" :checked="allSelected" @click="toggleSelectAll()" title="Select / unselect all">
@@ -359,11 +383,16 @@
         Page name
       </div>
       <div class="th lib-col-author">Created by</div>
-      <div class="th lib-col-source">Source</div>
+      <div class="th lib-col-editedby">Last edited by</div>
       <div class="th lib-col-edited">
         <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><circle cx="12" cy="12" r="9"/><polyline points="12 7 12 12 16 14"/></svg>
         Last edited time
       </div>
+      <div class="th lib-col-visited">
+        <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><circle cx="11" cy="11" r="8"/><line x1="21" y1="21" x2="16.65" y2="16.65"/></svg>
+        Last visited time
+      </div>
+      <div class="th lib-col-source">Source</div>
     </div>
     <div class="lib-table-body" id="lib-table-body">
     </div>

app/templates/local_workspace.html

@@ -166,9 +166,12 @@
 .bc-edit-go:hover{opacity:.85;}
 
 /* ── Preserved styles ── */
+/* NOTE : scopées à .ws-tree (arbre principal) — la variante globale
+   écrasait aussi les items .ws-tree-item de la sidebar (macro partagée),
+   d'où un espacement différent sur /local-workspace. */
 .ws-tree{list-style:none;margin:0;padding:0;}
-.ws-tree-item{display:flex;align-items:center;gap:10px;padding:8px 12px;border-radius:6px;cursor:pointer;transition:background 150ms;border:1px solid transparent;margin-bottom:2px;}
-.ws-tree-item:hover{background:var(--bg-tertiary);border-color:var(--border);}
+.ws-tree .ws-tree-item{display:flex;align-items:center;gap:10px;padding:8px 12px;border-radius:6px;cursor:pointer;transition:background 150ms;border:1px solid transparent;margin-bottom:2px;}
+.ws-tree .ws-tree-item:hover{background:var(--bg-tertiary);border-color:var(--border);}
 .ws-tree-item .icon{font-size:16px;width:20px;text-align:center;flex-shrink:0;}
 .ws-tree-item .name{flex:1;font-size:14px;color:var(--text);}
 .ws-tree-item .name.folder{font-weight:500;cursor:pointer;}
@@ -179,6 +182,8 @@
 .ws-tree-item .name.file .inline-rename-input{flex:0 1 auto;}
 .ws-tree-item .name.file .btn-open{flex:0 0 auto;}
 .ws-shared-badge{flex-shrink:0;margin-right:4px;font-size:13px;line-height:1;opacity:.9;filter:grayscale(1) brightness(1.6);}
+.ws-fav-badge{flex-shrink:0;margin-right:4px;font-size:13px;line-height:1;color:#f5a623;}
+.ws-pub-badge{flex-shrink:0;margin-right:4px;font-size:13px;line-height:1;}
 .ws-tree-item .actions{display:none;gap:2px;}
 .ws-tree-item:hover .actions{display:flex;}
 .ws-chevron{width:16px;height:16px;flex-shrink:0;display:inline-flex;align-items:center;justify-content:center;background:none;border:none;color:var(--text-tertiary);cursor:pointer;font-size:8px;padding:0;border-radius:3px;transition:transform 150ms,color 100ms;}
@@ -480,7 +485,11 @@
 .preview-leave-end{transform:translateX(100%);opacity:0;}
 </style>
 
-<script type="application/json" id="lw-config" nonce="{{ csp_nonce() }}">{{ {"current_folder_id": current_folder_id, "workspace_id": workspace_id} | tojson }}</script>
+<script type="application/json" id="lw-config" nonce="{{ csp_nonce() }}">{{ {"current_folder_id": current_folder_id, "workspace_id": workspace_id, "workspace_name": workspace_name} | tojson }}</script>
+<link rel="stylesheet" href="/static/css/highlight.css?v={{ asset_version }}">
+<script data-cfasync="false" src="/static/js/highlight.min.js?v={{ asset_version }}"></script>
+<script data-cfasync="false" src="/static/js/highlight-extra.min.js?v={{ asset_version }}"></script>
+<script data-cfasync="false" src="/static/js/code-highlight.js?v={{ asset_version }}"></script>
 <script data-cfasync="false" src="/static/js/local_workspace.js?v={{ asset_version }}"></script>
 <script data-cfasync="false" nonce="{{ csp_nonce() }}">console.log('FLOWDECK v2.4.6 — ws-split about to render, _wsInitData exists:', typeof _wsInitData !== 'undefined');</script>
 
@@ -528,7 +537,6 @@
     <a href="/workspaces" class="btn-icon nav-arrow" title="All workspaces">{{ fd_icon("home",14) }}</a>
     <button class="btn-icon nav-arrow" @click="goBack()" title="Back" :disabled="!canGoBack">{{ fd_icon("chevron-left",14) }}</button>
     <button class="btn-icon nav-arrow" @click="goForward()" title="Forward" :disabled="!canGoForward">{{ fd_icon("chevron-right",14) }}</button>
-    {% if breadcrumbs %}
     <div class="unified-breadcrumb">
       <span class="bc-icon">{{ fd_icon("folder",13) }}</span>
       <a href="/local-workspace" @click.prevent="navigateToRoot()" class="bc-root">{{ workspace_name }}</a>
@@ -549,7 +557,6 @@
       </div>
       <button class="bc-edit-trigger" x-show="!editingPath" @click="startPathEdit()" title="Edit path">{{ fd_icon("edit",11) }}</button>
     </div>
-    {% endif %}
     <div class="ws-toolbar" style="margin-left:auto;">
       <!-- View dropdown -->
       <div class="lib-toolbar-wrap" style="position:relative;">
@@ -729,6 +736,8 @@
             <span class="name folder" @click.stop="navigateToFolder(node.id)"
                   @dblclick.prevent.stop="startInlineRename(node, $event)">
               <span x-show="node.is_shared" class="ws-shared-badge" title="Partagé" x-text="'👥'"></span>
+              <span x-show="node.favorited" class="ws-fav-badge" title="Favorite">★</span>
+              <span x-show="node.published" class="ws-pub-badge" title="Published to web">🌐</span>
               <span x-show="renamingId !== node.id" x-text="node.name"></span>
               <span class="child-count" x-show="node.child_count > 0 && renamingId !== node.id"
                     x-text="'(' + node.child_count + ')'"></span>
@@ -743,6 +752,8 @@
                   @mouseleave="hidePreview()"
                   title="Click to rename, double-click to open">
               <span x-show="node.is_shared" class="ws-shared-badge" title="Partagé" x-text="'👥'"></span>
+              <span x-show="node.favorited" class="ws-fav-badge" title="Favorite">★</span>
+              <span x-show="node.published" class="ws-pub-badge" title="Published to web">🌐</span>
               <span x-show="renamingId !== node.id" x-text="node.name"></span>
               <input x-show="renamingId === node.id" class="inline-rename-input" x-model="renamingName"
                      @keydown.enter="commitRename(node)" @keydown.escape="cancelRename()"

app/templates/public_page.html

@@ -173,8 +173,17 @@
 </footer>
 
 <script src="/static/js/katex.min.js?v=0.16.11"></script>
+<link rel="stylesheet" href="/static/css/highlight.css?v={{ asset_version }}">
+<script src="/static/js/highlight.min.js?v={{ asset_version }}"></script>
+<script src="/static/js/highlight-extra.min.js?v={{ asset_version }}"></script>
+<script src="/static/js/code-highlight.js?v={{ asset_version }}"></script>
 <script data-cfasync="false" nonce="{{ csp_nonce() }}">
 document.addEventListener('DOMContentLoaded', function () {
+  // hljs — syntax-highlight code blocks (~230 file types via FDHighlight).
+  try {
+    if (window.FDHighlight) FDHighlight.highlightAllIn(document.querySelector('.pub-content') || document);
+    else if (window.hljs) document.querySelectorAll('.pub-content pre code').forEach(function (el) { try { window.hljs.highlightElement(el); } catch (e) {} });
+  } catch (e) {}
   if (window.katex) {
     document.querySelectorAll('div[data-katex]').forEach(function (el) {
       var tex = el.getAttribute('data-katex') || '';

