Add admin panel, WebSocket support, and API versioning

Introduce an admin portal (React + Nginx), WebSocket routing, and
API versioning middleware with `/api/v1/` prefix deprecation.
Add master API key authentication, new Prometheus metrics for AI
token consumption and active WebSockets, and extend S3 config
with a public endpoint URL. Update test paths and fixtures to
align with the new routing structure.

docs/SHAARLI-INTEGRATION.md

@@ -0,0 +1,45 @@
+# Shaarli Integration
+
+The **Imago Hub** can be integrated directly with **Shaarli** (the PHP-based bookmark manager) to automatically process and enhance your bookmarks when you save them.
+
+## Overview
+
+When you save a new bookmark in Shaarli, you might want to:
+1. Automatically fetch the URL and summarize the page content using AI
+2. Generate tags automatically based on the page content
+3. Archive full-page screenshots to Imago for offline reading
+
+This repository includes a reference implementation for this architecture using the Imago Python SDK.
+
+## Components
+
+The `integration/shaarli/` directory contains:
+- `hub_plugin.py` — A Python class (`ShaarliHubPlugin`) that acts as a middleware interacting with the `imago-client` SDK.
+- `config_example.env` — Example environment variable file for the plugin layer.
+- `example_usage.py` — A standalone script demonstrating how the plugin summarizes URLs.
+
+## Setup Instructions
+
+1. **Install the SDK:** The script requires the Imago Python SDK.
+   ```bash
+   cd sdk
+   pip install imago-client
+   ```
+
+2. **Configure Environment variables:**
+   Copy `integration/shaarli/config_example.env` to `.env` and fill it with your **Imago Admin API Key** or a specific Client API Key provisioned with `ai:use` and `images:write` scopes.
+
+3. **Deploy as a microservice or hook:**
+   Since Shaarli is written in PHP, to hook python scripts into Shaarli you have a few options:
+   
+   - **Option A (WebHook):** Deploy a small FastAPI/Flask wrapper around `hub_plugin.py` and configure Shaarli's WebHook plugin to POST data to your wrapper whenever a new link is created.
+   - **Option B (Cron/Background Job):** Have a Python daemon watch the Shaarli RSS/Atom feed or poll its REST API, grab new entries, process them via `hub_plugin.py`, and PATCH the bookmark back in Shaarli with the newly generated Description and Tags using the Shaarli REST API.
+
+## Example Flow (Option B)
+
+1. You save `https://github.com/imago-project/imago` to Shaarli.
+2. The Python daemon detects the new bookmark via Shaarli's API.
+3. The daemon calls `plugin.process_bookmark(url)`.
+4. Imago leverages OpenRouter/Gemini to fetch and summarize the GitHub page.
+5. Imago returns `{"title": "...", "description": "...", "suggested_tags": ["github", "python", "ai"]}`.
+6. The daemon PATCHes the Shaarli API to enrich the bookmark.