feat(ai): commandes @/ & skills, analyse d'images et capacites des modeles (#81)

backend/ai_chat.py

@@ -17,6 +17,7 @@ from __future__ import annotations
 
 import json
 import logging
+import re
 from collections.abc import AsyncIterator
 from dataclasses import dataclass, field
 from typing import Any
@@ -162,6 +163,46 @@ def _gemini_tools(tools: list[dict[str, Any]] | None) -> list[dict[str, Any]] |
     return [{"functionDeclarations": declarations}]
 
 
+_DATA_URL_RE = re.compile(r"^data:([^;,]+);base64,(.*)$", re.DOTALL)
+
+
+def _content_to_gemini_parts(content: Any) -> list[dict[str, Any]]:
+    """Convert OpenAI-style message content to Gemini ``parts``.
+
+    Accepts either a plain string or a multimodal content array
+    (``[{"type": "text", ...}, {"type": "image_url", ...}]``). Data URLs are
+    turned into ``inlineData`` parts so images can be sent to vision models.
+    """
+    if content is None:
+        return [{"text": ""}]
+    if isinstance(content, str):
+        return [{"text": content}]
+
+    parts: list[dict[str, Any]] = []
+    if isinstance(content, list):
+        for item in content:
+            if isinstance(item, str):
+                parts.append({"text": item})
+                continue
+            if not isinstance(item, dict):
+                continue
+            item_type = item.get("type")
+            if item_type == "text":
+                parts.append({"text": item.get("text", "")})
+            elif item_type == "image_url":
+                url = (item.get("image_url") or {}).get("url", "")
+                match = _DATA_URL_RE.match(url or "")
+                if match:
+                    parts.append({
+                        "inlineData": {"mimeType": match.group(1), "data": match.group(2)},
+                    })
+                elif url:
+                    parts.append({"fileData": {"fileUri": url}})
+    if not parts:
+        parts.append({"text": ""})
+    return parts
+
+
 def _gemini_contents(messages: list[dict[str, Any]]) -> tuple[str, list[dict[str, Any]]]:
     """Split OpenAI-style messages into Gemini ``system`` text + ``contents``."""
     system_parts: list[str] = []
@@ -169,7 +210,8 @@ def _gemini_contents(messages: list[dict[str, Any]]) -> tuple[str, list[dict[str
     for msg in messages:
         role = msg.get("role")
         if role == "system":
-            system_parts.append(msg.get("content") or "")
+            content = msg.get("content") or ""
+            system_parts.append(content if isinstance(content, str) else "")
         elif role == "tool":
             contents.append({
                 "role": "user",
@@ -183,7 +225,7 @@ def _gemini_contents(messages: list[dict[str, Any]]) -> tuple[str, list[dict[str
         else:
             contents.append({
                 "role": "user" if role == "user" else "model",
-                "parts": [{"text": msg.get("content") or ""}],
+                "parts": _content_to_gemini_parts(msg.get("content")),
             })
     return "\n".join(p for p in system_parts if p).strip(), contents