docs/DATA_MODEL.md

@@ -0,0 +1,333 @@
+# FlowDeck — Modèle de données (v7.69.8)
+
+Document de référence du schéma de persistance, généré à partir du code réel :
+
+- `app/db.py` — schéma de base (baseline, « version 1 »), créé de façon idempotente par `init_db()` (lignes 14–839) et connexion SQLite via `get_conn()` (lignes 843–858) ;
+- `app/migrations.py` — migrations versionnées (1–38, table `schema_version`) ;
+- `app/services/webhook_outbound.py` — table `webhook_subscriptions` (créée au boot, l. 272–286) ;
+- `app/models/requests.py` / `app/models/responses.py` — couche Pydantic de l'API.
+
+Total : **105 tables relationnelles + 1 table virtuelle FTS5** (`pages_fts`), réparties entre le baseline (`app/db.py`) et 33 migrations effectuant des créations de tables (`app/migrations.py`).
+
+---
+
+## 1. Moteur de base de données
+
+### 1.1 SQLite uniquement (pas de dual-engine)
+
+**Correction vis-à-vis de certaines notes d'architecture : FlowDeck ne supporte PAS PostgreSQL.** Le code est exclusivement SQLite :
+
+- `app/db.py` l.4 : `import sqlite3` (stdlib) — aucune dépendance `psycopg`/`pg8000`/SQLAlchemy nulle part (`requirements.txt`, `pyproject.toml` vérifiés) ;
+- le module entier est typé `sqlite3.Connection` (y compris `app/migrations.py` l.16) ;
+- fonctionnalités SQLite spécifiques utilisées : `PRAGMA journal_mode=WAL`, `PRAGMA foreign_keys=ON`, `PRAGMA busy_timeout=5000` (`get_conn`, l.846–850), `INTEGER PRIMARY KEY AUTOINCREMENT`, `ALTER TABLE … ADD COLUMN` (sans `IF NOT EXISTS`, d'où le pattern `try/except sqlite3.OperationalError`), déclencheurs (`TRIGGER`), extension **FTS5**, `DROP TABLE`/`RENAME` pour les reconstructions (`comments`, `tags`) ;
+- `app/config.py` l.58 : `database_url: str = "sqlite:////data/flowdeck.db"` — seul le préfixe `sqlite:///` est interprété par la propriété `db_path` (l.151–163, avec gestion `:memory:` et des chemins Windows type `C:\…`) ; toute autre valeur retombe sur `/data/flowdeck.db`.
+
+La « compatibilité » mentionnée dans certains documents est donc en réalité : une **base SQLite unique par instance**, en mode WAL (un writer, multi-readers), avec `PRAGMA busy_timeout=5000` pour supporter la concurrence (tests xdist, scheduler d'automations). Les backups (v5.2.0, `backup_*` dans `app/config.py`) sont des copies du fichier SQLite.
+
+### 1.2 Stratégie de migrations
+
+Décrite dans l'en-tête de `app/migrations.py` (l.1–13) :
+
+| Composant | Rôle |
+|---|---|
+| `app/db.py::init_db()` | Schéma **baseline** (« version 1 ») : `CREATE TABLE IF NOT EXISTS` + ALTER garde-fous (`try/except OperationalError`). Réexécuté à chaque boot, idempotent. |
+| `schema_version` (`version INTEGER PK, name TEXT, applied_at`) | Registre des migrations appliquées ; version maximale lue par `current_version()` (l.66–71). Scellée à `BASELINE_VERSION = 1` au premier démarrage (l.86–101). |
+| `@register(version, name)` (l.28–38) | Décorateur d'enregistrement dans `MIGRATIONS`, trié par version, doublons refusés. |
+| `apply_migrations(conn)` (l.86–111) | Appelée en fin de `init_db()` (`app/db.py` l.829–831) ; applique exactement une fois chaque migration `> version courante`. |
+| `_apply_one()` (l.113–130) | **Une migration = une transaction** (`BEGIN` explicite, `rollback` complet à l'échec, puis `INSERT INTO schema_version`) — DDL tout-ou-rien (audit A31). |
+| `columns(conn, table)` (l.53–64) | Helper unique `PRAGMA table_info` pour tester l'existence de colonnes avant un ALTER. |
+| `fts5_available()` (l.74–83) | Détection de FTS5 ; sinon repli sur recherche `LIKE`. |
+
+Version de schéma actuelle : **38** (`v7.1.1 meeting block`, dernier `@register`). Les migrations 4 et 17 et 23–38 ont été enregistrées hors ordre chronologique de version (versions 4, 17, 23 insérées après 21) — le tri à l'enregistrement garantit l'ordre d'application.
+
+---
+
+## 2. Tables par domaine
+
+Légende : **PK** = clé primaire, **FK** = clé étrangère (avec cascade quand indiquée). Sauf mention, PK = `id INTEGER PK AUTOINCREMENT`.
+
+### 2.1 Kanban & intégration Gitea (héritage v0.x–v1.x)
+
+Créées dans `app/db.py` (l.22–140, 430–478).
+
+| Table | But | Clés / colonnes clés |
+|---|---|---|
+| `boards` | Board Kanban par dépôt Gitea | `UNIQUE(project_owner, project_name)` ; `columns_json` (liste des colonnes), `wip_limits_json` |
+| `cards` | Position d'une issue Gitea dans une colonne | FK `board_id → boards` ; `gitea_issue_id`, `column_name`, `position`, `priority`, `due_date` |
+| `col_mapping` | Mapping colonne ↔ label Gitea | FK `board_id → boards` ; `UNIQUE(board_id, column_name)` ; `close_issue` |
+| `notes` | Notes par projet | `UNIQUE(project_owner, project_name, title)` ; `content` |
+| `checklists` | Checklists par issue | `board_id`, `gitea_issue_id` (pas de FK déclarée), `title`, `position`, `UNIQUE(board_id, gitea_issue_id, title)` |
+| `checklist_items` | Lignes de checklist | FK `checklist_id → checklists ON DELETE CASCADE` ; `content`, `checked`, `position` |
+| `project_properties` | Propriétés par projet (ancêtre de `collection_properties`, conservé) | `UNIQUE(project_owner, project_name, name)` ; `prop_type`, `options_json` |
+| `property_values` | Valeurs de ces propriétés par issue | FK `property_id → project_properties ON DELETE CASCADE` ; `UNIQUE(property_id, gitea_issue_id)` |
+| `ai_keywords` | Fréquentité des mots-clés détectés par l'IA | `UNIQUE(project_owner, project_name, keyword)` ; `color`, `usage_count` |
+| `gitea_private_pages` | Pages privées rattachées à un dépôt Gitea (v2.7) | FK `user_id → users` ; `gitea_owner`, `gitea_repo`, `content` |
+| `user_tokens` | Token Gitea par utilisateur | `gitea_user_id INTEGER UNIQUE`, `gitea_token` |
+| `projects` | Registre forge-agnostique des projets (gitea/github/builtin), synchronisé par cron (mig. 6, `migrations.py` l.260–331) | `UNIQUE(proj_type, owner, name)` ; `clone_url`, `default_branch`, `last_synced_at` |
+
+### 2.2 Utilisateurs & authentification
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `users` | Compte local ou fédéré | `login UNIQUE`, `email`, `password_hash`, `is_admin`, `is_active`, `login_attempts`, `locked_until`, `auth_method` (local\|gitea), `avatar_url/color`, `sidebar_config` (JSON), `notification_prefs` (JSON), `timezone`, `totp_secret_enc`, `totp_backup_hashes` (JSON) | db.py l.22 ; ALTER db.py l.328–470, mig. 10/11/28 |
+| `login_history` | Journal des connexions | FK `user_id → users` ; `ip_address`, `user_agent` | db.py l.40 |
+| `user_oauth_tokens` | Tokens OAuth par provider (Gitea, Google, M365…) | FK `user_id → users` ; `UNIQUE(user_id, provider)` ; `access_token`, `refresh_token`, `expires_at` | db.py l.55 |
+| `user_sessions` | Sessions cookie révocables (une ligne = un cookie signé) | **PK `id TEXT`** (payload du cookie) ; FK `user_id → users ON DELETE CASCADE` ; `revoked`, `last_seen_at` | mig. 6 (l.294–316) |
+| `api_tokens` | Tokens Bearer publics (sha256 stocké) | FK `user_id → users ON DELETE CASCADE` ; `token_hash UNIQUE`, `token_prefix`, `scopes`, `expires_at`, `revoked` | mig. 6 + mig. 20 (l.884–900) |
+| `webauthn_credentials` | Passkeys / 2FA WebAuthn | FK `user_id → users ON DELETE CASCADE` ; `credential_id UNIQUE`, `public_key`, `sign_count` | mig. 28 (l.1311–1324) |
+| `sso_config` | Config SSO entreprise (SAML 2.0 ou OIDC), la table prime sur `.env` dès qu'un admin sauvegarde | FK `workspace_id → workspaces` ; secrets chiffrés au repos (`client_secret`, `sp_private_key` — `app/services/sso_provisioning.py`) ; `attribute_mapping`/`groups_mapping` (JSON), `active` | mig. 23 (l.1495–1523) |
+| `sso_login_history` | Audit de chaque tentative SSO (succès + rejets) | FK `user_id → users` ; `success`, `error_message`, `sso_identifier` | mig. 23 (l.1525–1545) |
+| `sso_requests` | Store mono-usage anti-replay (AuthnRequest ID, état OIDC, PKCE, relay state) | **PK `id TEXT`** ; `kind`, `code_verifier`, `next_path`, `used` | mig. 23 (l.1547–1555) |
+| `scim_tokens` | Tokens Bearer de provisioning SCIM (hash unique) | `token_hash UNIQUE`, `revoked`, FK `created_by → users` | mig. 28 (l.1285–1294) |
+| `domain_claims` | Domaines d'entreprise vérifiés (DNS/txt), auto-jonction + `enforce_sso` | `domain UNIQUE`, `txt_token`, FK `workspace_id → workspaces` | mig. 28 (l.1296–1307) |
+| `llm_config` | Config LLM globale de l'instance — **ligne unique** `CHECK (id = 1)` | `provider`, `model`, `api_key`, `api_base`, `verified`, `verified_model`, `last_error` ; précédence : ligne DB > `settings.llm_*` | db.py l.769–784 |
+| `user_llm_keys` | Clés API LLM par utilisateur et provider | FK `user_id → users ON DELETE CASCADE` ; `UNIQUE(user_id, provider)` ; `models_json` (cache des modèles), `default_model`, `verified*` | db.py l.788–812 |
+
+### 2.3 Workspaces, collaboration & permissions
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `workspaces` | Espace de travail multi-utilisateur | FK `owner_id → users` ; `settings_json` | db.py l.222 |
+| `workspace_members` | Adhésion + rôle (`owner/admin/editor/viewer/commenter`) | FK `workspace_id → workspaces CASCADE`, `user_id → users` ; `UNIQUE(workspace_id, user_id)` | db.py l.230 |
+| `teamspaces` | Namespace de pages/bases dans un workspace (v7.3.0) ; `private=1` = invisibilité 404 | FK `workspace_id → workspaces CASCADE` ; `UNIQUE(workspace_id, name)` | mig. 29 (l.1380–1391) |
+| `teamspace_members` | Membres + rôle d'un teamspace | FK `teamspace_id → teamspaces CASCADE`, `user_id → users` ; `UNIQUE(teamspace_id, user_id)` | mig. 29 (l.1393–1403) |
+| `user_groups` | Groupes réutilisables par workspace (ACL) | FK `workspace_id → workspaces CASCADE` ; `UNIQUE(workspace_id, name)` | mig. 18 (l.721–730) |
+| `group_members` | Adhésion à un groupe | FK `group_id → user_groups CASCADE`, `user_id → users CASCADE` ; `UNIQUE(group_id, user_id)` | mig. 18 (l.732–740) |
+| `page_permissions` | Grant explicite sur une page de l'éditeur | FK `page_id → pages CASCADE` ; `user_id XOR group_id` (`CHECK`) ; `role` (viewer/commenter/editor/owner), `grant_type`, `granted_by` | mig. 18 (l.745–760) |
+| `collection_permissions` | Idem pour les bases (collections) | FK `collection_id → collections CASCADE` ; mêmes colonnes | mig. 18 (l.766–780) |
+| `property_permissions` | Grant par propriété de collection (view/edit) | FK `collection_id`, `property_id → collection_properties CASCADE` | mig. 18 (l.782–797) |
+| `permission_audit_log` | Traînée immuable grant/revoke | `resource_type`, `resource_id`, `action`, `old_role`, `new_role`, `performed_by → users`, `ip_address` | mig. 18 (l.799–815) |
+| `page_shares` | Partage de page (user, groupe ou email) | FK `page_id → pages CASCADE`, `shared_with_user_id → users`, `shared_with_group_id → user_groups` ; `permission` (view/edit/comment) | db.py l.451–478 |
+| `guest_shares` | Accès sans compte via `/g/<token>` (v7.3.0) | FK `page_id → pages CASCADE` ; `token UNIQUE`, `role`, `expires_at`, `revoked` | mig. 29 (l.1440–1452) |
+| `favorites` | Favoris d'un utilisateur (reconstruite en v2.2 pour pointer `pages`) | FK `user_id → users`, `page_id → pages` ; `UNIQUE(user_id, page_id)` | db.py l.260–270 |
+| `recents` | Pages récemment accédées (vue « Recents ») | FK `user_id → users`, `page_id → pages` ; `source_type` (local/gitea…), `UNIQUE(user_id, page_id)` | db.py l.481–493 |
+| `notifications` | Notifications in-app (mention/commentaire/page) | FK `user_id → users CASCADE`, `actor_id → users` ; `ntype`, `resource_type/resource_id`, `is_read` ; index `(user_id, is_read)` | db.py l.592–613 |
+| `comments` | Commentaires (fil, résolus) ancrés sur pages ou lignes de base — **reconstruite** v4.9 : cible polymorphe | FK `user_id → users` ; `target_type` (`page` \| `collection_page`) + `target_id`, `parent_id` (réponses), `anchor_block_id/anchor_start/anchor_end` (ancre inline) | db.py l.241 (v2.0) puis l.620–649 (v4.9) |
+| `comment_reactions` | Réactions emoji d'un commentaire | FK `comment_id → comments CASCADE`, `user_id → users CASCADE` ; `UNIQUE(comment_id, user_id, emoji)` | mig. 29 (l.1418–1428) |
+| `text_reactions` | Réactions ancrées sur une plage de texte d'un bloc (façon Notion, v7.64) | FK `page_id → pages CASCADE`, `user_id → users CASCADE` ; `block_id TEXT`, offsets `anchor_start/anchor_end`, `UNIQUE(page_id, block_id, start, end, emoji, user_id)` | mig. 37 (l.1683–1707) |
+| `page_follows` | Abonnement aux changements d'une page | PK composite `(page_id → pages CASCADE, user_id → users CASCADE)` | mig. 29 (l.1430–1437) |
+| `page_verifications` | Badge « page vérifiée ✅ » avec expiration | **PK `page_id → pages CASCADE`** (1 par page) ; `verified_by → users`, `expires_at` | mig. 29 (l.1407–1416) |
+| `tags` | Étiquettes par utilisateur (`user_id=0` = globales) | FK `user_id → users` ; `UNIQUE(name, user_id)` ; `color` | db.py l.282–290 |
+| `page_tags` | Association page ↔ tag | PK composite `(page_id → pages CASCADE, tag_id → tags CASCADE)` | db.py l.292–299 |
+| `custom_emojis` | Emojis personnalisés uploadés par workspace | FK `workspace_id` (défaut 1, **non déclaré**) ; `name`, `url` | mig. 8 (l.371–388) |
+
+### 2.4 Pages & éditeur de blocs
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `pages` | Page de l'éditeur (wiki) — table pivot du produit | FK `workspace_id → workspaces`, `parent_id → pages` (hiérarchie), `collection_id → collections` (base plein écran, v4.6), `collection_row_id → collection_pages CASCADE` (page-ombre de contenu d'une ligne de base, v6.5, mig. 22 l.997–1020), `locked_by → users` ; colonnes notables : `content` (markdown legacy **ou JSON de blocs** selon `content_format`), `content_format`, `parent_section`, `share_mode`, `published`/`is_published`/`publish_slug`/`is_shared`, `deleted_at` (corbeille), `sort_order`, `cover_url`, `page_icon`, `is_locked`, `full_width`, `font_small`, `search_excluded`, `permission_type` (inherit/restricted/private), `teamspace_id` (ALTER, **sans FK déclarée**), `sync_version` (trigger de concurrence optimiste, mig. 17) | db.py l.154–166 + ALTER l.328–470 ; mig. 7/13/18/22/24/25/29 |
+| `page_versions` | Snapshots annulables de l'éditeur de blocs (undo/restore, une ligne par save) | FK `page_id → pages CASCADE` ; `title`, `blocks_json` (liste complète des blocs), `note`, `user_id → users` | mig. 7 (l.336–365) |
+| `page_history` | Historique legacy des **lignes de base** (`collection_pages`) | FK `page_id → collection_pages CASCADE` ; `change_type`, `snapshot_json` | db.py l.249 |
+| `page_global_templates` | Modèles de page globaux créés par les utilisateurs | `blocks_json` (format éditeur), FK `created_by → users SET NULL` | mig. 13 (l.582–597) |
+| `synced_blocks` | Bloc synchronisé : source de vérité du contenu, réutilisé sur plusieurs pages | `content` (JSON blocs), `created_by → users` (sans FK), `workspace` (TEXT), `synced` via `page_synced_blocks` | mig. 15 (l.626–662) |
+| `page_synced_blocks` | Référence page ↔ bloc synchronisé (désynchronisable par page) | FK `page_id → pages CASCADE`, `synced_block_id → synced_blocks CASCADE` ; `block_index`, `UNIQUE(page_id, synced_block_id)` | mig. 15 (l.646–662) |
+| `page_views` | Compteur de vues quotidien par page (même pattern que `site_views`) | FK `page_id → pages CASCADE` ; PK composite `(page_id, day)`, `views` | mig. 29 (l.1454–1461) |
+| `pages_fts` | **Table virtuelle FTS5** (`title`, `body`) pour la recherche plein texte de la palette de commandes ; synchronisée par 3 triggers `AFTER INSERT/UPDATE/DELETE` sur `pages` ; backfill au boot ; repli `LIKE` si FTS5 absent | rowid = `pages.id` | mig. 3 (l.156–205) |
+
+### 2.5 Bases de données (collections, style Notion)
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `collections` | Base de données (collection) | `schema_json` (schéma des propriétés), FK `gitea_owner/gitea_repo` (texte, verrouillage sync), `workspace_id → workspaces`, `created_by → users`, `is_locked`, `is_inline` + `parent_page_id` (base inline), `is_task` + `task_assignee_prop/task_status_prop/task_due_prop → collection_properties SET NULL` (mapping My Tasks, mig. 30 l.1558–1585), `permission_type`, `teamspace_id` (sans FK), `form_config_json`, `sync_version` | db.py l.172 + ALTER db.py l.495–520, mig. 18/24/29/30 |
+| `collection_pages` | **Ligne** d'une base (row) | FK `collection_id → collections CASCADE`, `parent_id → collection_pages` (hiérarchie interne), `property_values_json` (valeurs par propriété), `gitea_issue_id/gitea_issue_number` (sync Kanban), `cover_url`, `external_event_id` (calendrier, mig. 27), `permission_type`, `sync_version` | db.py l.187 ; mig. 10/18/27 |
+| `collection_views` | Vues sauvegardées par base (table/board/calendar/gallery/list…) | FK `collection_id → collections CASCADE` ; `view_type`, **`config_json`** (filtres/tri/groupement/columns visibles), `position`, `created_by` (NULL = vue partagée, sinon personnelle — mig. 10) | db.py l.210 ; mig. 10 (l.472–478) |
+| `collection_properties` | Schéma des propriétés d'une base (remplace `project_properties`) | FK `collection_id → collections CASCADE` ; `prop_type`, `options_json` (select/multi-select), `number_format`, relationnels : `related_collection_id → collections SET NULL`, `relation_property_id/target_property_id → collection_properties SET NULL`, `rollup_function`, `formula_expression`, `validation_json` (mig. 4), `group_name` (mig. 10), `button_automation_id → automations SET NULL` (bouton natif, mig. 26) ; `UNIQUE(collection_id, name)` | db.py l.222–241 ; mig. 4/10/26 |
+| `collection_data_sources` | Sources de données liées (base liée type Notion « Linked database ») | FK `collection_id → collections CASCADE`, `source_collection_id → collections` ; `is_linked`, `UNIQUE(collection_id, source_collection_id)` | db.py l.480–493 |
+| `collection_dashboards` | Dashboard = combinaison de vues/widgets sur une page | FK `collection_id → collections CASCADE` ; **`layout_json`** (`{"columns":1,"widgets":[]}`) | db.py l.531–542 |
+| `database_templates` | Modèles de bases prêts à l'emploi (seed idempotent, v5.3.0 : Meeting notes, etc.) | `name`, `icon`, `description`, `schema_json` | db.py l.272 + mig. 4 (l.390–411) ; seed via `app/services/db_templates.py::SEED_TEMPLATES` |
+| `page_templates` | Modèles de **ligne** par base, éventuellement récurrents | FK `collection_id → collections CASCADE` ; `property_values_json`, `content_json` (blocs), `is_recurring` + `recurrence_rule` | db.py l.294–303 + ALTER l.512–522 |
+| `page_dependencies` | Dépendances entre lignes (bloque / bloqué par) | FK `page_id`, `dependency_id → collection_pages CASCADE` ; `dependency_type`, `auto_shift` (overlap/…) ; `UNIQUE(page_id, dependency_id, dependency_type)` | db.py l.551–563 |
+| `sprints` | Sprints Scrum par base de tâches | FK `collection_id → collections CASCADE` ; `start_date/end_date` (TEXT), `status` (planning/active/done), `goal`, `auto_complete` | db.py l.570–582 |
+| `sprint_pages` | Lignes affectées à un sprint + vélocité | FK `sprint_id → sprints CASCADE`, `page_id → collection_pages CASCADE` ; `velocity_points`, `status_at_start`, `UNIQUE(sprint_id, page_id)` | db.py l.584–594 |
+| `reminder_log` | Journal anti-doublon des rappels : 1 ligne par (ligne de base, date d'occurrence) — le rappel ne fire qu'une fois même après restart | FK `page_id → collection_pages CASCADE` ; `UNIQUE(page_id, occurrence_date)` ; alimenté par le scan `reminders_enabled` (`app/config.py` l.75–77) — **il n'existe pas de table `reminders`** ; les dates de rappel proviennent des propriétés `due` des lignes | mig. 11 (l.485–509) |
+
+### 2.6 Automatisation, workers & webhooks
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `automations` | Règles if-this-then-that (moteur v5.1.0) : `trigger_type` event/cron/button, `event`, `cron_expression`, `collection_id` (TEXT sans FK, scope optionnel), **`condition_json`** (clauses), **`actions_json`** (descripteurs), `enabled`, `run_count`, `last_run_at`, `trigger_mode` (`any`\|`all`, fenêtre de 5 min, mig. 26) | | mig. 5 (l.208–257) + mig. 26 (l.1196–1201) |
+| `automation_steps` | Étapes ordonnées d'une automatisation v2 (trigger/condition/delay/action, v7.0.0) ; l'engine retombe sur `condition_json/actions_json` si aucune step | FK `automation_id → automations CASCADE` ; `kind`, `position`, `config_json` | mig. 26 (l.1156–1172) |
+| `automation_runs` | Historique d'exécution (audit + UI Settings) | FK `automation_id → automations CASCADE` ; `status` (fired/skipped/error), `trigger_source`, `collection_id`/`page_id` (TEXT/INT sans FK déclarée), `detail` | mig. 5 (l.242–257) |
+| `workers` | Snippets Python sandboxés (cron ou manuel), partageables (v7.0.0) | `slug UNIQUE`, FK `workspace_id → workspaces CASCADE`, `code_py`, `schedule_cron`, `shared`, `daily_budget_s` (budget CPU quotidien), `created_by → users` (sans FK déclarée) | mig. 26 (l.1174–1192) |
+| `worker_runs` | Logs d'exécution des workers | FK `worker_id → workers CASCADE` ; `status`, `logs`, `duration_ms` | mig. 26 (l.1194–1210) |
+| `webhook_subscriptions` | Abonnement sortant URL + event + secret (créée au boot hors migrations, appelée par `init_db` l.834–837) | `url`, `event`, `secret`, `active` | `app/services/webhook_outbound.py` l.272–286 |
+| `webhook_deliveries` | Journal de livraison (retry ledger) | FK `webhook_id → webhook_subscriptions CASCADE` ; `status` (pending/retrying/success/error/superseded), `http_code`, `attempt`, `event`, `next_retry_at` (epoch, scheduler v6.4), `payload` (JSON), `duration_ms` | mig. 20 (l.913–928) + mig. 21 (l.940–957) |
+| `idempotency_keys` | Support d'`Idempotency-Key` sur les POST de l'API v2 (réponse rejouée) | **PK `key TEXT`** ; FK `user_id → users CASCADE` ; `response_json`, `status_code` | mig. 20 (l.930–938) |
+
+### 2.7 IA / Agent FlowDeck
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `agents` | Définition d'agent (personnel ou workspace) | FK `workspace_id → workspaces`, `created_by → users` ; `agent_type`, `system_instructions`, `model`, `scope_json`, `trigger_json`, `approval_mode` (auto/gated), `is_active`, `UNIQUE(workspace_id, name)` | db.py l.660–680 |
+| `agent_conversations` | Conversation d'un utilisateur avec un agent | FK `agent_id → agents CASCADE`, `user_id → users` ; `status`, `context_json`, override `provider`/`model` (ALTER db.py l.820–827), `memory_enabled` (mig. 31) | db.py l.681–691 |
+| `agent_messages` | Messages (rôles user/assistant/tool), coût tokens | FK `conversation_id → agent_conversations CASCADE` ; `tool_calls_json`, `model`, `tokens_used` | db.py l.692–702 |
+| `agent_actions` | Journal des tool calls exécutés (audit + undo) | FK `conversation_id → … CASCADE` ; `tool_name`, `target_type/target_id`, `payload_json`, `result_json`, `status`, `undo_snapshot_json`, `executed_by → users` | db.py l.703–716 |
+| `agent_skills` | Skills réutilisables (prompt + outils autorisés) | FK `workspace_id → workspaces`, `created_by → users` ; `prompt_template`, `allowed_tools_json`, `UNIQUE(workspace_id, name)` | db.py l.717–729 |
+| `agent_triggers` | Déclencheurs d'agent (manual/cron/événement) | FK `agent_id → agents CASCADE` ; `trigger_type`, `config_json`, `is_active`, `last_fired_at` | db.py l.730–740 |
+| `agent_feedback` | 👍/👎 sur les réponses (rating CHECK up/down) | FK `conversation_id`/`message_id`/`user_id` en `SET NULL` ; `snippet`, `comment` | db.py l.742–756 |
+| `agent_policies` | Gouvernance par workspace (1 ligne/workspace) | FK `workspace_id → workspaces CASCADE`, `UNIQUE(workspace_id)` ; `allowed_tools_json`, `max_steps`, `require_approval` | mig. 28 (l.1326–1338) |
+| `agent_approvals` | File d'approbation des actions d'écriture gated | `conversation_id` (INT **sans FK déclarée**), `tool`, `args_json`, `status` (pending/…), `requester_id`/`approver_id → users` | mig. 28 (l.1340–1354) |
+| `agent_memory` | Résumé mémorisé par conversation (upsert 1:1, v7.54.0) | **PK `conversation_id → agent_conversations CASCADE`** ; `kind` (summary), `content` | mig. 31 (l.1661–1681) |
+| `agent_connectors` | Catalogue de connecteurs d'agent (custom/Discord/Telegram/MCP, v7.55–7.57) | `name`, `url`, `secret_encrypted` (Fernet), `kind`, `auth` (bearer/…), `tools_json` (cache d'outils MCP), `enabled`, `status`, `created_by` (sans FK) | mig. 32 (l.1640–1659) + mig. 34 (l.1613–1624) |
+| `connector_tokens` | Tokens OAuth par (kind, user) — Google / M365, chiffrés Fernet | **PK composite `(kind, user_id)`** ; `tokens_enc` | mig. 33 (l.1626–1638) |
+| `plugins` | Registre des modules activables/désactivables de l'instance (web-tools, web-clipper, automations) — seed idempotent | **PK `slug TEXT`** ; `enabled` | mig. 35 (l.1589–1611) |
+| `semantic_embeddings` | Recherche sémantique hybride + Ask AI (v6.9) : vecteurs « hashed-TF » 256 dims **sans dépendance externe** ; chunks par ressource polymorphe | **PK composite `(resource_type, resource_id, chunk_id)`** (page \| collection…) ; `chunk_text`, `embedding BLOB`, `model` | mig. 25 (l.1096–1133) |
+| `semantic_index_state` | État d'indexation incrémentale (dernier index par ressource) | **PK composite `(resource_type, resource_id)`** ; `indexed_at` | mig. 25 (l.1122–1133) |
+
+### 2.8 Sites publics & formulaires
+
+(mig. 24, `migrations.py` l.1021–1093)
+
+| Table | But | Clés / colonnes clés |
+|---|---|---|
+| `sites` | Mini-site public multi-pages (façon Notion Sites) | `slug UNIQUE`, FK `root_page_id → pages CASCADE`, `custom_domain UNIQUE`, `theme`, `password_hash`, `expires_at`, `noindex`, `analytics_id`, `created_by → users` |
+| `site_pages` | Arbre public ordonné des pages du site | FK `site_id → sites CASCADE`, `page_id → pages CASCADE` ; `position`, `UNIQUE(site_id, page_id)` |
+| `site_views` | Compteur de vues jour/site (upsert, **pas d'IP brute** — RGPD) | FK `site_id → sites CASCADE` ; PK composite `(site_id, day)`, `views` |
+| `form_responses` | Journal anonymisé des soumissions de formulaire public | FK `collection_id → collections CASCADE`, `row_id → collection_pages SET NULL` (ligne créée) ; `ip_hash` (hash tournant quotidien, pas d'IP) ; config du form dans `collections.form_config_json` |
+
+### 2.9 Calendrier & réunions
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `calendar_links` | Lien par utilisateur entre une collection et un calendrier externe (Google REST ou CalDAV générique) | FK `user_id → users CASCADE`, `collection_id → collections CASCADE` ; `provider`, `tokens_enc` (Fernet), `calendar_id`, `date_property`, `sync_token`, `UNIQUE(user_id, provider, calendar_id)` | mig. 27 (l.1231–1245) |
+| `meeting_transcripts` | Note de réunion (audio + transcription + résumé IA), machine à états du bloc Notion-like (idle/recording/paused/processing/done/failed) | FK `page_id → pages CASCADE` ; `audio_path`, `transcript`, `summary`, `language` ; v7.1.1 : `status`, `mode`, `channels_json`, `instruction_snapshot`, `consent_json`, `segments_json`, `processing_steps_json`, `summary_json`, `generated_title`, `completeness_class`, `quality_score`, `duration_ms/paused_ms`, **`event_occurrence_id`** (index unique partiel : une occurrence calendrier = au plus une note) | mig. 27 (l.1247–1271) + mig. 38 (l.1730–1800) |
+| `meeting_consent_log` | Journal append-only du consentement d'enregistrement | FK `transcript_id → meeting_transcripts CASCADE` ; `method` (start_attestation/…), `attested_by → users`, `participants_json` | mig. 38 (l.1767–1777) |
+| `meeting_distributions` | Journal append-only des diffusions de compte-rendu (copy_link/…) | FK `transcript_id → meeting_transcripts CASCADE` ; `channel`, `actor_id → users`, `recipients_json`, `status` | mig. 38 (l.1779–1788) |
+
+### 2.10 Import/export & offline (PWA, clipper)
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `import_jobs` | Job d'import asynchrone (polling UI) — Notion/CSV/markdown | **PK `id TEXT`** (uuid) ; `source`, `filename`, `status` (queued/running/done/error), `error`, `report_json` | mig. 9 (l.434–451) |
+| `import_items` | Dédup : 1 ligne par page importée, clé (workspace, source, external_id) → re-import idempotent | FK `workspace_id`/`page_id` **sans REFERENCES déclaré** ; `UNIQUE(workspace_id, source, external_id)` | mig. 9 (l.418–432) |
+| `offline_sync_queue` | File serveur des mutations reçues des clients hors-ligne (`/api/v2/sync/batch`), rejouables/auditées par device | FK `user_id → users CASCADE` ; `device_id`, `type` (page_create/update/delete/move, collection_*), `payload` (JSON), `client_timestamp`, `server_version`, `status` (pending/syncing/synced/failed), `retries`, `error` ; détection de conflit édit/édit via `sync_version` (mig. 17, triggers sur `pages`, `collection_pages`, `collections`) | mig. 16 (l.668–700) |
+| `extension_devices` | Devices de l'extension Web Clipper (token hashé, scopes) | FK `user_id → users CASCADE` ; `device_id`, `token_hash`, `scopes`, `revoked`, `UNIQUE(user_id, extension_name, device_id)` | mig. 19 (l.838–857) |
+| `extension_clips` | Clippings capturés (page cible + workspace cible) | FK `user_id → users CASCADE`, `target_page_id → pages SET NULL`, `target_workspace_id → workspaces SET NULL` ; `clip_type`, `source_url` | mig. 19 (l.861–877) |
+
+### 2.11 API & audit
+
+| Table | But | Clés / colonnes clés | Emplacement |
+|---|---|---|---|
+| `api_audit_log` | Traînée immuable des mutations de l'API publique v2 | FK `user_id → users SET NULL`, `token_id → api_tokens SET NULL` ; `action`, `resource_type`, `resource_id`, `ip_address`, `detail` | mig. 20 (l.895–911) |
+| `schema_version` | Registre des migrations appliquées (meta) | **PK `version INTEGER`** ; `name`, `applied_at` | `app/migrations.py` l.41–51 |
+
+### 2.12 Divers
+
+`favorites`, `recents`, `notes`, `ai_keywords`, `login_history` déjà couverts ci-dessus. À noter également : les tables temporaires de reconstruction vues dans le code (`comments_new`, `tags_new`) **ne sont pas** des tables permanentes (pattern recreate → copy → drop → rename).
+
+---
+
+## 3. Relations clés (graphe FK simplifié)
+
+```mermaid
+graph TD
+    USERS[users]
+    WS[workspaces]
+    PAGES[pages]
+    COLS[collections]
+    CP[collection_pages]
+    CPV[collection_properties]
+    CV[collection_views]
+
+    WS -->|owner_id| USERS
+    WS -->|workspace_members| USERS
+    WS -->|workspace_id| PAGES
+    WS -->|workspace_id| COLS
+    PAGES -->|parent_id self| PAGES
+    PAGES -->|collection_id| COLS
+    PAGES -->|collection_row_id CASCADE| CP
+    CP -->|collection_id CASCADE| COLS
+    CP -->|parent_id self| CP
+    CPV -->|collection_id CASCADE| COLS
+    CPV -->|related_collection_id| COLS
+    CPV -->|relation/target_property_id self| CPV
+    CPV -->|button_automation_id| AUTO[automations]
+    CV -->|collection_id CASCADE| COLS
+    AUTO -->|collection_id texte sans FK| COLS
+    AGENTS[agents] -->|workspace_id| WS
+    CONV[agent_conversations] -->|agent_id CASCADE| AGENTS
+    CONV -->|user_id| USERS
+    MSG[agent_messages + actions] -->|conversation_id CASCADE| CONV
+    GRP[user_groups] -->|workspace_id CASCADE| WS
+    PPERM[page/collection/property_permissions] -->|user_id XOR group_id| USERS
+    PPERM -->|group_id| GRP
+    SITES[sites/site_pages] -->|root/page CASCADE| PAGES
+    TRANS[meeting_transcripts] -->|page_id CASCADE| PAGES
+    CAL[calendar_links] -->|user_id| USERS
+    CAL -->|collection_id CASCADE| COLS
+    CLIPS[extension_clips] -->|target_page_id| PAGES
+```
+
+Points de relation à retenir :
+
+1. **`users` est le hub d'identité** : plus de 30 FK entrantes (sessions, tokens, permissions, audit, contenu). `workspaces` est le hub d'isolation multi-tenant (`owner_id`, `workspace_members` définit le rôle).
+2. **Hiérarchie `pages ↔ collection_pages`** : deux univers de « pages ». `pages` = éditeur de blocs wiki (FK self `parent_id`, soft-delete `deleted_at`, corbeille) ; `collection_pages` = lignes de base. Le pont v6.5 : chaque ligne peut avoir une page-ombre `pages.collection_row_id` (CASCADE) qui porte son contenu blocs — versions, synced blocks et temps réel opèrent sur elle.
+3. **Chaîne de cascades principales** : `collections → collection_pages/collection_views/collection_properties → …` et `pages → page_versions/page_shares/page_synced_blocks/page_permissions/guest_shares/text_reactions`. Les suppressions de ressource nettoyent l'éditorial ; les grants d'audit (`*_audit_log`, `api_audit_log`, `permission_audit_log`) privilégient `SET NULL` sur l'acteur pour survivre à la suppression du compte.
+4. **Permissions en 3 niveaux** (mig. 18) : `pages/collections/collection_pages.permission_type` (`inherit`|`restricted`|`private`) pilote si les tables `*_permissions` (user **XOR** group via `CHECK`) font foi ; les groupes (`user_groups`/`group_members`) sont workspace-scopés ; `page_shares` reste le chemin de partage simple, `guest_shares` le chemin sans compte.
+5. **Sync sémantique Gitea** : `boards/cards/col_mapping/checklists/property_values` et `collection_pages.gitea_issue_id` + `collections.gitea_owner/repo` relient les entités FlowDeck aux issues Gitea (jointures par ID numérique, **pas de FK** — système externe).
+6. **Polymorphisme non-FK** : `notifications.resource_type/resource_id`, `semantic_embeddings.resource_type/resource_id`, `permission_audit_log.resource_type/resource_id`, `import_items.page_id`, `agent_approvals.conversation_id`, `automations.collection_id`, `custom_emojis.workspace_id`, `teamspaces`-id sur `pages`/`collections` (ALTER sans FK) — cohérence maintenue applicativement, pas par le moteur.
+
+---
+
+## 4. Colonnes JSON structurantes
+
+Le modèle est fortement « document-oriented » : le JSON TEXT est le format de stockage des structures riches (choix assumé du SQLite sans JSON natif, requêtage via `json_extract` côté applicatif).
+
+| Domaine | Table.colonne | Contenu |
+|---|---|---|
+| Kanban | `boards.columns_json` / `wip_limits_json` | liste ordonnée des colonnes ; limites WIP par colonne |
+| Kanban/legacy | `project_properties.options_json` | options de select |
+| Collections | `collections.schema_json` | schéma complet des propriétés d'une base (déclaratif, dupliqué avec `collection_properties`) |
+| Collections | `collections.form_config_json` | formulaire public (champs, validations) |
+| Lignes | `collection_pages.property_values_json` | map `{property_name → valeur}` de la ligne |
+| Vues | `collection_views.config_json` | **config de vue** : filtres, sorts, groupements, colonnes visibles/cache, sizing — cœur du système de vues multiples |
+| Propriétés | `collection_properties.options_json` / `validation_json` | options select, règles de validation |
+| Dashboard | `collection_dashboards.layout_json` | colonnes + widgets |
+| Modèles | `database_templates.schema_json`, `page_templates.property_values_json` / `content_json` | blueprint de base / ligne |
+| Éditeur | `pages.content` (avec `content_format='blocks'`) | **tableau JSON des blocs** (heading, todo, callout, synced_block, meeting, etc.) — l'unité de contenu fondamentale |
+| Historique | `page_versions.blocks_json`, `page_history.snapshot_json`, `page_global_templates.blocks_json`, `synced_blocks.content` | snapshots de liste de blocs |
+| Workspace | `workspaces.settings_json`, `users.sidebar_config`, `users.notification_prefs`, `users.totp_backup_hashes` | préférences |
+| Agent | `agents.scope_json`/`trigger_json`, `agent_conversations.context_json`, `agent_messages.tool_calls_json`, `agent_actions.payload_json`/`result_json`/`undo_snapshot_json`, `agent_skills.allowed_tools_json`, `agent_triggers.config_json`, `agent_policies.allowed_tools_json`, `agent_approvals.args_json` | orchestration, audit et undo |
+| LLM | `user_llm_keys.models_json` | cache des modèles distants |
+| Automatisation | `automations.condition_json`/`actions_json`, `automation_steps.config_json` | DSL des règles (fallback steps ↔ legacy colonnes) |
+| Sync offline | `offline_sync_queue.payload`, `idempotency_keys.response_json` | mutation client + réponse rejouée |
+| Webhooks | `webhook_deliveries.payload` | corps livré |
+| Réunions | `meeting_transcripts.channels_json`/`consent_json`/`segments_json`/`processing_steps_json`/`summary_json`, `meeting_consent_log.participants_json`, `meeting_distributions.recipients_json` | pipeline transcription |
+| Imports | `import_jobs.report_json` | rapport par fichier |
+| SSO | `sso_config.attribute_mapping`, `groups_mapping`, `sso_login_history` (mapping) | attributs d'assertion |
+
+---
+
+## 5. Couche Pydantic (`app/models/`)
+
+Les modèles Pydantic ne couvrent **qu'une petite partie** du contrat API — la majorité des routes FastAPI valide en dur dans les handlers :
+
+- `app/models/requests.py` (75 l.) :
+  - `FileSaveRequest` — sauvegarde de fichier via Gitea (`path`, `content`, `message`, `sha` optionnel pour update) ; `@model_validator(mode="after")` restreint l'extension via `ALLOWED_EXTENSIONS`/`_ext` (`app/middleware/security`) ;
+  - `IssueCreateRequest` / `IssueUpdateRequest` — création/mutation d'issues Gitea (title ≤ 500, `state` pattern `^(open|closed)$`, labels/milestone/assignee en ids virgule) ;
+  - `CardMoveRequest` — déplacement de carte Kanban (owner/repo/issue_id/column) ;
+  - `ColMappingRequest` — mapping colonne ↔ label ;
+  - `UploadValidationResult` — DTO de validation d'upload (filename/size/extension/valid/error).
+- `app/models/responses.py` (38 l.) : deux enveloppes standardisées, utilisées en `response_model` OpenAPI des routes publiques :
+  - `ErrorResponse {error, detail?}` ;
+  - `SuccessResponse {status="ok", data?: dict}`.
+
+Aucun modèle ORM (pas de SQLAlchemy) : l'accès aux données est du **SQL brut `sqlite3`** via le context-manager `get_conn()` (`app/db.py` l.843–858, `row_factory = sqlite3.Row`), et les conversions dict↔JSON des colonnes `*_json` sont faites manuellement dans les services.
+
+---
+
+## 6. Fiches de synthèse technique
+
+| Aspect | Valeur réelle (vérifiée dans le code) |
+|---|---|
+| Moteur | SQLite unique, fichier `settings.db_path` (`database_url` = `sqlite:///…`), WAL + `busy_timeout=5000`, `foreign_keys=ON` |
+| ORM | aucun — SQL brut |
+| Pilote | `sqlite3` stdlib |
+| Migrations | maison, versionnées : baseline v1 (`init_db` idempotent) + registry `@register` (`app/migrations.py`), table `schema_version`, transaction par migration |
+| Version de schéma | 38 |
+| Triggers | FTS5 (`pages_fts_ai/ad/au`), bumps `sync_version` (`*_sync_version_bu`) sur `pages`/`collection_pages`/`collections` |
+| Full-text | FTS5 optionnel, repli `LIKE` |
+| Recherche vectorielle | `semantic_embeddings` (hash TF 256-dim en BLOB, PK composite polymorphe, pas d'extension vec0) |
+| Chiffrement au repos | Fernet sur `calendar_links.tokens_enc`, `connector_tokens.tokens_enc`, secrets SSO ; hash sha256 pour `api_tokens.token_hash`, `scim_tokens.token_hash`, `extension_devices.token_hash` ; `password_hash` (users) |
+| Sauvegardes | copie planifiée du fichier SQLite (`backup_*`, `app/config.py` l.66–70) |
+| Concurrence | verrou applicatif d'écriture indirect via WAL + busy_timeout ; migration async du 510 call sites synchrones encore en cours (note A21, `app/db.py` l.848–852) |

