✨ Feature / Enhancement PR

🔗 Epic / Issue

Closes #1910, #172

🚀 Summary (1-2 sentences)

Added configurable AUTO_REFRESH_SERVERS feature that automatically syncs gateway tools, resources, and prompts with the database during health checks. When enabled, the system detects changes from MCP servers and performs add/update/delete operations only when necessary.

The tools, prompts, and resources are compared against those stored in the database, and any required add, update, or delete operations are performed.

🔧 Usage

Enable via environment variable:

AUTO_REFRESH_SERVERS=true

1. HEALTH_CHECK_INTERVAL=30 # Optional: shorter interval for testing
2. Add an mcp server with 3 tools in the gateway.
3. Update the mcp server by reducing the number of tools or adding some tools.
4. Wait for the health check and notice that the new tools are automatically refreshed in the DB.
5. Same applies for prompts and resources as well.

🧪 Checks

• make lint passes
• make test passes
• CHANGELOG updated (if user-facing)

📓 Notes (optional)

flowchart TD
    A[Health Check Loop] -->|Every HEALTH_CHECK_INTERVAL| B{Gateway Healthy?}
    B -->|No| C[Handle Failure]
    B -->|Yes| D{AUTO_REFRESH_SERVERS?}
    D -->|No| E[Update last_seen]
    D -->|Yes| F[Fetch tools from MCP Server]
    F --> G{Any Changes?}
    G -->|No| E
    G -->|Yes| H[Sync DB: Add/Update/Delete]
    H --> I[Invalidate Caches]
    I --> E

• The feature reuses existing _update_or_create_tools/resources/prompts helper methods
• Uses fresh_db_session() to avoid holding DB connections during HTTP calls to MCP servers
• SQLAlchemy's db.dirty tracking detects in-place updates to determine if commit is needed
• Tool lookup cache is invalidated per-gateway for efficiency