docs/V71_Calendar_Meetings.md

@@ -1,8 +1,8 @@
-# V7.1.0 — Calendar sync + Meeting Notes
+# V7.1.0 — Calendar sync + Meeting Notes · V7.1.1 — Bloc façon Notion
 
-> **Statut** : ✅ Livré en v7.1.0 (2026-09-28) — `services/calendar_sync.py`, `services/meetings.py`, `routers/meetings.py`, migration 27, 15 tests verts · **Roadmap** : `ROADMAP.md § v7.1.0`
+> **Statut** : ✅ Livré en v7.1.0 (2026-09-28) — `services/calendar_sync.py`, `services/meetings.py`, `routers/meetings.py`, migration 27, 15 tests verts · **Refonte Notion livrée en v7.1.1** (migration 38, `static/js/meeting_block.js`, `static/css/meeting.css`, 5 tests verts `tests/test_v711_meeting_notion.py`) · **Roadmap** : `ROADMAP.md § v7.1.0`
 > Notes d'implémentation : pas de lib CalDAV/Google (REST + XML bruts) ; refresh OAuth hors scope (relink explicite sur 401) ; conflit = LWW auto + notif (pas de picker « garder ma version ») ; free/busy au jour (pas de créneaux horaires).
-> **Référence Notion** : Notion Calendar (Google sync) + AI Meeting Notes → trigger Custom Agents (07/2026).
+> **Référence Notion** : Notion Calendar (Google sync) + AI Meeting Notes → trigger Custom Agents (07/2026). Spec détaillée : `docs/architecture-meeting-notion-flowdeck.md` (§3A parcours visuel des 6 états, §3B maquettes ASCII, §6.5 machine à états, §12.7 étapes Thinking, §12.8 complétude + titre).
 > **Existant** : vues jour/semaine/mois, moteur `recurrence.py`, `reminders.py` (scan 60s, dédup), timezones, `notify_assignment()`, template Meeting Notes.
 
 ---
@@ -41,3 +41,12 @@ Upload `.mp3/.wav/.m4a` (100 MB) sur page Meeting → `meeting_transcripts` →
 ## 3. Tests (`tests/test_v71_calendar_meetings.py`, ~15)
 
 OAuth mock (tokens chiffrés roundtrip), pull/push, conflit LWW + notif, freebusy, transcription mock → résumé blocs, trigger émis, quota 100 MB refusé.
+
+## 4. V7.1.1 — Refonte du bloc façon Notion (§3A/3B de `architecture-meeting-notion-flowdeck.md`)
+
+- **Bloc à états** (`static/js/meeting_block.js` surcharge `window.FDMeeting`, styles `static/css/meeting.css`) : `idle → recording → paused → processing → done/failed`, révélation progressive `Notes → Notes+Transcript → Summary+Notes+Transcript` (Summary en tête, actif par défaut après Stop), bouton Start à menu (source/mode, langue), onde en direct + chrono, pastille Thinking, barre `Share this summary` (Copy link / Email / Slack tracés), pied instructions + attestation de consentement puis bandeau. Le titre généré par l'IA ne touche jamais le titre de la page hôte. Anciens blocs migrés à la volée.
+- **Capture Web 2 modes** : A micro (`getUserMedia`), B onglet + micro (`getDisplayMedia` + micro, canaux séparés, alerte « je n'entends que vous » après 20 s de canal distant muet) ; STT navigateur quand dispo, transcript manuel + import audio en secours, tampon IndexedDB avec reprise après rechargement (recording → paused, processing → suivi repris).
+- **Backend** (migration 38) : colonnes état/mode/instructions photo/consent/segments/steps/résumé structuré/titre/complétude + tables append-only `meeting_consent_log`, `meeting_distributions` + unicité une occurrence = une note. Endpoints : `POST /transcripts` (création idempotente), `/consents`, `/sessions`, `/pause`, `/resume`, `/stop` (analyse inline), `/segments`, `/transcript` (groupé par source), `/processing-steps`, `/processing:retry`, `/generated-title` (PATCH), `/distributions`, `/notes` (PATCH), `GET /meetings/history` + `/meetings/instructions`.
+- **Pipeline IA** (`services/meetings.py`) : étapes Thinking persistées (lecture → analyse → résolution des références → compréhension → complétude → résumé → titre → citations), résumé contraint Overview/Décisions/Actions/Questions avec citations `¹ [seg N]` obligatoires et états vides explicites, classification `complete/brief/incomplete/fragmentary` (jamais de refus : résumé dégradé + titre-diagnostic `Brief or Incomplete Meeting Recording`), audit via `meeting.summarized` inchangé.
+- **Hub** (`base.html`) : onglets Upcoming / Today / Past, recherche plein texte, CTA contextuels (Join & transcribe / Prepare / View summary).
+- **Tests** : `tests/test_v711_meeting_notion.py` (5) — porte de consentement, machine à états complète avec résumé dégradé + titre-diagnostic, résumé structuré complet, unicité par occurrence, instructions + historique.

e2e/regression_local_workspace_breadcrumb.spec.js

@@ -0,0 +1,121 @@
+/**
+ * /local-workspace — le chemin du header suit la navigation dans les
+ * dossiers, et une navigation partielle vers un sous-dossier remonte la
+ * page Alpine (avant : « Cannot redefine property: $nextTick » → tous les
+ * composants restés visibles à la fois).
+ *
+ * Parcours : sidebar Docs → sidebar Sub → clic dans la vue sur Leaf →
+ * breadcrumb interne Docs → crumb workspace du header.
+ */
+const { test, expect } = require('@playwright/test');
+
+const FD_BASE = process.env.FD_BASE_URL || 'http://localhost:8080';
+const FD_PASS = process.env.FD_PASS || 'e2e-secret-123';
+
+test('local-workspace: chemin header + navigation dossiers (partielle et interne)', async ({ page }) => {
+  test.setTimeout(180000);
+  const errors = [];
+  page.on('console', m => { if (m.type() === 'warning' && m.text().includes('Cannot redefine')) errors.push(m.text()); });
+  page.on('pageerror', e => errors.push('PAGEERROR: ' + e.message));
+
+  // ── 1. Utilisateur + workspace + arborescence Docs/Sub/Leaf ──
+  await page.goto(`${FD_BASE}/auth/login?provider=local`);
+  await page.click('button:has-text("Register")');
+  await page.fill('#name', 'LWBC');
+  await page.fill('#email', 'lwbc' + Date.now() + '@flowdeck.local');
+  await page.fill('#password', FD_PASS);
+  await page.click('.btn-primary');
+  await page.waitForURL('**/workspaces', { timeout: 20000 });
+  await page.waitForTimeout(800);
+  await page.keyboard.press('Escape').catch(() => {});
+
+  const ids = await page.evaluate(async () => {
+    const csrf = decodeURIComponent((document.cookie.match(/csrf_token=([^;]+)/) || [])[1] || '');
+    const H = { 'Content-Type': 'application/json', 'X-CSRF-Token': csrf };
+    const r = await fetch('/api/workspaces', {
+      method: 'POST', headers: H, body: JSON.stringify({ name: 'LWBC WS ' + Date.now() })
+    });
+    const d = await r.json();
+    await fetch('/api/workspaces/' + d.id + '/select', { method: 'POST', headers: H });
+    const mk = (name, type, parent) => fetch('/api/local-workspace/items', {
+      method: 'POST', headers: H, body: JSON.stringify({ name, type, parent_id: parent || null })
+    }).then(x => x.json());
+    const docs = await mk('Docs', 'folder');
+    const sub = await mk('Sub', 'folder', docs.id);
+    const leaf = await mk('Leaf', 'folder', sub.id);
+    await mk('Page In Leaf', 'page', leaf.id);
+    return { wsName: '', docs: docs.id, sub: sub.id, leaf: leaf.id, ws: d.id };
+  });
+  await page.goto(`${FD_BASE}/local-workspace`);
+  await page.waitForFunction(() => typeof window._wsData !== 'undefined', { timeout: 15000 });
+  const wsName = await page.evaluate(() => {
+    const el = document.querySelector('.ws-main .unified-breadcrumb .bc-root');
+    return el ? el.textContent.trim() : '';
+  });
+  expect(wsName).toBeTruthy();
+
+  const headerPath = () => page.evaluate(() => {
+    const el = document.querySelector('.header-breadcrumb');
+    return el ? el.innerText.replace(/\s+/g, ' ').replace(/…/g, '...').trim() : null;
+  });
+  expect(await headerPath()).toContain(wsName);
+
+  // Composants Alpine peints à l'état brut (x-show non évalué) — hors pages
+  // légitimement visibles (arbre, ligne « parent », barre de tags).
+  // checkVisibility() prend en compte les ancêtres (display:contents sur
+  // .bc-edit-wrap ferme l'input sans que son propre display change).
+  const rawPainted = () => page.evaluate(() => {
+    const sels = ['.ws-dropdown', '.ws-select-menu', '.upload-progress', '.undo-toast',
+      '.move-modal-overlay', '.ws-search-bar', '.ws-select-bar', '.more-menu', '.filter-row',
+      '.bc-edit-wrap'];
+    return sels.filter(s => [...document.querySelectorAll(s)]
+      .some(e => e.checkVisibility({ opacityProperty: true, visibilityProperty: true })));
+  });
+
+  await page.goto(`${FD_BASE}/local-workspace`);
+  await page.waitForFunction(() => typeof window._wsData !== 'undefined', { timeout: 15000 });
+  await expect.poll(headerPath, { timeout: 10000 }).toContain(wsName);
+
+  // ── 2. Sidebar : clic sur Docs puis sur Sub (navigation partielle) ──
+  const clickSidebarFolder = async (id) => {
+    await page.evaluate((fid) => {
+      const el = document.querySelector('.sidebar-item.ws-tree-item[data-id="' + fid + '"] .tree-folder-link');
+      if (el) el.click();
+    }, id);
+    await page.waitForURL('**folder=' + id, { timeout: 15000 });
+    await page.waitForTimeout(1200);
+  };
+
+  await clickSidebarFolder(ids.docs);
+  expect(await headerPath()).toBe(`Home / ${wsName} / Docs`);
+  expect(await rawPainted()).toEqual([]);
+
+  await clickSidebarFolder(ids.sub);
+  expect(await headerPath()).toBe(`Home / ${wsName} / Docs / Sub`);
+  expect(await rawPainted()).toEqual([]);
+
+  // ── 3. Clic sur un sous-dossier DANS la vue (sans rechargement) ──
+  await page.evaluate((fid) => {
+    const el = document.querySelector('.ws-main [data-folder-id="' + fid + '"] .name.folder');
+    if (el) el.click();
+  }, ids.leaf);
+  await expect.poll(headerPath, { timeout: 10000 }).toContain('/ Sub / Leaf');
+  expect(await headerPath()).toContain(wsName);
+  expect(await rawPainted()).toEqual([]);
+
+  // ── 4. Breadcrumb interne : clic sur Docs → les deux chemins suivent ──
+  const inPage = () => page.evaluate(() =>
+    [...document.querySelectorAll('.ws-main .unified-breadcrumb a')].map(a => a.textContent.trim()));
+  expect(await inPage()).toEqual([expect.stringContaining(wsName), 'Docs', 'Sub', 'Leaf']);
+  await page.click('.ws-main .unified-breadcrumb a:has-text("Docs")');
+  await expect.poll(headerPath, { timeout: 10000 }).toBe(`Home / ${wsName} / Docs`);
+  expect(await inPage()).toEqual([expect.stringContaining(wsName), 'Docs']);
+
+  // ── 5. Crumb workspace du header → retour à la racine ──
+  await page.click('.header-breadcrumb a:has-text("' + wsName + '")');
+  await page.waitForURL('**/local-workspace', { timeout: 15000 });
+  await expect.poll(headerPath, { timeout: 10000 }).toBe(`Home / ${wsName}`);
+  expect(await rawPainted()).toEqual([]);
+
+  expect(errors).toEqual([]);
+});

scripts/vendor_hljs.py

@@ -0,0 +1,113 @@
+"""Vendor highlight.js (common bundle + extra languages + themes) into static/.
+
+Run: python3 scripts/vendor_hljs.py
+Fetches from cdnjs 11.9.0 (pinned for stability) and writes:
+  static/js/highlight.min.js        (common bundle, ~40 languages)
+  static/js/highlight-extra.min.js  (extra languages concatenated)
+  static/css/highlight.css          (github light + github-dark scoped on [data-theme])
+"""
+import re
+import urllib.request
+
+VERSION = "11.9.0"
+BASE = f"https://cdnjs.cloudflare.com/ajax/libs/highlight.js/{VERSION}"
+
+COMMON_JS = f"{BASE}/highlight.min.js"
+LANG_URL = BASE + "/languages/{lang}.min.js"
+LIGHT_CSS = BASE + "/styles/github.min.css"
+DARK_CSS = BASE + "/styles/github-dark.min.css"
+
+# Extra languages NOT in the common bundle, covering a maximum of file types.
+# NOTE (verified 11.9.0): "fish", "toml" and "cobol" have NO standalone grammar
+# on cdnjs for this version — code-highlight.js maps them to the closest
+# bundled grammar (fish→shell, toml→ini, cobol→plaintext).
+EXTRA_LANGS = [
+    # shells / ops
+    "powershell", "dos", "fish",
+    # build / containers / config
+    "dockerfile", "cmake", "toml", "graphql", "protobuf", "thrift",
+    # web servers / protocols (absent du bundle common cdnjs 11.9.0)
+    "apache", "http", "nginx",
+    # JVM / functional / misc popular
+    "scala", "groovy", "haskell", "elixir", "erlang", "clojure",
+    "ocaml", "fsharp", "nim", "d", "julia", "matlab",
+    "dart", "latex", "vim", "awk", "delphi",
+    # asm / hdl / legacy
+    "armasm", "x86asm", "vhdl", "verilog", "ada", "fortran", "cobol",
+    "gcode", "stata", "stan",
+]
+
+
+def fetch(url: str) -> str:
+    req = urllib.request.Request(url, headers={"User-Agent": "FlowDeck-vendor/1.0"})
+    with urllib.request.urlopen(req, timeout=30) as r:
+        return r.read().decode("utf-8", "replace")
+
+
+def scope_css(css: str, scope: str) -> str:
+    """Prefix every rule whose selector mentions .hljs with the theme scope."""
+    out = []
+    for chunk in css.split("}"):
+        if "{" not in chunk:
+            out.append(chunk)
+            continue
+        selectors, body = chunk.split("{", 1)
+        parts = []
+        for sel in selectors.split(","):
+            sel = sel.strip()
+            if not sel:
+                continue
+            if ".hljs" in sel:
+                parts.append(f"{scope} {sel}")
+            else:
+                parts.append(sel)
+        out.append(", ".join(parts) + "{" + body)
+    return "}".join(out)
+
+
+def main() -> None:
+    from pathlib import Path
+
+    root = Path(__file__).resolve().parent.parent
+    js_dir = root / "static" / "js"
+    css_dir = root / "static" / "css"
+
+    common = fetch(COMMON_JS)
+    (js_dir / "highlight.min.js").write_text(common, encoding="utf-8")
+    print(f"highlight.min.js: {len(common)} bytes")
+
+    extra_parts = []
+    missing = []
+    for lang in EXTRA_LANGS:
+        try:
+            code = fetch(LANG_URL.format(lang=lang))
+            extra_parts.append(f"\n/* hljs lang: {lang} */\n" + code)
+            print(f"  + {lang}: {len(code)} bytes")
+        except Exception as exc:  # noqa: BLE001
+            missing.append(lang)
+            print(f"  ! {lang} missing: {exc}")
+    (js_dir / "highlight-extra.min.js").write_text(
+        "/* FlowDeck — highlight.js extra languages (cdnjs %s) */\n" % VERSION
+        + "\n".join(extra_parts),
+        encoding="utf-8",
+    )
+
+    light = fetch(LIGHT_CSS)
+    dark = fetch(DARK_CSS)
+    css = (
+        "/* FlowDeck — highlight.js themes (cdnjs %s): github (light) + "
+        "github-dark scoped on html[data-theme=\"dark\"] */\n" % VERSION
+        + light
+        + "\n"
+        + scope_css(dark, 'html[data-theme="dark"]')
+        + "\n/* FlowDeck tweaks: keep code blocks integrated with app surfaces */\n"
+        + ".hljs{background:transparent;}\n"
+    )
+    (css_dir / "highlight.css").write_text(css, encoding="utf-8")
+    print(f"highlight.css: {len(css)} bytes")
+    if missing:
+        print("MISSING:", missing)
+
+
+if __name__ == "__main__":
+    main()

static/css/app.css

@@ -540,41 +540,42 @@ html.fd-sidebar-collapsed .editor-statusbar { left: 0; }
 .sidebar-item.empty-hint:hover { background: none; }
 .sidebar-item.empty-hint .page-name { font-size: 12px; }
 
-/* Shared subgroups (Par moi / Avec moi) */
+/* Shared subgroups (Par moi / Avec moi) — compact */
 .sidebar-subgroup {
-  padding-top: 2px;
+  padding-top: 1px;
 }
 .sidebar-subgroup + .sidebar-subgroup {
-  margin-top: 6px;
-  padding-top: 6px;
+  margin-top: 4px;
+  padding-top: 4px;
   border-top: 1px solid var(--border);
 }
 .sidebar-subgroup-label {
   display: flex;
   align-items: center;
   gap: 6px;
-  padding: 4px 8px;
-  font-size: 11px;
+  padding: 2px 8px;
+  font-size: 10px;
   font-weight: 600;
   text-transform: uppercase;
   letter-spacing: 0.04em;
   color: var(--text-tertiary);
 }
 
-/* Sections */
+/* Sections — compact (v7.70 : une seule ouverte à la fois, paddings réduits) */
 .sidebar-section {
-  padding: var(--space-xs) var(--space-md);
+  padding: 1px 6px;
 }
 
 .sidebar-section-header {
   display: flex;
   align-items: center;
   justify-content: space-between;
-  padding: 4px 6px;
+  padding: 3px 4px;
   border-radius: var(--radius-sm);
   cursor: pointer;
   transition: background var(--transition);
   user-select: none;
+  min-height: 24px;
 }
 
 .sidebar-section-header:hover {
@@ -589,8 +590,8 @@ html.fd-sidebar-collapsed .editor-statusbar { left: 0; }
 .sidebar-section-title {
   display: flex;
   align-items: center;
-  gap: 3px;
-  font-size: 12px;
+  gap: 2px;
+  font-size: 11px;
   font-weight: 600;
   color: var(--text-dim);
   text-transform: none;
@@ -754,6 +755,10 @@ html.fd-sidebar-collapsed .editor-statusbar { left: 0; }
 /* Customize Sidebar Panel */
 .sidebar-customize-panel {
   padding: 8px 12px;
+  /* v7.69.6 : au-dessus de .customize-overlay (z 998) — sinon chaque clic,
+     même sur le panneau, touche l'overlay qui referme le mode customize. */
+  position: relative;
+  z-index: 999;
 }
 
 .scp-header {
@@ -853,14 +858,14 @@ html.fd-sidebar-collapsed .editor-statusbar { left: 0; }
   display: flex;
   align-items: center;
   gap: 4px;
-  padding: 2px 4px;
+  padding: 1px 4px;
   padding-left: 4px;
   border-radius: var(--radius-sm);
-  font-size: 13px;
+  font-size: 12.5px;
   color: var(--text-primary);
   cursor: pointer;
   transition: background var(--transition);
-  min-height: 24px;
+  min-height: 22px;
 }
 
 .sidebar-item:hover {
@@ -1037,6 +1042,80 @@ html.fd-sidebar-collapsed .editor-statusbar { left: 0; }
   color: var(--text-secondary);
 }
 
+/* ── Bouton "New" + popover "Create new" (v7.70, style Notion) ──
+   Popover ouvert vers le haut, bouton pilule en bas de la sidebar. */
+.fd-new-wrap {
+  position: relative;
+  display: flex;
+  justify-content: center;
+}
+.fd-new-btn {
+  flex: 1;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  gap: 8px;
+  padding: 8px 16px;
+  border-radius: 20px;
+  background: var(--bg-hover);
+  color: var(--text-primary);
+  font-size: 14px;
+  font-weight: 500;
+  border: 1px solid transparent;
+  cursor: pointer;
+  transition: background 150ms, border-color 150ms;
+}
+.fd-new-btn:hover,
+.fd-new-btn.open {
+  background: var(--bg-active);
+  border-color: var(--border-light);
+}
+.fd-new-popover {
+  position: absolute;
+  bottom: calc(100% + 8px);
+  left: 0;
+  right: 0;
+  background: var(--bg-modal);
+  border: 1px solid var(--border-light);
+  border-radius: 10px;
+  box-shadow: 0 8px 24px rgba(0,0,0,.35);
+  padding: 6px;
+  z-index: 500;
+  animation: fd-new-pop-in 150ms ease-out;
+  transform-origin: bottom center;
+}
+@keyframes fd-new-pop-in {
+  from { opacity: 0; transform: translateY(6px); }
+  to { opacity: 1; transform: translateY(0); }
+}
+.fd-new-pop-header {
+  font-size: 12px;
+  font-weight: 600;
+  color: var(--text-dim);
+  padding: 6px 10px 4px;
+}
+.fd-new-pop-item {
+  display: flex;
+  align-items: center;
+  gap: 10px;
+  width: 100%;
+  padding: 8px 10px;
+  border-radius: 6px;
+  background: transparent;
+  border: none;
+  color: var(--text-secondary);
+  font-size: 14px;
+  text-align: left;
+  cursor: pointer;
+  transition: background 120ms, color 120ms;
+}
+.fd-new-pop-item:hover,
+.fd-new-pop-item.active {
+  background: var(--bg-hover);
+  color: var(--text-primary);
+}
+.fd-new-pop-item svg { flex-shrink: 0; }
+
 /* ═══════════════════════════════════════════════════════════
    CONTEXT MENU
    ═══════════════════════════════════════════════════════════ */
@@ -2654,6 +2733,35 @@ button.breadcrumb-current { max-width: 260px; }
   background: var(--bg-tertiary);
   border-bottom: 1px solid var(--border);
 }
+.code-lang-bar {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+.code-lang-select {
+  background: transparent;
+  border: none;
+  outline: none;
+  font-size: 12px;
+  font-weight: 600;
+  color: var(--text-dim);
+  cursor: pointer;
+  padding: 0;
+  max-width: 220px;
+}
+.code-lang-select:hover {
+  color: var(--text-primary);
+}
+.code-lang-select option {
+  background: var(--bg-modal, #1f1f1f);
+  color: var(--text-primary, #eee);
+  font-weight: 400;
+}
+/* highlight.js inside editor code blocks: theme colors the tokens,
+   the block keeps its app surface (transparent hljs background). */
+.block-code code.hljs {
+  background: transparent;
+}
 .block-code pre {
   margin: 0;
   padding: 12px 16px;

static/css/highlight.css

@@ -0,0 +1,23 @@
+/* FlowDeck — highlight.js themes (cdnjs 11.9.0): github (light) + github-dark scoped on html[data-theme="dark"] */
+pre code.hljs{display:block;overflow-x:auto;padding:1em}code.hljs{padding:3px 5px}/*!
+  Theme: GitHub
+  Description: Light theme as seen on github.com
+  Author: github.com
+  Maintainer: @Hirse
+  Updated: 2021-05-15
+
+  Outdated base version: https://github.com/primer/github-syntax-light
+  Current colors taken from GitHub's CSS
+*/.hljs{color:#24292e;background:#fff}.hljs-doctag,.hljs-keyword,.hljs-meta .hljs-keyword,.hljs-template-tag,.hljs-template-variable,.hljs-type,.hljs-variable.language_{color:#d73a49}.hljs-title,.hljs-title.class_,.hljs-title.class_.inherited__,.hljs-title.function_{color:#6f42c1}.hljs-attr,.hljs-attribute,.hljs-literal,.hljs-meta,.hljs-number,.hljs-operator,.hljs-selector-attr,.hljs-selector-class,.hljs-selector-id,.hljs-variable{color:#005cc5}.hljs-meta .hljs-string,.hljs-regexp,.hljs-string{color:#032f62}.hljs-built_in,.hljs-symbol{color:#e36209}.hljs-code,.hljs-comment,.hljs-formula{color:#6a737d}.hljs-name,.hljs-quote,.hljs-selector-pseudo,.hljs-selector-tag{color:#22863a}.hljs-subst{color:#24292e}.hljs-section{color:#005cc5;font-weight:700}.hljs-bullet{color:#735c0f}.hljs-emphasis{color:#24292e;font-style:italic}.hljs-strong{color:#24292e;font-weight:700}.hljs-addition{color:#22863a;background-color:#f0fff4}.hljs-deletion{color:#b31d28;background-color:#ffeef0}
+html[data-theme="dark"] pre code.hljs{display:block;overflow-x:auto;padding:1em}html[data-theme="dark"] code.hljs{padding:3px 5px}html[data-theme="dark"] /*!
+  Theme: GitHub Dark
+  Description: Dark theme as seen on github.com
+  Author: github.com
+  Maintainer: @Hirse
+  Updated: 2021-05-15
+
+  Outdated base version: https://github.com/primer/github-syntax-dark
+  Current colors taken from GitHub's CSS
+*/.hljs{color:#c9d1d9;background:#0d1117}html[data-theme="dark"] .hljs-doctag, html[data-theme="dark"] .hljs-keyword, html[data-theme="dark"] .hljs-meta .hljs-keyword, html[data-theme="dark"] .hljs-template-tag, html[data-theme="dark"] .hljs-template-variable, html[data-theme="dark"] .hljs-type, html[data-theme="dark"] .hljs-variable.language_{color:#ff7b72}html[data-theme="dark"] .hljs-title, html[data-theme="dark"] .hljs-title.class_, html[data-theme="dark"] .hljs-title.class_.inherited__, html[data-theme="dark"] .hljs-title.function_{color:#d2a8ff}html[data-theme="dark"] .hljs-attr, html[data-theme="dark"] .hljs-attribute, html[data-theme="dark"] .hljs-literal, html[data-theme="dark"] .hljs-meta, html[data-theme="dark"] .hljs-number, html[data-theme="dark"] .hljs-operator, html[data-theme="dark"] .hljs-selector-attr, html[data-theme="dark"] .hljs-selector-class, html[data-theme="dark"] .hljs-selector-id, html[data-theme="dark"] .hljs-variable{color:#79c0ff}html[data-theme="dark"] .hljs-meta .hljs-string, html[data-theme="dark"] .hljs-regexp, html[data-theme="dark"] .hljs-string{color:#a5d6ff}html[data-theme="dark"] .hljs-built_in, html[data-theme="dark"] .hljs-symbol{color:#ffa657}html[data-theme="dark"] .hljs-code, html[data-theme="dark"] .hljs-comment, html[data-theme="dark"] .hljs-formula{color:#8b949e}html[data-theme="dark"] .hljs-name, html[data-theme="dark"] .hljs-quote, html[data-theme="dark"] .hljs-selector-pseudo, html[data-theme="dark"] .hljs-selector-tag{color:#7ee787}html[data-theme="dark"] .hljs-subst{color:#c9d1d9}html[data-theme="dark"] .hljs-section{color:#1f6feb;font-weight:700}html[data-theme="dark"] .hljs-bullet{color:#f2cc60}html[data-theme="dark"] .hljs-emphasis{color:#c9d1d9;font-style:italic}html[data-theme="dark"] .hljs-strong{color:#c9d1d9;font-weight:700}html[data-theme="dark"] .hljs-addition{color:#aff5b4;background-color:#033a16}html[data-theme="dark"] .hljs-deletion{color:#ffdcd7;background-color:#67060c}
+/* FlowDeck tweaks: keep code blocks integrated with app surfaces */
+.hljs{background:transparent;}

static/css/meeting.css

@@ -0,0 +1,180 @@
+/* FlowDeck — AI Meeting Notes, panneau façon Notion (v7.1.1).
+ * Coquille fixe : en-tête / rangée onglets+contrôles / corps / pied (§3B.9).
+ * Tout le contenu du corps est aligné sur la marge gauche des onglets ;
+ * tout ce qui est « action globale » est aligné à droite. */
+
+.fm-tabs { display: flex; gap: 2px; align-items: center; }
+.fm-tab {
+  background: none; border: none; color: var(--text-dim); font-size: 12.5px;
+  cursor: pointer; padding: 6px 12px; border-radius: 99px; font-weight: 500;
+  display: inline-flex; align-items: center; gap: 6px; white-space: nowrap;
+}
+.fm-tab.active { background: var(--bg-tertiary); color: var(--text-primary); }
+.fm-mic {
+  font-size: 10px; border: 1px solid var(--border); border-radius: 4px;
+  padding: 0 3px; font-style: normal;
+}
+.fm-tab.active .fm-mic { border-color: var(--accent); color: var(--accent); }
+
+.fm-row { display: flex; align-items: center; justify-content: space-between; gap: 8px; flex-wrap: wrap; }
+.fm-controls { display: flex; align-items: center; gap: 6px; margin-left: auto; position: relative; }
+.fm-icon {
+  background: none; border: none; color: var(--text-dim); cursor: pointer;
+  font-size: 14px; padding: 5px 7px; border-radius: 6px; line-height: 1;
+}
+.fm-icon:hover { background: var(--bg-hover); color: var(--text-primary); }
+
+/* Bouton de démarrage à menu (split button) */
+.fm-split { display: flex; position: relative; }
+.fm-start { border-radius: 8px 0 0 8px !important; }
+.fm-chevron { border-radius: 0 8px 8px 0 !important; border-left: 1px solid rgba(255,255,255,.35) !important; padding: 7px 9px !important; }
+.fm-menu {
+  position: absolute; top: calc(100% + 6px); right: 0; z-index: 60;
+  background: var(--bg-modal); border: 1px solid var(--border);
+  border-radius: 10px; box-shadow: var(--shadow-modal);
+  padding: 8px; min-width: 280px;
+}
+.fm-menu-label { font-size: 11px; color: var(--text-dim); text-transform: uppercase; letter-spacing: .4px; padding: 6px 8px 2px; }
+.fm-menu-item {
+  display: block; width: 100%; text-align: left; background: none; border: none;
+  color: var(--text-primary); font-size: 13px; padding: 7px 8px; border-radius: 6px; cursor: pointer;
+}
+.fm-menu-item:hover { background: var(--bg-hover); }
+.fm-menu-item.sel { color: var(--accent); font-weight: 600; }
+.fm-menu-row { display: flex; gap: 6px; }
+.fm-menu-row .fm-menu-item { width: auto; }
+.fm-menu-check { display: flex; gap: 8px; align-items: center; font-size: 12.5px; color: var(--text-secondary); padding: 8px; cursor: pointer; }
+.fm-setup {
+  display: flex; flex-direction: column; gap: 3px;
+  font-size: 12.5px; color: var(--text-secondary);
+  padding: 2px 8px 6px;
+}
+/* Bande de vérification pré-enregistrement (toujours visible à l'état idle) */
+.fm-setupstrip {
+  display: flex; align-items: center; gap: 12px; flex-wrap: wrap;
+  font-size: 12px; color: var(--text-secondary);
+  border: 1px dashed var(--border); border-radius: 8px; padding: 6px 10px;
+}
+.fm-dim { color: var(--text-dim); }
+
+/* Enregistrement : onde + chrono + pastilles */
+.fm-wave { height: 30px; width: 140px; }
+.fm-timer { font-family: monospace; font-size: 12px; color: var(--text-secondary); min-width: 44px; text-align: center; }
+.fm-btn-stop { color: #e5484d; }
+.fm-paused-badge {
+  font-size: 11px; padding: 3px 10px; border-radius: 99px;
+  background: #f5a62333; color: #f5a623; white-space: nowrap;
+}
+.fm-thinking {
+  font-size: 11.5px; padding: 4px 12px; border-radius: 99px;
+  background: var(--bg-tertiary); color: var(--text-primary); white-space: nowrap;
+}
+
+/* Panneau réglages */
+.fm-settings {
+  border: 1px solid var(--border); border-radius: 8px; padding: 8px 10px;
+  display: flex; flex-direction: column; gap: 8px; background: var(--bg-tertiary);
+}
+.fm-set-row { display: flex; align-items: center; justify-content: space-between; gap: 10px; font-size: 12.5px; color: var(--text-secondary); }
+.fm-set-row select {
+  background: var(--bg-tertiary); color: var(--text-primary);
+  border: 1px solid var(--border); border-radius: 6px; padding: 4px 8px; font-size: 12.5px;
+}
+.fm-chip {
+  background: var(--bg-tertiary); border: 1px solid var(--border);
+  color: var(--text-secondary); border-radius: 99px; padding: 4px 12px;
+  font-size: 12px; cursor: pointer;
+}
+.fm-chip.active { color: var(--text-primary); border-color: var(--accent); }
+
+/* Corps */
+.fm-body { display: flex; flex-direction: column; gap: 8px; }
+.fm-promise { color: var(--text-dim); font-size: 13px; padding: 2px 0; }
+.fm-hint { color: var(--text-dim); font-size: 12px; }
+.fm-head { display: flex; align-items: center; gap: 6px; }
+.fm-cal { color: var(--text-dim); font-size: 16px; }
+.fm-chev { color: var(--text-dim); font-size: 11px; }
+.fm-rule { height: 1px; background: var(--border); margin: 0 -2px; }
+
+/* Étapes Thinking */
+.fm-steps { display: flex; flex-direction: column; gap: 6px; padding: 4px 0; }
+.fm-step { display: flex; gap: 10px; align-items: baseline; font-size: 13px; color: var(--text-dim); }
+.fm-step.done { color: var(--text-secondary); }
+.fm-step.running { color: var(--text-primary); font-weight: 600; }
+.fm-step.failed { color: #e5484d; }
+.fm-step-ico { width: 18px; text-align: center; }
+
+/* Barre de diffusion */
+.fm-sharebar {
+  border: 1px solid var(--border); border-radius: 8px; padding: 8px 12px;
+  display: flex; align-items: center; justify-content: space-between; gap: 8px; flex-wrap: wrap;
+  font-size: 13px; font-weight: 600;
+}
+.fm-share-btns { display: flex; gap: 6px; align-items: center; flex-wrap: wrap; }
+.fm-chipbtn { font-size: 12px; padding: 5px 10px; font-weight: 500; }
+
+/* Résumé structuré */
+.fm-summary { display: flex; flex-direction: column; }
+.fm-sum-sec h4 { font-size: 13.5px; font-weight: 700; margin: 10px 0 4px; color: var(--text-primary); }
+.fm-bullet { display: flex; gap: 8px; padding: 3px 0; font-size: 13px; line-height: 1.5; }
+.fm-bullet > span { flex: 1; }
+.fm-cite {
+  background: none; border: none; color: var(--accent); cursor: pointer;
+  font-size: 11px; vertical-align: super; padding: 0 1px; font-weight: 700;
+}
+.fm-cite:hover { text-decoration: underline; }
+.fm-cite-static { color: var(--accent); font-size: 11px; vertical-align: super; }
+.fm-sum-note { color: var(--text-dim); font-size: 12px; font-style: italic; margin-top: 8px; }
+.fm-class {
+  font-size: 12px; color: #f5a623; background: #f5a62314;
+  border: 1px solid #f5a62344; border-radius: 8px; padding: 6px 10px;
+}
+
+/* Transcript groupé par source + rail temporel */
+.fm-src { color: var(--text-dim); font-size: 12.5px; margin-top: 8px; }
+.fm-seg { display: flex; gap: 8px; padding: 2px 0; font-size: 13px; border-radius: 6px; }
+.fm-seg .fm-ts { color: var(--text-dim); font-family: monospace; font-size: 11px; flex-shrink: 0; width: 34px; padding-top: 3px; }
+.fm-rail { width: 2px; background: var(--border); border-radius: 2px; flex-shrink: 0; }
+.fm-seg-body { flex: 1; min-width: 0; }
+.fm-who {
+  background: none; border: none; color: var(--accent); font-weight: 600;
+  cursor: pointer; font-size: 12.5px; padding: 0 6px 0 0;
+}
+.fm-who:hover { text-decoration: underline; }
+.fm-flash { background: #2383e233; }
+
+/* Pied de bloc + bandeau */
+.fm-foot {
+  display: flex; align-items: center; gap: 10px; font-size: 12px;
+  color: var(--text-dim); flex-wrap: wrap;
+}
+.fm-foot select {
+  background: transparent; color: var(--text-primary);
+  border: 1px solid transparent; border-radius: 6px; font-size: 12px; cursor: pointer;
+}
+.fm-foot select:hover { border-color: var(--border); }
+.fm-sep { opacity: .5; }
+.fm-consent { flex: 1; min-width: 180px; }
+.fm-foot-icons { display: flex; gap: 2px; margin-left: auto; }
+.fm-banner {
+  display: flex; align-items: center; justify-content: space-between; gap: 8px;
+  font-size: 12px; color: var(--text-secondary);
+  border-top: 1px solid var(--border); padding-top: 8px;
+}
+.fm-warn {
+  font-size: 12.5px; color: #f5a623; background: #f5a62314;
+  border: 1px solid #f5a62344; border-radius: 8px; padding: 7px 10px;
+}
+.fm-manual { font-size: 12.5px; color: var(--text-secondary); }
+.fm-manual summary { cursor: pointer; color: var(--text-dim); padding: 4px 0; }
+.fm-manual-row { display: flex; gap: 8px; align-items: center; margin-top: 6px; flex-wrap: wrap; }
+
+/* Hub Meetings : recherche */
+.sb-search-row { padding: 0 8px 6px; }
+.sb-search-input {
+  width: 100%; box-sizing: border-box; background: var(--bg-tertiary);
+  border: 1px solid var(--border); border-radius: 6px; color: var(--text-primary);
+  font-size: 12px; padding: 5px 9px; outline: none;
+}
+.sb-search-input:focus { border-color: var(--accent); }
+.sb-search-input::-webkit-search-cancel-button { cursor: pointer; }

static/js/code-highlight.js

@@ -0,0 +1,348 @@
+/* ═══════════════════════════════════════════════════════════════════
+   FlowDeck — central syntax-highlighting helper (highlight.js, vendored).
+
+   Single source of truth mapping file extensions / well-known filenames to
+   highlight.js language ids. Covers ~230 file types (code, config, IaC,
+   docs). Every code viewer (Gitea file view, page-editor code blocks & file
+   viewer, workspace previews, public pages) must go through this helper so a
+   new language is added in ONE place.
+
+   Requires (load order): highlight.min.js, highlight-extra.min.js.
+   All functions are defensive: without hljs they degrade to escaped plain
+   text — never throw, never use eval (CSP-safe).
+   ═══════════════════════════════════════════════════════════════════ */
+(function () {
+  'use strict';
+
+  /* Extension (lowercase, no dot) → highlight.js language id.
+     Ids must exist in static/js/highlight.min.js (common bundle) or
+     static/js/highlight-extra.min.js — anything else falls back to the
+     alias table / plaintext via normalizeLang(). */
+  var EXT_TO_LANG = {
+    // ── Python ──
+    py: 'python', pyw: 'python', pyi: 'python', gyp: 'python', gypi: 'python',
+    // ── JavaScript / TypeScript ──
+    js: 'javascript', mjs: 'javascript', cjs: 'javascript', jsx: 'javascript',
+    ts: 'typescript', mts: 'typescript', cts: 'typescript', tsx: 'typescript',
+    // ── Web markup / styles ──
+    html: 'xml', htm: 'xml', xhtml: 'xml', xht: 'xml', xml: 'xml',
+    xsd: 'xml', xsl: 'xml', xslt: 'xml', rss: 'xml', atom: 'xml',
+    plist: 'xml', wsf: 'xml', svg: 'xml', vue: 'xml', svelte: 'xml',
+    astro: 'xml', axml: 'xml',
+    css: 'css', scss: 'scss', sass: 'scss', less: 'less',
+    // ── Data / config ──
+    json: 'json', json5: 'json', jsonc: 'json', geojson: 'json',
+    webmanifest: 'json', avsc: 'json', har: 'json',
+    yaml: 'yaml', yml: 'yaml',
+    toml: 'ini', ini: 'ini', cfg: 'ini', conf: 'ini', env: 'ini',
+    properties: 'ini', prop: 'ini', gitconfig: 'ini', gitmodules: 'ini',
+    editorconfig: 'ini', inf: 'ini', url: 'ini', reg: 'ini',
+    // ── Markdown / docs ──
+    md: 'markdown', markdown: 'markdown', mkd: 'markdown', mdwn: 'markdown',
+    mdown: 'markdown', rmd: 'markdown', qmd: 'markdown',
+    tex: 'latex', latex: 'latex', sty: 'latex', cls: 'latex', dtx: 'latex',
+    // ── SQL ──
+    sql: 'sql', pgsql: 'sql', psql: 'sql', mysql: 'sql', mssql: 'sql',
+    // ── Shells ──
+    // NOTE: no `.shell` extension key (unused); the `shell` grammar stays
+    // available via `fish`, aliases and the language picker.
+    sh: 'bash', bash: 'bash', zsh: 'bash', ksh: 'bash', mksh: 'bash',
+    fish: 'shell', yash: 'bash',
+    ps1: 'powershell', psm1: 'powershell', psd1: 'powershell',
+    bat: 'dos', cmd: 'dos',
+    // ── Systems ──
+    c: 'c', h: 'c',
+    cpp: 'cpp', cc: 'cpp', cxx: 'cpp', 'c++': 'cpp', hh: 'cpp',
+    hpp: 'cpp', hxx: 'cpp', hcc: 'cpp', ino: 'cpp', pde: 'cpp',
+    cs: 'csharp', csx: 'csharp', cake: 'csharp',
+    rs: 'rust',
+    go: 'go',
+    swift: 'swift',
+    kt: 'kotlin', kts: 'kotlin',
+    java: 'java', jsp: 'java',
+    scala: 'scala', sc: 'scala',
+    groovy: 'groovy', gvy: 'groovy', gradle: 'groovy',
+    dart: 'dart',
+    lua: 'lua',
+    // NOTE: no `.perl` / `.t` / `.mat` / `.app` / `.gpg` / `.wasm` keys on
+    // purpose (unused, ambiguous or binary extensions).
+    pl: 'perl', pm: 'perl', pod: 'perl',
+    php: 'php', phtml: 'php-template',
+    r: 'r', jl: 'julia',
+    rb: 'ruby', gemspec: 'ruby', rake: 'ruby', rbw: 'ruby',
+    // ── Functional / misc ──
+    hs: 'haskell', lhs: 'haskell',
+    ex: 'elixir', exs: 'elixir', eex: 'elixir', heex: 'elixir', leex: 'elixir',
+    erl: 'erlang', hrl: 'erlang', escript: 'erlang',
+    clj: 'clojure', cljs: 'clojure', cljc: 'clojure', edn: 'clojure',
+    ml: 'ocaml', mli: 'ocaml',
+    fs: 'fsharp', fsi: 'fsharp', fsx: 'fsharp', fsscript: 'fsharp',
+    nim: 'nim', nims: 'nim',
+    d: 'd', di: 'd',
+    m: 'objectivec', mm: 'objectivec',
+    vb: 'vbnet', vbs: 'vbnet', bas: 'vbnet',
+    pas: 'delphi', pp: 'delphi', dpr: 'delphi',
+    // ── HDL / asm / legacy ──
+    vhd: 'vhdl', vhdl: 'vhdl',
+    v: 'verilog', vh: 'verilog', sv: 'verilog', svh: 'verilog',
+    asm: 'x86asm', s: 'x86asm', nasm: 'x86asm', inc: 'x86asm',
+    ada: 'ada', ads: 'ada', adb: 'ada',
+    f: 'fortran', for: 'fortran', f90: 'fortran', f95: 'fortran',
+    f03: 'fortran', f08: 'fortran',
+    cob: 'plaintext', cbl: 'plaintext', cpy: 'plaintext',
+    // ── Build / IaC / containers ──
+    dockerfile: 'dockerfile',
+    tf: 'plaintext', tfvars: 'plaintext', hcl: 'plaintext',
+    cmake: 'cmake',
+    makefile: 'makefile', mk: 'makefile', mak: 'makefile',
+    ninja: 'ini',
+    gql: 'graphql', graphql: 'graphql', graphqls: 'graphql',
+    proto: 'protobuf',
+    thrift: 'thrift',
+    vim: 'vim', vimrc: 'vim', gvimrc: 'vim',
+    awk: 'awk',
+    gcode: 'gcode', gco: 'gcode', nc: 'gcode', ngc: 'gcode',
+    // ── Web servers / protocols ──
+    http: 'http', rest: 'http',
+    nginx: 'nginx',
+    apache: 'apache', htaccess: 'apache',
+    wat: 'wasm',
+    // ── Diffs / misc text ──
+    diff: 'diff', patch: 'diff', rej: 'diff',
+    log: 'plaintext', txt: 'plaintext', text: 'plaintext',
+    csv: 'plaintext', tsv: 'plaintext', tab: 'plaintext',
+    lock: 'plaintext', gitignore: 'plaintext', gitattributes: 'plaintext',
+    gitkeep: 'plaintext', dockerignore: 'plaintext', npmignore: 'plaintext',
+    bib: 'plaintext', bbl: 'plaintext', nfo: 'plaintext', asc: 'plaintext',
+  };
+
+  /* Exact basenames (lowercase) → language, checked BEFORE the extension
+     (Dockerfile / Makefile / go.mod … often carry no usable extension). */
+  var FILE_TO_LANG = {
+    'dockerfile': 'dockerfile', 'containerfile': 'dockerfile',
+    'docker-compose.yml': 'yaml', 'docker-compose.yaml': 'yaml',
+    'compose.yml': 'yaml', 'compose.yaml': 'yaml',
+    'makefile': 'makefile', 'gnumakefile': 'makefile',
+    'cmakelists.txt': 'cmake',
+    'go.mod': 'go', 'go.sum': 'plaintext', 'go.work': 'go',
+    'cargo.toml': 'ini', 'cargo.lock': 'plaintext',
+    'package.json': 'json', 'package-lock.json': 'json',
+    'composer.json': 'json', 'tsconfig.json': 'json',
+    'nginx.conf': 'nginx',
+    '.bashrc': 'bash', '.bash_profile': 'bash', '.bash_login': 'bash',
+    '.bash_logout': 'bash', '.zshrc': 'bash', '.zshenv': 'bash',
+    '.profile': 'bash', '.shrc': 'bash', '.kshrc': 'bash',
+    '.vimrc': 'vim', '.gvimrc': 'vim',
+    'rakefile': 'ruby', 'gemfile': 'ruby', 'vagrantfile': 'ruby',
+    'podfile': 'ruby', 'fastfile': 'ruby', 'dangerfile': 'ruby',
+    'brewfile': 'ruby', 'berksfile': 'ruby',
+    'build.gradle': 'groovy', 'settings.gradle': 'groovy',
+    'build.gradle.kts': 'kotlin', 'settings.gradle.kts': 'kotlin',
+    '.gitconfig': 'ini', '.gitmodules': 'ini', '.editorconfig': 'ini',
+    '.htaccess': 'apache', '.viminfo': 'vim',
+  };
+
+  /* Aliases users may type (markdown fences, old block labels, Prism ids…)
+     → canonical highlight.js id. */
+  var ALIAS_TO_LANG = {
+    'py': 'python', 'python3': 'python', 'gyp': 'python',
+    'js': 'javascript', 'jsx': 'javascript', 'mjs': 'javascript',
+    'cjs': 'javascript', 'node': 'javascript',
+    'ts': 'typescript', 'tsx': 'typescript',
+    'html': 'xml', 'htm': 'xml', 'xhtml': 'xml', 'svg': 'xml',
+    'markup': 'xml', 'vue': 'xml', 'svelte': 'xml',
+    'yml': 'yaml', 'yaml': 'yaml',
+    'md': 'markdown', 'mkd': 'markdown',
+    'sh': 'bash', 'shell': 'shell', 'zsh': 'bash', 'console': 'shell',
+    'ps1': 'powershell', 'ps': 'powershell', 'powershell': 'powershell',
+    'bat': 'dos', 'cmd': 'dos', 'batch': 'dos',
+    'shell-session': 'shell', 'shellsession': 'shell',
+    'c++': 'cpp', 'cc': 'cpp', 'hpp': 'cpp', 'h': 'c',
+    'c#': 'csharp', 'csharp': 'csharp', 'cs': 'csharp',
+    'kt': 'kotlin', 'kts': 'kotlin',
+    'rb': 'ruby',
+    'rs': 'rust',
+    'golang': 'go',
+    'pl': 'perl', 'pm': 'perl',
+    'm': 'objectivec', 'mm': 'objectivec', 'objc': 'objectivec',
+    'r': 'r',
+    'jl': 'julia',
+    'toml': 'ini', 'ini': 'ini', 'cfg': 'ini', 'conf': 'ini',
+    'env': 'ini', 'properties': 'ini',
+    'tex': 'latex', 'latex': 'latex',
+    'docker': 'dockerfile',
+    'make': 'makefile',
+    'patch': 'diff',
+    'text': 'plaintext', 'plain': 'plaintext', 'plain text': 'plaintext',
+    'plaintext': 'plaintext', 'txt': 'plaintext', 'log': 'plaintext',
+    'none': 'plaintext', '': 'plaintext',
+  };
+
+  /* Language id → short human label for the code-block badge / selector. */
+  var LANG_LABELS = {
+    python: 'Python', javascript: 'JavaScript', typescript: 'TypeScript',
+    xml: 'HTML/XML', css: 'CSS', scss: 'SCSS', less: 'Less',
+    json: 'JSON', yaml: 'YAML', ini: 'INI/Config', markdown: 'Markdown',
+    latex: 'LaTeX', sql: 'SQL', bash: 'Bash', shell: 'Shell',
+    powershell: 'PowerShell', dos: 'Batch', c: 'C', cpp: 'C++',
+    csharp: 'C#', rust: 'Rust', go: 'Go', swift: 'Swift', kotlin: 'Kotlin',
+    java: 'Java', scala: 'Scala', groovy: 'Groovy', dart: 'Dart',
+    lua: 'Lua', perl: 'Perl', php: 'PHP', 'php-template': 'PHP',
+    r: 'R', julia: 'Julia', ruby: 'Ruby', haskell: 'Haskell',
+    elixir: 'Elixir', erlang: 'Erlang', clojure: 'Clojure',
+    ocaml: 'OCaml', fsharp: 'F#', nim: 'Nim', d: 'D',
+    objectivec: 'Objective-C', matlab: 'MATLAB', vbnet: 'VB.NET',
+    delphi: 'Delphi', vhdl: 'VHDL', verilog: 'Verilog',
+    x86asm: 'ASM', armasm: 'ARM ASM', ada: 'Ada', fortran: 'Fortran',
+    dockerfile: 'Docker', cmake: 'CMake', makefile: 'Makefile',
+    graphql: 'GraphQL', protobuf: 'Protobuf', thrift: 'Thrift',
+    vim: 'Vim', awk: 'AWK', gcode: 'G-code',
+    http: 'HTTP', nginx: 'Nginx', apache: 'Apache', wasm: 'WASM',
+    diff: 'Diff', plaintext: 'Plain Text',
+  };
+
+  /** Languages offered in the code-block language picker (id order). */
+  var PICKER_LANGS = [
+    'plaintext', 'python', 'javascript', 'typescript', 'xml', 'css', 'scss',
+    'json', 'yaml', 'ini', 'markdown', 'sql', 'bash', 'shell', 'powershell',
+    'dos', 'c', 'cpp', 'csharp', 'rust', 'go', 'swift', 'kotlin', 'java',
+    'scala', 'groovy', 'dart', 'lua', 'perl', 'php', 'r', 'julia', 'ruby',
+    'haskell', 'elixir', 'erlang', 'clojure', 'ocaml', 'fsharp', 'nim', 'd',
+    'objectivec', 'dockerfile', 'cmake', 'makefile', 'graphql', 'protobuf',
+    'latex', 'vim', 'http', 'nginx', 'diff',
+  ];
+
+  function hasHljs() {
+    return typeof window.hljs !== 'undefined' && !!window.hljs.getLanguage;
+  }
+
+  function escHtml(s) {
+    return String(s == null ? '' : s)
+      .replace(/&/g, '&')
+      .replace(/</g, '&lt;')
+      .replace(/>/g, '&gt;');
+  }
+
+  /** Normalize any user label / extension / Prism id to a bundled hljs id. */
+  function normalizeLang(lang) {
+    var l = String(lang == null ? '' : lang).trim().toLowerCase();
+    if (!l || l === 'plain text' || l === 'none') return 'plaintext';
+    if (ALIAS_TO_LANG[l]) return ALIAS_TO_LANG[l];
+    if (!hasHljs()) return l || 'plaintext';
+    try {
+      if (window.hljs.getLanguage(l)) return l;
+    } catch (e) { /* ignore → plaintext */ }
+    return 'plaintext';
+  }
+
+  /** Best highlight.js id for a file path (basename map, then extension). */
+  function langForFile(path) {
+    var name = String(path || '').replace(/\\/g, '/').split('/').pop().toLowerCase();
+    if (!name) return 'plaintext';
+    if (FILE_TO_LANG[name]) return FILE_TO_LANG[name];
+    // dotfiles like ".bashrc" without another dot are covered above; a
+    // trailing ".xyz" is the extension otherwise.
+    var dot = name.lastIndexOf('.');
+    if (dot < 0) return 'plaintext';
+    var ext = name.slice(dot + 1);
+    if (FILE_TO_LANG['.' + name]) return FILE_TO_LANG['.' + name];
+    if (EXT_TO_LANG[ext]) return EXT_TO_LANG[ext];
+    return 'plaintext';
+  }
+
+  function labelFor(lang) {
+    var n = normalizeLang(lang);
+    return LANG_LABELS[n] || n;
+  }
+
+  /** Highlight `text` → HTML string (escaped when hljs/lang missing). */
+  function highlightText(text, lang) {
+    var src = String(text == null ? '' : text);
+    var n = normalizeLang(lang);
+    if (n === 'plaintext' || !hasHljs()) return escHtml(src);
+    try {
+      return window.hljs.highlight(src, { language: n }).value;
+    } catch (e) {
+      return escHtml(src);
+    }
+  }
+
+  /**
+   * Highlight a <code> element in place. The element's TEXT is preserved
+   * (highlight.js reads textContent), so this is safe for contenteditable
+   * code blocks: sync() re-reads text, spans are transparent.
+   * Re-highlighting is skipped when neither language nor text changed
+   * (cheap length+prefix fingerprint) — e.g. repeated editor renders.
+   */
+  function highlightBlock(el, lang) {
+    if (!el) return;
+    var n = normalizeLang(lang || el.getAttribute('data-hljs-lang') || langClassOf(el));
+    el.setAttribute('data-hljs-lang', n);
+    if (n !== 'plaintext') {
+      el.classList.add('hljs');
+      el.classList.add('language-' + n);
+    }
+    if (!hasHljs() || n === 'plaintext') return;
+    try {
+      var fp = fingerprint(el);
+      if (el.dataset && el.dataset.hljsFp === fp) return; // unchanged
+      window.hljs.highlightElement(el);
+      if (el.dataset) el.dataset.hljsFp = fp;
+    } catch (e) { /* keep plain text */ }
+  }
+
+  function langClassOf(el) {
+    if (!el || !el.className) return '';
+    var m = /(?:^|\s)language-([\w-]+)/.exec(el.className);
+    return m ? m[1] : '';
+  }
+
+  function fingerprint(el) {
+    var t = el.textContent || '';
+    return (el.getAttribute('data-hljs-lang') || '') + '|' + t.length + ':' + t.slice(0, 64);
+  }
+
+  /**
+   * Highlight every `pre code` viewer block under `root`:
+   * `[data-hljs]` (explicit, e.g. editor code blocks) and
+   * `code[class*=language-]` (Gitea viewer, exports, public pages).
+   * Editable code blocks are included — their text round-trips losslessly.
+   */
+  function highlightAllIn(root) {
+    var scope = root || document;
+    if (!hasHljs() || !scope.querySelectorAll) return 0;
+    var count = 0;
+    try {
+      var nodes = scope.querySelectorAll(
+        'pre code[data-hljs], pre code[data-hljs-lang], pre code[class*="language-"]'
+      );
+      for (var i = 0; i < nodes.length; i++) {
+        var el = nodes[i];
+        var lang = el.getAttribute('data-hljs-lang') || langClassOf(el) || 'plaintext';
+        var n = normalizeLang(lang);
+        if (n === 'plaintext') continue;
+        el.setAttribute('data-hljs-lang', n);
+        el.classList.add('hljs');
+        var fp = n + '|' + (el.textContent || '').length + ':' + (el.textContent || '').slice(0, 64);
+        if (el.dataset && el.dataset.hljsFp === fp) continue;
+        window.hljs.highlightElement(el);
+        if (el.dataset) el.dataset.hljsFp = fp;
+        count++;
+      }
+    } catch (e) { /* never break rendering */ }
+    return count;
+  }
+
+  window.FDHighlight = {
+    EXT_TO_LANG: EXT_TO_LANG,
+    FILE_TO_LANG: FILE_TO_LANG,
+    PICKER_LANGS: PICKER_LANGS,
+    normalizeLang: normalizeLang,
+    langForFile: langForFile,
+    labelFor: labelFor,
+    highlightText: highlightText,
+    highlightBlock: highlightBlock,
+    highlightAllIn: highlightAllIn,
+    hasHljs: hasHljs,
+  };
+})();

static/js/gitea_workspace.js

@@ -19,7 +19,8 @@ var _regGiteaWs = () => {
     fileSize: 0,
     fileSha: '',
     fileLoading: false,
-    fileLanguage: 'text',
+    fileLanguage: 'plaintext',
+    fileLanguageLabel: 'Plain Text',
     editContent: '',
     commitMessage: 'Update via FlowDeck',
     commitHistory: JSON.parse(localStorage.getItem('fd_commit_msgs') || '[]'),
@@ -290,17 +291,27 @@ var _regGiteaWs = () => {
         this.fileContent = '[Error loading file]';
       }
       this.fileLoading = false;
-      // Highlight after Alpine renders
+      // Highlight after Alpine renders (FDHighlight covers ~230 file types;
+      // legacy Prism kept as a last-resort fallback).
       var self = this;
       this.$nextTick(function() {
         var block = self.$refs.codeBlock;
-        if (block && block.textContent && typeof Prism !== 'undefined') {
+        if (!block || !block.textContent) return;
+        if (window.FDHighlight) {
+          try {
+            FDHighlight.highlightBlock(block, self.fileLanguage);
+            self.fileLanguageLabel = FDHighlight.labelFor(self.fileLanguage);
+          } catch (e) { /* plain text */ }
+        } else if (typeof Prism !== 'undefined') {
           Prism.highlightElement(block);
         }
       });
     },
 
     getLanguage(path) {
+      if (window.FDHighlight) {
+        try { return FDHighlight.langForFile(path); } catch (e) { /* fall through */ }
+      }
       var ext = (path || '').split('.').pop().toLowerCase();
       var map = {
         js:'javascript', jsx:'javascript', ts:'typescript', tsx:'typescript',

static/js/library.js

@@ -35,6 +35,9 @@ function libraryPage() {
     shareDir: 'all',
     showAuthor: false,
     showSource: false,
+    showEditedBy: true,
+    showEditedTime: true,
+    showVisited: true,
 
     // Tags
     workspaceTags: [],
@@ -90,6 +93,12 @@ function libraryPage() {
       for (var i = 0; i < this.flatItems.length; i++) {
         var item = this.flatItems[i];
 
+        // ── En-tête de section de date (onglet AI Meeting Notes) ──
+        if (item._isSection) {
+          html += '<div class="lib-section-header">' + self._escHtml(item.title) + '</div>';
+          continue;
+        }
+
         // ── Synthetic "+ Add new" placeholder row for an expanded parent ──
         if (item._isAddNew) {
           var pad = 30 + (item.depth || 0) * 20;
@@ -154,10 +163,19 @@ function libraryPage() {
         html += '<div class="row-cell lib-col-author"><div class="row-author">';
         html += '<div class="author-avatar">' + (item.author_initial || '?') + '</div>';
         html += '<span>' + (item.author || '\u2014') + '</span></div></div>';
-        // Source
-        html += '<div class="row-cell lib-col-source"><span class="row-source">' + (item.source_label || 'Private') + '</span></div>';
+        // Last edited by
+        html += '<div class="row-cell lib-col-editedby"><div class="row-author">';
+        html += '<div class="author-avatar">' + (item.edited_by_initial || '?') + '</div>';
+        html += '<span>' + (item.edited_by || '\u2014') + '</span></div></div>';
         // Edited
         html += '<div class="row-cell lib-col-edited"><span class="row-time">' + self.formatTime(item.updated_at) + '</span></div>';
+        // Last visited
+        html += '<div class="row-cell lib-col-visited"><span class="row-time">' +
+          (item.visited_at ? self.formatTime(item.visited_at) : '\u2014') + '</span></div>';
+        // Source (icône de source + libellé) — en dernier, comme l'en-tête
+        html += '<div class="row-cell lib-col-source"><span class="row-source">' +
+          getSvgIcon(item.source_icon || 'file', 13) +
+          '<span>' + (item.source_label || 'Private') + '</span></span></div>';
         html += '</div>';
       }
       body.innerHTML = html;
@@ -202,6 +220,7 @@ bindHtmlItem(el, item) {
 },
     _renderIcon(item) {
       if (!item) return getSvgIcon('file', 14);
+      if (item.is_meeting) return getSvgIcon('calendar', 14);
       // Custom emoji (image URL), icon name, or unicode emoji
       if (item.page_icon) return '<span style="font-size:14px;line-height:1;display:inline-flex;align-items:center;">' + (window.fdIconHtml ? window.fdIconHtml(item.page_icon, 14) : this._escHtml(item.page_icon)) + '</span>';
       if (item.is_folder || item.has_children) return getSvgIcon('folder', 14);
@@ -226,13 +245,47 @@ _renderIcon(item) {
           return (r.tags || []).some(function(t) { return t.name === tf; });
         });
       }
+      if (this.tab === 'meetings') result = this._groupMeetings(result);
       // Use splice for Alpine 3.14.9 reactivity (array reference reassign fails)
       this.flatItems.splice(0, this.flatItems.length, ...result);
       this.selectedCount = Object.keys(this.selected).length;
-      this.allSelected = this.flatItems.length > 0 && this.selectedCount === this.flatItems.length;
+      var selectable = result.filter(function(r) { return !r._isSection; });
+      this.allSelected = selectable.length > 0 && this.selectedCount === selectable.length;
       this._renderTable();
     },
 
+    // Regroupement par date pour l'onglet AI Meeting Notes : sections
+    // Today / Yesterday / This week / Last week / This month / Older.
+    _meetingBucket(updatedAt) {
+      var d = new Date(String(updatedAt || '').replace(' ', 'T'));
+      if (isNaN(d.getTime())) return 5;
+      var now = new Date();
+      var start = new Date(now.getFullYear(), now.getMonth(), now.getDate());
+      var diff = Math.floor((start.getTime() - new Date(d.getFullYear(), d.getMonth(), d.getDate()).getTime()) / 86400000);
+      if (diff <= 0) return 0;
+      if (diff === 1) return 1;
+      if (diff <= 7) return 2;
+      if (diff <= 14) return 3;
+      if (diff <= 30) return 4;
+      return 5;
+    },
+
+    _groupMeetings(items) {
+      var labels = ['Today', 'Yesterday', 'This week', 'Last week', 'This month', 'Older'];
+      var buckets = [[], [], [], [], [], []];
+      for (var i = 0; i < items.length; i++) {
+        if (items[i]._isSection) continue;
+        buckets[this._meetingBucket(items[i].updated_at)].push(items[i]);
+      }
+      var out = [];
+      for (var b = 0; b < buckets.length; b++) {
+        if (!buckets[b].length) continue;
+        out.push({ _isSection: true, id: 'sec-' + b, title: labels[b] });
+        for (var j = 0; j < buckets[b].length; j++) out.push(buckets[b][j]);
+      }
+      return out;
+    },
+
     // ── Global dispatch (bypasses Alpine x-for reactivity issues) ──
     toggleItemById(id) {
       var item = this._findItem(id);
@@ -399,7 +452,7 @@ _renderIcon(item) {
     },
 
     _canAddChild() {
-      return this.tab !== 'favorites' && this.tab !== 'shared' && this.tab !== 'published' && this.tab !== 'repository';
+      return this.tab !== 'favorites' && this.tab !== 'shared' && this.tab !== 'published' && this.tab !== 'repository' && this.tab !== 'meetings';
     },
 
     _getCsrf() {
@@ -443,13 +496,14 @@ _renderIcon(item) {
     emptyText: '',
 
     _recomputeEmpty() {
-      var icons = {recents:'clock',favorites:'star',shared:'users',published:'globe',private:'lock',workspace:'folder',repository:'folder'};
+      var icons = {recents:'clock',meetings:'calendar',favorites:'star',shared:'users',published:'globe',private:'lock',workspace:'folder',repository:'folder'};
       this.emptyIcon = icons[this.tab] || 'file';
-      var titles = {recents:'No recent pages',favorites:'No favorites',shared:'No shared pages',
+      var titles = {recents:'No recent pages',meetings:'No meeting notes',favorites:'No favorites',shared:'No shared pages',
                     published:'No published pages',private:'No private pages',workspace:'No workspace pages',
                     repository:'Empty repository'};
       this.emptyTitle = titles[this.tab] || 'No items';
       var texts = {recents:'Pages you\'ve recently viewed or edited will appear here.',
+                   meetings:'Create a meeting note from the sidebar (+) — it will appear here, grouped by date.',
                    favorites:'Right-click a page in the sidebar and select "Add to Favorites".',
                    shared:'Shared pages from your team will appear here.',
                    published:'Pages published to the web will appear here.',
@@ -478,7 +532,7 @@ _renderIcon(item) {
       console.log('[Library] workspaceId=' + this.workspaceId + ' isGitea=' + this.isGiteaActive);
       var params = new URLSearchParams(window.location.search);
       var t = params.get('tab');
-      if (t && ['recents','favorites','shared','published','private','workspace','repository'].indexOf(t) !== -1) {
+      if (t && ['recents','meetings','favorites','shared','published','private','workspace','repository'].indexOf(t) !== -1) {
         this.tab = t;
       }
       this._recomputeEmpty();
@@ -524,7 +578,7 @@ _renderIcon(item) {
         if (this.tab === 'repository' && this.isGiteaActive) {
           url += sep + 'gitea_owner=' + encodeURIComponent(this.giteaOwner) + '&gitea_repo=' + encodeURIComponent(this.giteaRepo);
           sep = '&';
-        } else if (this.tab !== 'repository' && this.tab !== 'recents' && this.tab !== 'favorites') {
+        } else if (this.tab !== 'repository' && this.tab !== 'recents' && this.tab !== 'favorites' && this.tab !== 'meetings') {
           // Hierarchical tree view: show only root-level pages at the top
           url += sep + 'tree=1';
           sep = '&';
@@ -573,6 +627,9 @@ _renderIcon(item) {
     toggleColumn(col) {
       if (col === 'author') this.showAuthor = !this.showAuthor;
       else if (col === 'source') this.showSource = !this.showSource;
+      else if (col === 'editedby') this.showEditedBy = !this.showEditedBy;
+      else if (col === 'editedtime') this.showEditedTime = !this.showEditedTime;
+      else if (col === 'visited') this.showVisited = !this.showVisited;
     },
 
     doSort() {
@@ -650,6 +707,7 @@ _renderIcon(item) {
       } else {
         var s = {};
         for (var i = 0; i < this.flatItems.length; i++) {
+          if (this.flatItems[i]._isSection || this.flatItems[i]._isAddNew) continue;
           s[this.flatItems[i].id] = true;
         }
         this.selected = s;

static/js/local_workspace.js

@@ -1,6 +1,12 @@
+// Config de la page courante — relue à CHAQUE exécution du script : sur une
+// navigation partielle htmx le <script> est ré-exécuté (avec le nouveau
+// #lw-config) alors que le corps du fichier est verrouillé par le garde
+// ci-dessous. Avec un `const` capturé à la première exécution, la page
+// gardait le folder/workspace du premier chargement.
+var LW = (function(){try{var el=document.getElementById('lw-config');return el?JSON.parse(el.textContent):{}}catch{return {}}})();
+
 if (!window.__fdLocalWorkspaceScriptsLoaded) {
 window.__fdLocalWorkspaceScriptsLoaded = true;
-const LW=(()=>{try{const el=document.getElementById('lw-config');return el?JSON.parse(el.textContent):{}}catch{return {}}})();
 
 console.log('FLOWDECK v2.4.6 — IIFE running, about to set _wsInitData');
 const _wsInitData = (function() {  // lexical : NON propriété globalThis →
@@ -150,6 +156,19 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
       try {
         // Expose globally for event handlers (BEFORE await)
         window._wsData = this;
+        // Navigation partielle : l'objet est partagé entre les montages →
+        // on recale l'état sur la page courante (#lw-config fraîchement lu)
+        // et on remet à zéro les panneaux/menus de la vue précédente.
+        if (LW && LW.current_folder_id != null) this.currentFolder = LW.current_folder_id;
+        this.searchQuery = '';
+        this.filterType = '';
+        this.editingPath = false;
+        this.previewItem = null;
+        this.uploading = false;
+        this.undoVisible = false;
+        this.moveModalOpen = false;
+        this.isDragging = false;
+        this.clearSelection();
         var url = '/api/local-workspace/tree';
         if (this.currentFolder && this.currentFolder > 0) {
           url += '?folder=' + this.currentFolder;
@@ -197,6 +216,32 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
       } catch(e) { console.error('Tree load error:', e); this.loaded = true; }
       this._recompute();
       this._initResizeHandle();
+      this._syncHeaderPath();
+    },
+
+    /** Recalcule le chemin du header (Home / <workspace> / <dossier> / …) à
+        partir de la pile de dossiers courante — même mécanique que le rendu
+        serveur, mais sans rechargement. */
+    _syncHeaderPath() {
+      var stack = this.folderStack || [];
+      var crumbs = [
+        { label: 'Home', url: '/workspaces', id: null, icon: 'home', menu: false, home: true },
+        { label: (LW && LW.workspace_name) || 'Workspace', url: '/local-workspace',
+          id: null, icon: 'folder', menu: false }
+      ];
+      for (var i = 0; i < stack.length; i++) {
+        var last = (i === stack.length - 1);
+        crumbs.push({
+          label: stack[i].name,
+          url: last ? null : '/local-workspace?folder=' + stack[i].id,
+          id: stack[i].id,
+          icon: 'folder',
+          menu: true
+        });
+      }
+      window.dispatchEvent(new CustomEvent('flowdeck:breadcrumb-changed', {
+        detail: { workspace_id: (LW && LW.workspace_id) || 0, crumbs: crumbs }
+      }));
     },
 
     _initResizeHandle() {
@@ -486,6 +531,12 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
                 if (b.type === 'code' || b.type === 'code_block') {
                   tag = 'pre';
                   style += 'background:var(--bg-tertiary);padding:10px 14px;border-radius:6px;font-family:monospace;font-size:12px;white-space:pre-wrap;';
+                  var cLang = (window.FDHighlight && b.language) ? FDHighlight.normalizeLang(b.language) : '';
+                  var cAttr = (cLang && cLang !== 'plaintext') ? ' data-hljs data-hljs-lang="' + cLang + '" class="hljs language-' + cLang + '"' : '';
+                  var cHtml = (window.FDHighlight && cLang && cLang !== 'plaintext')
+                    ? (function(){ try { return FDHighlight.highlightText(txt, cLang); } catch (e) { return ''; } })() || this._escHtml(txt)
+                    : this._escHtml(txt);
+                  return '<' + tag + ' style="' + style + '"><code' + cAttr + '>' + (txt ? cHtml : '<br>') + '</code></' + tag + '>';
                 }
                 return '<' + tag + ' style="' + style + '">' + (txt ? this._escHtml(txt) : '<br>') + '</' + tag + '>';
               }, this).join('');
@@ -493,7 +544,7 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
               preview = '<div style="text-align:center;padding:30px;color:var(--text-tertiary)">Empty document</div>';
             }
           } catch {
-            preview = '<pre style="background:var(--bg-tertiary);padding:14px;border-radius:6px;overflow:auto;font-size:12px;font-family:monospace;white-space:pre-wrap;">' + this._escHtml(String(content).substring(0, 3000)) + '</pre>';
+            preview = this._codeHtml(String(content).substring(0, 3000), node.name, 3000);
           }
 
         } else if (format === 'markdown') {
@@ -501,15 +552,16 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
           preview = '<div style="white-space:pre-wrap;line-height:1.6;font-size:13px;">' + this._escHtml(String(content).substring(0, 5000)) + '</div>';
 
         } else {
-          // Unknown format — show as text
+          // Unknown format — show as (highlighted) text
           var display = String(content || '').trim();
           if (display) {
-            preview = '<pre style="background:var(--bg-tertiary);padding:14px;border-radius:6px;overflow:auto;font-size:12px;font-family:monospace;white-space:pre-wrap;">' + this._escHtml(display.substring(0, 5000)) + '</pre>';
+            preview = this._codeHtml(display, node.name, 5000);
           } else {
             preview = '<div style="text-align:center;padding:30px;color:var(--text-tertiary)">No content</div>';
           }
         }
         body.innerHTML = preview || '<div style="color:var(--text-tertiary);text-align:center;padding:20px">(empty)</div>';
+        try { if (window.FDHighlight) FDHighlight.highlightAllIn(body); } catch (e) {}
       } catch {
         body.innerHTML = '<div style="text-align:center;padding:20px;color:var(--text-tertiary)">Preview not available</div>';
       }
@@ -519,6 +571,23 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
       return String(s||'').replace(/&/g,'&amp;').replace(/</g,'&lt;').replace(/>/g,'&gt;');
     },
 
+    // ── hljs: code-aware <pre> for previews (FDHighlight, ~230 file types).
+    // Falls back to escaped plain text when the helper is unavailable.
+    _codeHtml(text, name, maxLen) {
+      var t = String(text == null ? '' : text).substring(0, maxLen || 5000);
+      if (window.FDHighlight) {
+        try {
+          var lang = FDHighlight.langForFile(name || '');
+          var inner = FDHighlight.highlightText(t, lang);
+          var cls = lang !== 'plaintext' ? ' class="hljs language-' + lang + '"' : '';
+          return '<pre style="background:var(--bg-tertiary);padding:14px;border-radius:6px;overflow:auto;font-size:12px;font-family:monospace;white-space:pre-wrap;">'
+            + '<code' + cls + '>' + inner + '</code></pre>';
+        } catch (e) { /* fallback below */ }
+      }
+      return '<pre style="background:var(--bg-tertiary);padding:14px;border-radius:6px;overflow:auto;font-size:12px;font-family:monospace;white-space:pre-wrap;">'
+        + this._escHtml(t) + '</pre>';
+    },
+
     // ── Content snippet for content view ──
     contentSnippetMap: {},
 
@@ -765,7 +834,10 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
         html += '<span class="icon">'+icon+'</span>';
         if (c.is_folder) {
           var countBadge = (c.child_count > 0) ? ' <span class="child-count">('+c.child_count+')</span>' : '';
-          html += '<span class="name folder" onclick="window._wsData.navigateToFolder('+escId+')">'+c.name+countBadge+'</span>';
+          var statusBadges = (c.is_shared ? '<span class="ws-shared-badge" title="Partag\u00e9" onclick="event.stopPropagation()">👥</span>' : '')
+            + (c.favorited ? '<span class="ws-fav-badge" title="Favorite">★</span>' : '')
+            + (c.published ? '<span class="ws-pub-badge" title="Published to web">🌐</span>' : '');
+          html += '<span class="name folder" onclick="window._wsData.navigateToFolder('+escId+')">'+statusBadges+c.name+countBadge+'</span>';
         } else {
           // Click to rename, double-click to open
           html += '<span class="name file" onclick="event.stopPropagation();window._wsData.startRenderChildRename('+escId+',\''+escName+'\')"'
@@ -774,6 +846,8 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
                + ' onmouseenter="var d=window._wsData;if(d)d.showPreview(event,{id:'+escId+',name:\''+escName+'\',is_folder:false,content_format:\''+(c.content_format||'')+'\'})"'
                + ' onmouseleave="var d=window._wsData;if(d)d.hidePreview()">'
                 + (c.is_shared ? '<span class="ws-shared-badge" title="Partag\u00e9" onclick="event.stopPropagation()">👥</span>' : '')
+                + (c.favorited ? '<span class="ws-fav-badge" title="Favorite">★</span>' : '')
+                + (c.published ? '<span class="ws-pub-badge" title="Published to web">🌐</span>' : '')
                 + '<span id="rc-name-'+escId+'">'+c.name+'</span>'
                + '<input id="rc-input-'+escId+'" class="inline-rename-input" style="display:none"'
                + ' onkeydown="if(event.key===\'Enter\')window._wsData.commitRenderChildRename('+escId+')"'
@@ -956,6 +1030,7 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
         this._saveExpandedState();
         this.previewItem = null;
         this._recompute();
+        this._syncHeaderPath();
         var self = this;
         this.$nextTick(function() { self._syncCheckboxes(); });
       } catch(e) { console.error('Load folder error:', e); }
@@ -1230,8 +1305,9 @@ const _wsInitData = (function() {  // lexical : NON propriété globalThis →
         } else {
           this.previewContent = 'Loading...';
           this._fetchRetry(url).then(function(t) {
-            var content = t.substring(0, 500);
-            self.previewContent = content; self.previewCache[cacheKey] = content;
+            var html = self._codeHtml(t, node.name, 500);
+            // Tooltip body is already monospace/pre-wrap: keep it lightweight.
+            self.previewContent = html; self.previewCache[cacheKey] = html;
           }).catch(function() { self.previewContent = '(preview not available)'; });
         }
       } else {
@@ -2043,9 +2119,32 @@ console.log('FLOWDECK v2.4.6 — IIFE done, _wsInitData keys:', Object.keys(_wsI
 // A20 ph3 : registre Alpine.data (le build CSP ne résout que le registre ;
 // x-data="wsInitData()" au lieu de l'identifiant global _wsInitData).
 var _regWs = function () {
-  // objet partagé JS↔Alpine via closure LEXICALE (const hors window →
-  // non snapshoté par ji ; probe4 : factory retournant l'objet = accepté).
-  Alpine.data('wsInitData', function () { return _wsInitData; });
+  // Objet partagé JS↔Alpine via closure LEXICALE (const hors window →
+  // non snapshoté par ji). L'état reste unique, MAIS la factory Alpine
+  // doit retourner une enveloppe FRAÎCHE à chaque montage : Alpine pose ses
+  // magics ($nextTick…) avec Object.defineProperty non-configurable. En
+  // retournant le même objet, le 2e montage (navigation partielle
+  // /local-workspace → /local-workspace?folder=) levait
+  // « Cannot redefine property: $nextTick », initTree échouait et TOUT le
+  // contenu restait au state brut (tous les x-show visibles à la fois).
+  // L'enveloppe délègue lecture/écriture au proxy réactif partagé.
+  Alpine.data('wsInitData', function () {
+    var R = Alpine.reactive(_wsInitData);
+    var t = {};
+    Object.keys(_wsInitData).forEach(function (k) {
+      if (typeof _wsInitData[k] === 'function') {
+        t[k] = _wsInitData[k];
+      } else {
+        Object.defineProperty(t, k, {
+          get: function () { return R[k]; },
+          set: function (v) { R[k] = v; },
+          enumerable: true,
+          configurable: true,
+        });
+      }
+    });
+    return t;
+  });
   // Bloc preview : le parseur le place HORS de la div racine (structure
   // pré-existante) → montage distinct DÉLÉGUANT vers l'objet partagé.
   // Wrapper = objet unique (les magics $nextTick… ne sont redéfinissables

static/js/page_editor_scripts.js

@@ -13,6 +13,7 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
     ]},{name:'AI',items:[
       {id:'ai_write',name:'Write with AI',icon:'✨'},{id:'ai_summarize',name:'Summarize',icon:'📝'},
       {id:'ai_translate',name:'Translate',icon:'🌐'},{id:'ai_continue',name:'Continue writing',icon:'➡️'},
+      {id:'meeting',name:'AI Meeting Notes',icon:'🎙️'},
     ]},
     {name:'MEDIA',items:[
       {id:'image',name:'Image',icon:'🖼️'},{id:'embed',name:'Embed',icon:'🔗'},
@@ -398,7 +399,7 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
 
     // ── Block type shortcuts (shown on the right of each slash item) ──
     // Raccourcis = préfixes markdown (alignés sur la description mobile Notion : -, [], >, ---)
-    const _SHORTCUT={text:'T',heading_1:'#',heading_2:'##',heading_3:'###',heading_4:'####',bulleted_list:'-',numbered_list:'1.',to_do:'[]',toggle:'>',quote:'"',callout:'!',table_of_contents:'TOC',math:'∑',columns:'‖',divider:'---',code:'<>',table:'|',image:'[img]',embed:'[🔗]',bookmark:'[🔖]',video:'[🎬]',audio:'[🎵]',database:'DB',ai_write:'/ai',ai_summarize:'/ai',ai_translate:'/ai',ai_continue:'/ai',duplicate:'Ctrl+D',turn_into:'⋯',copy_link:'#',delete:'⌫'};
+    const _SHORTCUT={text:'T',heading_1:'#',heading_2:'##',heading_3:'###',heading_4:'####',bulleted_list:'-',numbered_list:'1.',to_do:'[]',toggle:'>',quote:'"',callout:'!',table_of_contents:'TOC',math:'∑',columns:'‖',divider:'---',code:'<>',table:'|',image:'[img]',embed:'[🔗]',bookmark:'[🔖]',video:'[🎬]',audio:'[🎵]',database:'DB',ai_write:'/ai',ai_summarize:'/ai',ai_translate:'/ai',ai_continue:'/ai',meeting:'/meet',duplicate:'Ctrl+D',turn_into:'⋯',copy_link:'#',delete:'⌫'};
     const _GROUP_LABEL={BASIC:'Basic blocks',AI:'AI writing',MEDIA:'Media',DATA:'Data',ACTIONS:'Actions',SYNCED:'Synced'};
     // ── Hover preview card (matches Notion's visual hint) ──
     function _previewHtml(blockId){
@@ -426,6 +427,7 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
         case 'video': return p('<div style="border:1px solid var(--border);border-radius:6px;padding:14px;text-align:center;color:var(--text-dim);font-size:12px">🎬 Inline video player</div>');
         case 'audio': return p('<div style="border:1px solid var(--border);border-radius:6px;padding:14px;text-align:center;color:var(--text-dim);font-size:12px">🎵 Inline audio player</div>');
         case 'database': return p('<div style="border:1px solid var(--border);border-radius:6px;padding:10px;font-size:12px">📊 Inline database table</div>');
+        case 'meeting': return p('<div style="border:1px solid var(--border);border-radius:6px;padding:10px;font-size:12px">🎙️ <b>AI Meeting Notes</b><br><span style="color:var(--text-dim)">Record, transcribe and summarize this meeting.</span></div>');
         default: return p('<div style="font-size:13px;color:var(--text-dim)">Text block</div>');
       }
     }
@@ -751,6 +753,28 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
     };
     window.AIAC=AIAC;
 
+    // ── Code block renderer (hljs: highlight.js syntax highlighting) — literal
+    // text (esc, real newlines — NOT mdEsc, whose <br> would break
+    // highlight.js textContent reads) plus a language picker driving
+    // FDHighlight (~230 file types). Highlight spans are transparent to
+    // sync(): gt()/gtTok() re-read text only.
+    function renderCodeBlock(b){
+      var rawLang=b.language||'Plain Text';
+      var norm=(window.FDHighlight?FDHighlight.normalizeLang(rawLang):String(rawLang||'plaintext').toLowerCase().replace(/[^a-z0-9+#-]+/g,''))||'plaintext';
+      var ro=!!(window.E&&E.canEdit===false);
+      var head;
+      if(!ro&&window.FDHighlight){
+        var opts=FDHighlight.PICKER_LANGS.map(function(id){
+          return '<option value="'+id+'"'+(id===norm?' selected':'')+'>'+esc(FDHighlight.labelFor(id))+'</option>';
+        }).join('');
+        head='<div class="code-lang-label code-lang-bar"><select class="code-lang-select" onchange="E.setCodeLang(\''+b.id+'\',this.value)" title="Code language — syntax highlighting">'+opts+'</select></div>';
+      }else{
+        var label=(window.FDHighlight?FDHighlight.labelFor(norm):rawLang);
+        head='<div class="code-lang-label">'+esc(label)+'</div>';
+      }
+      return head+'<pre><code data-bid="'+b.id+'" data-hljs data-hljs-lang="'+norm+'" class="language-'+norm+'" contenteditable="'+(ro?'false':'true')+'" spellcheck="false">'+esc(b.content||'')+'</code></pre>';
+    }
+
     function renderBlock(b,idx){
       if(b.type==='meeting')return renderMeetingBlock(b,idx);
       if(b.type==='table')return renderTableBlock(b,idx);
@@ -813,7 +837,7 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
       let inner='';
       if(b.type==='to_do')inner=`<input type="checkbox" class="todo-checkbox" ${b.checked?'checked':''} onchange="E._doToggle('${b.id}',this.checked)"><div data-bid="${b.id}" contenteditable="true" data-placeholder="${ph}" spellcheck="false">${mdEsc(b.content)}</div>`;
       else if(b.type==='toggle')inner=`<div class="toggle-title-row"><button class="toggle-chevron ${b.expanded?'open':''}" onclick="E._doToggleExpand('${b.id}')">▶</button><div class="toggle-title" data-bid="${b.id}" contenteditable="true" data-placeholder="${ph}" spellcheck="false">${mdEsc(b.content)}</div></div><div class="toggle-children" ${(b.expanded===false)?'style="display:none"':''}>${(b.children||[]).map((ch)=>renderNestedBlock(ch)).join('')}</div>`;
-      else if(b.type==='code')inner=`<div class="code-lang-label">${b.language||'Plain Text'}</div><pre><code data-bid="${b.id}" contenteditable="true" spellcheck="false">${mdEsc(b.content)}</code></pre>`;
+      else if(b.type==='code')inner=renderCodeBlock(b);
       else if(b.type==='callout')inner=`<span class="callout-icon" onclick="E._toggleCalloutIcon('${b.id}')" title="Change icon">${b.icon||'💡'}</span><div data-bid="${b.id}" contenteditable="true" data-placeholder="${ph}" spellcheck="false">${mdEsc(b.content)}</div><div class="callout-icon-picker" id="callout-picker-${b.id}" style="display:none">${CALLOUT_ICONS.map(function(i){return '<button type="button" class="callout-icon-opt" onclick="E._setCalloutIcon(\''+b.id+'\',\''+i+'\')">'+i+'</button>';}).join('')}</div>`;
       else inner=`<div data-bid="${b.id}" contenteditable="true" data-placeholder="${ph}" spellcheck="false">${mdEsc(b.content)}</div>`;
       const tag=b.type.startsWith('heading_')?'h'+b.type.slice(-1):b.type==='quote'?'blockquote':'div';
@@ -1656,12 +1680,13 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
           if(mime==='application/pdf'||ext==='pdf'){
             this.blocks=[this.mkB('embed','',{src:url,embed_type:'pdf'})];this.render();return;
           }
-          const codeExts=['py','js','ts','jsx','tsx','html','htm','css','json','xml','yaml','yml','md','markdown','sql','sh','bash','ps1','bat','cmd','rb','go','rs','java','c','cpp','h','hpp','php','swift','kt','scala','r','toml','ini','cfg','conf','env','txt','log'];
+          const _fh=(typeof window!=='undefined'&&window.FDHighlight)?window.FDHighlight:null;
+          const codeExts=_fh?Object.keys(_fh.EXT_TO_LANG):['py','js','ts','jsx','tsx','html','htm','css','json','xml','yaml','yml','md','markdown','sql','sh','bash','ps1','bat','cmd','rb','go','rs','java','c','cpp','h','hpp','php','swift','kt','scala','r','toml','ini','cfg','conf','env','txt','log'];
           const langMap={'py':'python','js':'javascript','ts':'typescript','jsx':'javascript','tsx':'typescript','html':'html','htm':'html','css':'css','json':'json','xml':'xml','yaml':'yaml','yml':'yaml','md':'markdown','markdown':'markdown','sql':'sql','sh':'bash','bash':'bash','ps1':'powershell','bat':'plaintext','cmd':'plaintext','rb':'ruby','go':'go','rs':'rust','java':'java','c':'c','cpp':'cpp','h':'c','hpp':'cpp','php':'php','swift':'swift','kt':'kotlin','scala':'scala','r':'r','toml':'toml','ini':'ini','cfg':'ini','conf':'ini','env':'ini','txt':'plaintext','log':'plaintext'};
           if(mime.startsWith('text/')||codeExts.includes(ext)){
             try{const r=await fetch(url);const text=await r.text();
               if(ext==='md'||ext==='markdown'){this.blocks=this.md2b(text);}
-              else{this.blocks=[this.mkB('code',text,{language:langMap[ext]||'Plain Text'})];}
+              else{const lang=_fh?_fh.langForFile(name):(langMap[ext]||'Plain Text');this.blocks=[this.mkB('code',text,{language:lang})];}
               this.render();
             }catch(e){this.blocks=[this.mkB('paragraph','Error loading file: '+e.message)];this.render();}
             return;
@@ -1717,6 +1742,9 @@ const PD=(()=>{try{const el=document.getElementById('page-data');return el?JSON.
           if(b.type==='columns'&&b.children)b.children.forEach(function(ch){const che=ct.querySelector(`[data-bid="${ch.id}"]`);if(che)ch.content=gtMd(che)||'';});
         }},
         render(){const ct=document.getElementById('_blocksCt');if(!ct)return;let html='';for(let i=0;i<this.blocks.length;i++)html+=renderBlock(this.blocks[i],i);ct.innerHTML=html;
+         // hljs: syntax-highlight code blocks (FDHighlight skips unchanged
+         // blocks via content fingerprint; spans are sync-transparent).
+         try{if(window.FDHighlight)FDHighlight.highlightAllIn(ct);}catch(e){}
          // Render math blocks with KaTeX
          var self=this;
          ct.querySelectorAll('.math-display').forEach(function(el){var bid=el.dataset.bid;var b=self.blocks.find(function(x){return x.id===bid;});self._renderKatex(el,b?b.content:'');});
@@ -1983,6 +2011,8 @@ replaceBlock(idx,type){const b=this.blocks[idx];if(!b)return;this.sync();this.pu
 
         // ── v5.10.0: Turn into / couleurs / liens / déplacement ──
         _blockTargets(idx){const sel=_selIndicesSorted();return(sel.includes(idx))?sel:[idx];},
+        // hljs: code-block language picker (drives FDHighlight re-render).
+        setCodeLang(bid,lang){const idx=this.getIdx(bid);if(idx<0)return;const b=this.blocks[idx];if(!b||b.type!=='code')return;this.sync();this.pushHistory();b.language=(window.FDHighlight?FDHighlight.normalizeLang(lang):lang)||'Plain Text';this.dirty=true;this.autoSave();this.render();_rtSync();},
         turnInto(idx,type,cols){if(idx<0||idx>=this.blocks.length)return;if(!_BLOCK_TURN_TYPES.some(t=>t.id===type))return;const targets=this._blockTargets(idx);const focusId=this.blocks[idx].id;this.sync();this.pushHistory();targets.forEach(i=>{const b=this.blocks[i];if(!b)return;if(b.type===type)return;const keep=(b.content||'').trim();b.type=type;if(keep&&['image','embed','meeting','divider','table_of_contents','database','button','columns'].indexOf(b.type)<0){b.content=keep;}else{b.content='';}if(type==='toggle'){b.expanded=true;if(!Array.isArray(b.children))b.children=[];}if(type==='to_do')b.checked=false;if(type==='table'&&!Array.isArray(b.rows))b.rows=_tblNew(3,3);if(type==='code'&&!b.language)b.language='Plain Text';if(type==='columns'&&(!Array.isArray(b.children)||!b.children.length))b.children=[{id:genId(),type:'paragraph',content:''}];if(type==='columns'&&cols>=2){b.children=Array.from({length:cols},()=>({id:genId(),type:'paragraph',content:''}));}if(type==='synced'){b.synced_id=0;b._synced_content=[];b._synced_title='';}if(type==='button'){b.automation_id=0;b.automation_name='';}});
           if(type==='synced')setTimeout(()=>this._pickSyncedBlock(idx),80);
           this.dirty=true;this.autoSave();this.render();_rtSync();_selClear();
@@ -2702,6 +2732,23 @@ applyAIBlocks(text){
            }
            if(id.indexOf('ai_')===0){ this.aiSlash(id,idx); return; }
            if(id==='duplicate'){this.duplicateBlock(idx);return;}
+           if(id==='meeting'){
+             // v7.1.1 : bloc AI Meeting Notes (façon Notion) — même action que
+             // meetingNote(), mais en place dans le bloc courant (/meet).
+             const block=this.blocks[idx];this.sync();this.pushHistory();
+             const today=new Date();const iso=today.toISOString().slice(0,10);
+             block.type='meeting';block.content='';
+             block.meeting={v:2,title:'Meeting @'+iso,date:iso,state:'idle',
+               mode:'mic_only',instruction:'auto',lang:'fr',consent:null,
+               consentAudio:false,tab:'notes',channels:[],segments:[],segSeq:0,
+               notes:'',manual:'',interim:'',startedAt:0,elapsedBase:0,pausedMs:0,
+               steps:[],summaryMd:'',completeness:'',genTitle:'',titleEdited:false,
+               shareHidden:false,bannerHidden:false,transcriptId:null,error:'',
+               remoteSilentWarned:false,menuOpen:false,settingsOpen:false};
+             this.render();this.dirty=true;this.autoSave();_rtSync();
+             setTimeout(()=>{const el=document.querySelector('[data-meeting-bid="'+block.id+'"]');if(el&&el.scrollIntoView)el.scrollIntoView({behavior:'smooth',block:'center'});},60);
+             return;
+           }
           if(id==='delete'){this.removeBlock(idx);return;}
           if(id==='copy_link'){const blk=this.blocks[idx];if(blk)this.copyBlockLink(blk.id);return;}
           if(id==='turn_into'){const blk=this.blocks[idx];if(!blk)return;const el=this.getEl(blk.id);const rk=el?el.getBoundingClientRect():null;blockMenuOpen(rk?rk.left:window.innerWidth/2,rk?rk.bottom:window.innerHeight/2,blk.id);return;}

static/sw.js

@@ -20,7 +20,7 @@ const PRECACHE_URLS = [
   '/static/css/design-tokens.css?v=5.2.0',
   '/static/css/components.css?v=5.2.0',
   '/static/css/katex.min.css?v=0.16.11',
-  '/static/css/prism.css',
+  '/static/css/highlight.css',
   // JS
   '/static/js/htmx.min.js',
   '/static/js/alpine.csp.min.js',
@@ -28,7 +28,9 @@ const PRECACHE_URLS = [
   '/static/js/app.js?v=2.4.8',
   '/static/js/offline.js?v=6.0.0',
   '/static/js/katex.min.js?v=0.16.11',
-  '/static/js/prism.min.js',
+  '/static/js/highlight.min.js',
+  '/static/js/highlight-extra.min.js',
+  '/static/js/code-highlight.js',
   // Icônes & métadonnées
   '/static/favicon.ico',
   '/static/manifest.json',

tests/test_v696_sidebar_library_publish.py

@@ -0,0 +1,240 @@
+"""FlowDeck v7.69.6 — Library AI Meeting Notes, Recents sidebar, Publish unifié.
+
+- ``GET /api/library/meetings`` : notes de réunion groupées par date côté
+  client ; exclues des onglets recents/workspace/private.
+- Sidebar Recents : fichiers réellement récents (``_recent_sidebar_pages``),
+  icônes workspace, notes de réunion exclues.
+- Publish : ``is_published`` + ``published`` maintenus ensemble — la page
+  publiée apparaît dans la sidebar Published, la Library et le dialogue.
+"""
+from __future__ import annotations
+
+import json
+import secrets
+
+from conftest import anon_csrf  # noqa: F401
+
+from app.db import get_conn
+
+
+def _login(client):
+    from app.auth.session import SessionManager
+    login = f"v696_{secrets.token_hex(4)}"
+    with get_conn() as conn:
+        conn.execute(
+            "INSERT INTO users (login, full_name, email, is_admin) VALUES (?, 'V696', ?, 0)",
+            (login, f"{login}@test.com"),
+        )
+        uid = conn.execute("SELECT id FROM users WHERE login=?", (login,)).fetchone()["id"]
+        conn.commit()
+    return SessionManager.create_session({"id": uid, "login": login}), uid
+
+
+def _cookies(session):
+    return {"flowdeck_session": session}
+
+
+def _mkworkspace(uid, name="WS696"):
+    with get_conn() as conn:
+        cur = conn.execute("INSERT INTO workspaces (name, owner_id) VALUES (?, ?)",
+                           (name, uid))
+        conn.commit()
+        return cur.lastrowid
+
+
+def _mkpage(ws_id, title, with_meeting=False, updated="2026-10-09 12:00:00"):
+    content = json.dumps([{"id": "b1",
+                           "type": "meeting" if with_meeting else "paragraph",
+                           "content": ""}])
+    with get_conn() as conn:
+        cur = conn.execute(
+            "INSERT INTO pages (workspace, workspace_id, title, content, content_format,"
+            " parent_section, updated_at) VALUES ('t', ?, ?, ?, 'blocks', 'Private', ?)",
+            (ws_id, title, content, updated))
+        conn.commit()
+        return cur.lastrowid
+
+
+# ── Library meetings ─────────────────────────────────────────────────────
+
+def test_library_meetings_tab_lists_only_meeting_notes(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    mp = _mkpage(ws, "Sprint Review", True)
+    _mkpage(ws, "Plain notes", False)
+    r = client.get("/api/library/meetings", cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    items = r.json()["items"]
+    assert [i["id"] for i in items] == [mp]
+    assert items[0]["is_meeting"] is True
+    assert items[0]["url"] == f"/pages/{mp}"
+
+
+def test_library_browsing_tabs_hide_meeting_notes(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    _mkpage(ws, "Sprint Review", True, updated="2026-10-09 12:00:00")
+    plain = _mkpage(ws, "Plain notes", False, updated="2026-10-09 13:00:00")
+    for tab in ("recents", "workspace", "private"):
+        r = client.get(f"/api/library/{tab}", cookies=_cookies(session))
+        assert r.status_code == 200, r.text
+        ids = [i["id"] for i in r.json()["items"]]
+        assert plain in ids
+        assert not any(i.get("is_meeting") for i in r.json()["items"])
+        assert len([i for i in ids if isinstance(i, int)]) == len(ids)
+
+
+# ── Sidebar recents ──────────────────────────────────────────────────────
+
+def test_recent_sidebar_pages_excludes_meetings_and_uses_workspace_icons(client):
+    from app.routers.dashboard._common import _recent_sidebar_pages
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    _mkpage(ws, "Sprint Review", True)
+    plain = _mkpage(ws, "Plain notes", False)
+    recents = _recent_sidebar_pages(uid)
+    ids = [e["id"] for e in recents]
+    assert f"page/{plain}" in ids
+    assert not any("Sprint" in e["name"] for e in recents)
+    icons = {e["id"]: e["icon"] for e in recents}
+    assert icons[f"page/{plain}"] == "edit"  # même convention que le Workspace
+    assert all(e["url"].startswith("/pages/") for e in recents)
+
+
+# ── Publish unifié ───────────────────────────────────────────────────────
+
+def test_publish_sets_both_flags_and_lists(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    pid = _mkpage(ws, "Launch notes", False)
+    r = client.post(f"/api/pages/{pid}/publish", cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert r.json()["is_published"] is True
+    with get_conn() as conn:
+        row = conn.execute("SELECT is_published, published, publish_slug FROM pages"
+                           " WHERE id=?", (pid,)).fetchone()
+    assert row["is_published"] == 1 and row["published"] == 1
+    assert row["publish_slug"]
+    # Library Published + sidebar Published voient la page.
+    r = client.get("/api/library/published", cookies=_cookies(session))
+    assert pid in [i["id"] for i in r.json()["items"]]
+    from app.routers.board._common import _load_shared_sidebar_pages
+    _made, _received, published, _shared = _load_shared_sidebar_pages(uid)
+    assert any(p["db_id"] == pid for p in published)
+    # Le dialogue (page_data) reflète l'état.
+    r = client.get(f"/pages/{pid}", cookies=_cookies(session))
+    assert r.status_code == 200
+    assert '"published": true' in r.text
+    # Unpublish nettoie les deux drapeaux.
+    r = client.delete(f"/api/pages/{pid}/publish", cookies=_cookies(session))
+    assert r.json()["is_published"] is False
+    with get_conn() as conn:
+        row = conn.execute("SELECT is_published, published FROM pages WHERE id=?",
+                           (pid,)).fetchone()
+    assert row["is_published"] == 0 and row["published"] == 0
+    r = client.get("/api/library/published", cookies=_cookies(session))
+    assert pid not in [i["id"] for i in r.json()["items"]]
+
+
+def test_share_dialog_publish_path_sets_both_flags_and_slug(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    pid = _mkpage(ws, "Dialog publish", False)
+    r = client.post(f"/board/api/share/{pid}",
+                    json={"mode": "private", "published": True},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    with get_conn() as conn:
+        row = conn.execute("SELECT is_published, published, publish_slug FROM pages"
+                           " WHERE id=?", (pid,)).fetchone()
+    assert row["is_published"] == 1 and row["published"] == 1
+    assert row["publish_slug"]
+    r = client.get("/api/library/published", cookies=_cookies(session))
+    assert pid in [i["id"] for i in r.json()["items"]]
+
+
+def test_customize_panel_above_overlay():
+    from pathlib import Path
+    css = (Path(__file__).resolve().parent.parent
+           / "static" / "css" / "app.css").read_text(encoding="utf-8")
+    panel = css.split(".sidebar-customize-panel")[1].split("}")[0]
+    assert "z-index: 999" in panel
+
+
+def test_library_column_selectors_beat_row_rules():
+    """Show columns : les règles de visibilité doivent battre
+    ``.lib-table-header .th`` et ``.lib-row .row-cell`` (0,2,0) — avant,
+    seul ``:not(.show-editedtime)`` gagnait et seule Last edited time
+    s'affichait/masquait (vérifié en direct au navigateur)."""
+    from pathlib import Path
+    html = (Path(__file__).resolve().parent.parent
+            / "app" / "templates" / "library.html").read_text(encoding="utf-8")
+    for cls in ("lib-col-author", "lib-col-editedby",
+                "lib-col-source", "lib-col-visited"):
+        assert f".lib-table div.{cls}" in html, cls
+
+
+# ── v7.69.6b : provenance, badges, favoris ─────────────────────────────────
+
+def _save_blocks(client, session, pid, title="T"):
+    r = client.post(f"/board/api/pages/{pid}/blocks",
+                    json={"title": title,
+                          "blocks": [{"id": "b1", "type": "paragraph",
+                                      "content": "hello"}]},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+
+
+def test_library_provenance_columns(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    pid = _mkpage(ws, "Prov notes", False)
+    _save_blocks(client, session, pid, "Prov notes")
+    r = client.get("/api/library/recents", cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    items = [i for i in r.json()["items"] if i["id"] == pid]
+    assert len(items) == 1
+    it = items[0]
+    assert it["author"], "Created by must resolve (versions, sinon owner)"
+    assert it["edited_by"], "Last edited by must resolve"
+    assert it["source_icon"] in ("file", "link", "external-link")
+    assert it["visited_at"] == "", "not visited yet"
+    # Lecture → visite tracée.
+    assert client.get(f"/pages/{pid}", cookies=_cookies(session)).status_code == 200
+    r = client.get("/api/library/recents", cookies=_cookies(session))
+    it = [i for i in r.json()["items"] if i["id"] == pid][0]
+    assert it["visited_at"], "Last visited time must be tracked on view"
+
+
+def test_favorites_use_workspace_icons(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    pid = _mkpage(ws, "Fav notes", False)
+    with get_conn() as conn:
+        conn.execute("INSERT INTO favorites (user_id, page_id, position) VALUES (?,?,0)",
+                     (uid, pid))
+        conn.commit()
+    r = client.get("/api/library/favorites", cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    items = [i for i in r.json()["items"] if i["id"] == pid]
+    assert len(items) == 1
+    assert items[0]["icon"] in ("folder", "file", "edit", "image")
+
+
+def test_local_workspace_tree_status_flags(client):
+    from app.routers.dashboard._common import _build_tree_children
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    pid = _mkpage(ws, "Flagged notes", False)
+    with get_conn() as conn:
+        conn.execute("INSERT INTO favorites (user_id, page_id, position) VALUES (?,?,0)",
+                     (uid, pid))
+        conn.execute("UPDATE pages SET is_published=1, published=1, publish_slug='x',"
+                     " share_mode='anyone' WHERE id=?", (pid,))
+        conn.commit()
+        tree = _build_tree_children(conn, None, ws, uid)
+    nodes = [n for n in tree if n["id"] == pid]
+    assert len(nodes) == 1
+    assert nodes[0]["favorited"] is True
+    assert nodes[0]["published"] is True
+    assert nodes[0]["is_shared"] is True

tests/test_v711_meeting_notion.py

@@ -0,0 +1,328 @@
+"""FlowDeck v7.1.1 — Meeting block façon Notion (machine à états, consentement,
+segments, résumé structuré, partage)."""
+from __future__ import annotations
+
+import secrets
+
+import pytest
+from conftest import anon_csrf  # noqa: F401
+
+from app.db import get_conn
+
+
+def _login(client):
+    from app.auth.session import SessionManager
+    login = f"v711_{secrets.token_hex(4)}"
+    with get_conn() as conn:
+        conn.execute(
+            "INSERT INTO users (login, full_name, email, is_admin) VALUES (?, 'V711', ?, 0)",
+            (login, f"{login}@test.com"),
+        )
+        uid = conn.execute("SELECT id FROM users WHERE login=?", (login,)).fetchone()["id"]
+        conn.commit()
+    return SessionManager.create_session({"id": uid, "login": login}), uid
+
+
+def _cookies(session):
+    return {"flowdeck_session": session}
+
+
+def _mkpage(title="Meeting"):
+    with get_conn() as conn:
+        cur = conn.execute(
+            "INSERT INTO pages (workspace, title, content, content_format)"
+            " VALUES ('test', ?, '', 'blocks')", (title,))
+        conn.commit()
+        return cur.lastrowid
+
+
+def _mktr(client, session, pid=None, **kw):
+    pid = pid or _mkpage()
+    r = client.post("/api/v2/meetings/transcribe",
+                    files={"audio": ("rec.mp3", b"ID3" + b"\x00" * 50, "audio/mpeg")},
+                    data={"page_id": str(pid), **{k: str(v) for k, v in kw.items()}},
+                    cookies=_cookies(session))
+    assert r.status_code == 201, r.text
+    return r.json()["id"]
+
+
+def test_consent_gate_blocks_start(client):
+    session, _ = _login(client)
+    tid = _mktr(client, session)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/sessions",
+                    json={"mode": "mic_only"}, cookies=_cookies(session))
+    assert r.status_code == 400
+    assert "consent" in r.text.lower()
+
+
+def test_full_state_machine_offline_degraded(client):
+    session, _ = _login(client)
+    pid = _mkpage("Meeting @Today")
+    tid = _mktr(client, session, pid)
+    # consentement par attestation au démarrage (§3A.2)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/consents",
+                    json={"method": "start_attestation"}, cookies=_cookies(session))
+    assert r.status_code == 200
+    # start avec photo d'instructions
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/sessions",
+                    json={"mode": "tab_plus_mic", "language": "fr",
+                          "instruction_id": "standup"},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert r.json()["status"] == "recording"
+    # segments groupés par source, marque de troncature (§3A.7)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/segments",
+                    json={"segments": [
+                        {"channel": "self", "source_label": "Bruno mic",
+                         "speaker": "Bruno", "text": "What I like first is the plan",
+                         "start_ms": 0, "end_ms": 2000, "starts_mid": True}]},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    # pause / resume = états serveur (§6.5)
+    assert client.post(f"/api/v2/meetings/transcripts/{tid}/pause",
+                       json={}, cookies=_cookies(session)).status_code == 200
+    assert client.post(f"/api/v2/meetings/transcripts/{tid}/resume",
+                       cookies=_cookies(session)).status_code == 200
+    # stop → processing → résumé dégradé classé, titre-diagnostic (§12.8)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/stop",
+                    json={"duration_ms": 8000}, cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    body = r.json()
+    assert body["completeness"] in ("brief", "incomplete", "fragmentary")
+    assert body["generated_title"] == "Brief or Incomplete Meeting Recording"
+    assert "No action items could be identified" in body["summary"]
+    assert "¹" in body["summary"]
+    # steps Thinking persistés (§12.7)
+    r = client.get(f"/api/v2/meetings/transcripts/{tid}/processing-steps",
+                   cookies=_cookies(session))
+    steps = {s["key"]: s["state"] for s in r.json()["steps"]}
+    assert steps.get("resolving_references") == "done"
+    assert steps.get("classifying_completeness") == "done"
+    # transcript groupé par source
+    r = client.get(f"/api/v2/meetings/transcripts/{tid}/transcript",
+                   cookies=_cookies(session))
+    assert r.json()["groups"][0]["source"] == "Bruno mic"
+    assert r.json()["groups"][0]["segments"][0]["starts_mid_utterance"] is True
+    # partage tracé (§3A.5)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/distributions",
+                    json={"channel": "copy_link"}, cookies=_cookies(session))
+    assert r.status_code == 200
+    # le titre IA ne touche jamais le titre de page hôte (§3A.5)
+    with get_conn() as conn:
+        title = conn.execute("SELECT title FROM pages WHERE id=?", (pid,)).fetchone()[0]
+    assert title == "Meeting @Today"
+
+
+def test_complete_meeting_structured_summary(client):
+    session, _ = _login(client)
+    tid = _mktr(client, session)
+    client.post(f"/api/v2/meetings/transcripts/{tid}/consents",
+                json={"method": "verbal"}, cookies=_cookies(session))
+    client.post(f"/api/v2/meetings/transcripts/{tid}/sessions",
+                json={"mode": "mic_only"}, cookies=_cookies(session))
+    words = ("We discussed the pricing launch for september. " * 20
+             + "Alice will prepare the release notes. "
+             + "We decided to ship on Friday morning. "
+             + "Bob should send the invoice before Monday. " * 3)
+    client.post(f"/api/v2/meetings/transcripts/{tid}/segments",
+                json={"segments": [
+                    {"channel": "self", "speaker": "Alice", "text": words,
+                     "start_ms": 0, "end_ms": 120000}]},
+                cookies=_cookies(session))
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/stop",
+                    json={"duration_ms": 300000}, cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert r.json()["completeness"] == "complete"
+    assert "## Action Items" in r.json()["summary"]
+
+
+def test_unicity_per_occurrence(client):
+    session, _ = _login(client)
+    p1, p2 = _mkpage("A"), _mkpage("B")
+    with get_conn() as conn:
+        t1 = conn.execute(
+            "INSERT INTO meeting_transcripts (page_id, event_occurrence_id)"
+            " VALUES (?, 'google:evt1:2026-10-15')", (p1,)).lastrowid
+        conn.commit()
+    _ = t1
+    import sqlite3
+    with get_conn() as conn:
+        try:
+            conn.execute(
+                "INSERT INTO meeting_transcripts (page_id, event_occurrence_id)"
+                " VALUES (?, 'google:evt1:2026-10-15')", (p2,))
+            conn.commit()
+            pytest.fail("unicity should reject the duplicate")
+        except sqlite3.IntegrityError:
+            pass
+
+
+def test_instructions_and_history(client):
+    session, _ = _login(client)
+    r = client.get("/api/v2/meetings/instructions", cookies=_cookies(session))
+    assert "auto" in [i["id"] for i in r.json()["instructions"]]
+    _mktr(client, session)
+    r = client.get("/api/v2/meetings/history?query=rec",
+                   cookies=_cookies(session))
+    assert r.status_code == 200
+
+
+# ── v7.1.1b : arrêt à vide, notes de réunion, arbre Workspace ───────────────
+
+def _start_empty_session(client, session):
+    pid = _mkpage("Empty take")
+    tid = _mktr(client, session, pid)
+    client.post(f"/api/v2/meetings/transcripts/{tid}/consents",
+                json={"method": "start_attestation"}, cookies=_cookies(session))
+    client.post(f"/api/v2/meetings/transcripts/{tid}/sessions",
+                json={"mode": "mic_only"}, cookies=_cookies(session))
+    return tid
+
+
+def test_stop_empty_without_flag_is_400_and_recoverable(client):
+    session, _ = _login(client)
+    tid = _start_empty_session(client, session)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/stop",
+                    json={}, cookies=_cookies(session))
+    assert r.status_code == 400
+    # La ligne ne reste jamais coincée en « processing ».
+    with get_conn() as conn:
+        status = conn.execute("SELECT status FROM meeting_transcripts WHERE id=?",
+                              (tid,)).fetchone()[0]
+    assert status == "failed"
+
+
+def test_stop_empty_with_flag_gives_diagnostic(client):
+    session, _ = _login(client)
+    tid = _start_empty_session(client, session)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/stop",
+                    json={"duration_ms": 5000, "allow_empty": True},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    body = r.json()
+    assert body["completeness"] == "fragmentary"
+    assert body["generated_title"] == "Brief or Incomplete Meeting Recording"
+    assert "No audio was captured" in body["summary"]
+
+
+def test_retry_from_stuck_processing_with_flag(client):
+    session, _ = _login(client)
+    tid = _start_empty_session(client, session)
+    # Ligne coincée en processing (cas du bug rapporté, transcript id 2).
+    with get_conn() as conn:
+        conn.execute("UPDATE meeting_transcripts SET status='processing' WHERE id=?",
+                     (tid,))
+        conn.commit()
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/processing:retry",
+                    json={"allow_empty": True}, cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert r.json()["completeness"] == "fragmentary"
+
+
+def _mkworkspace(uid, name="WS"):
+    with get_conn() as conn:
+        cur = conn.execute("INSERT INTO workspaces (name, owner_id) VALUES (?, ?)",
+                           (name, uid))
+        conn.commit()
+        return cur.lastrowid
+
+
+def _mkblockpage(ws_id, title, with_meeting):
+    import json as _json
+    content = _json.dumps([{"id": "b1", "type": "meeting" if with_meeting else "paragraph",
+                            "content": ""}])
+    with get_conn() as conn:
+        cur = conn.execute(
+            "INSERT INTO pages (workspace, workspace_id, title, content, content_format,"
+            " parent_section) VALUES ('t', ?, ?, ?, 'blocks', 'Private')",
+            (ws_id, title, content))
+        conn.commit()
+        return cur.lastrowid
+
+
+def test_meeting_notes_endpoint_and_ownership(client):
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    mp = _mkblockpage(ws, "Sprint Review", True)
+    _mkblockpage(ws, "Plain notes", False)
+    # Page d'un autre utilisateur : invisible.
+    s2, uid2 = _login(client)
+    ws2 = _mkworkspace(uid2)
+    _mkblockpage(ws2, "Someone else", True)
+    r = client.get("/api/v2/meetings/notes", cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    ids = [n["id"] for n in r.json()["notes"]]
+    assert mp in ids
+    assert len(ids) == 1
+    assert r.json()["notes"][0]["status"] == "idle"
+
+
+def test_workspace_tree_excludes_meeting_notes(client):
+    from app.routers.board import _load_workspace_pages
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    _mkblockpage(ws, "Sprint Review", True)
+    _mkblockpage(ws, "Plain notes", False)
+    pages = _load_workspace_pages(str(ws))
+    names = [p["name"] for p in pages]
+    assert "Plain notes" in names
+    assert "Sprint Review" not in names
+
+
+def test_sidebar_tree_endpoint_keeps_alpine_bindings(client):
+    """Contrat dont dépendent refreshSidebarTree + Alpine.initTree.
+
+    Régression /local-workspace : le fragment doit porter les bindings
+    (@contextmenu/showWsContext) et exclure les notes de réunion.
+    """
+    session, uid = _login(client)
+    ws = _mkworkspace(uid)
+    _mkblockpage(ws, "Sprint Review", True)
+    _mkblockpage(ws, "Plain notes", False)
+    cookies = _cookies(session)
+    cookies["flowdeck_workspace"] = str(ws)
+    r = client.get("/api/sidebar/workspace-tree", cookies=cookies)
+    assert r.status_code == 200, r.text
+    assert "showWsContext" in r.text
+    assert "Plain notes" in r.text
+    assert "Sprint Review" not in r.text
+
+
+# ── v7.69.5 : fin du 400 en boucle ──────────────────────────────────────────
+
+def test_summarize_empty_with_flag_gives_diagnostic(client):
+    session, _ = _login(client)
+    tid = _mktr(client, session)
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/summarize",
+                    json={"allow_empty": True}, cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert "No audio was captured" in r.json()["summary"]
+    with get_conn() as conn:
+        row = conn.execute("SELECT status, completeness_class FROM meeting_transcripts"
+                           " WHERE id=?", (tid,)).fetchone()
+    assert row["status"] == "done"
+    assert row["completeness_class"] == "fragmentary"
+
+
+def test_topup_after_failure_then_retry(client):
+    """Le manuel collé après un échec part au retry — plus de 400 en boucle."""
+    session, _ = _login(client)
+    tid = _start_empty_session(client, session)
+    # Premier stop refusé (sans flag) → failed, pas coincé.
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/stop",
+                    json={}, cookies=_cookies(session))
+    assert r.status_code == 400
+    # L'utilisateur colle un transcript : top-up autorisé hors session live.
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/segments",
+                    json={"segments": [
+                        {"channel": "self", "speaker": "Alice",
+                         "text": "We decided to ship on Friday morning. Alice will prepare "
+                                 "the release notes for the team. Bob should send the invoice "
+                                 "before Monday morning without fail.",
+                         "start_ms": 0, "end_ms": 60000}]},
+                    cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    r = client.post(f"/api/v2/meetings/transcripts/{tid}/processing:retry",
+                    json={}, cookies=_cookies(session))
+    assert r.status_code == 200, r.text
+    assert "Action Items" in r.json()["summary"]