api_database-api/.trae/documents/Implement Tool Management System for Dynamic Agents.md

Implement Tool Management for Dynamic Agents

I will implement a system to manage "Tools" (Function Calls) and associate them with Dynamic Agents, allowing the frontend to easily configure which tools an agent can use.

1. Database Schema Expansion

I will create two new tables in PostgreSQL:

• tools: Stores reusable tool definitions (templates).
  • tool_id: SERIAL PRIMARY KEY
  • name: VARCHAR (Unique, e.g., "get_weather")
  • description: TEXT (For LLM understanding)
  • parameters: JSONB (JSON Schema for arguments)
  • created_at: TIMESTAMP
• agent_tools: Maps agents to tools (Many-to-Many).
  • agent_card_id: INTEGER (FK)
  • tool_id: INTEGER (FK)
  • config: JSONB (Optional, for agent-specific tool config if needed)

2. API Implementation (main.py)

Pydantic Models

• ToolCreate: Validation for creating new tool definitions.
• ToolResponse: API response format.
• AgentToolAssignment: For linking tools to agents.

New Endpoints

• Tool Management (CRUD):
  • POST /tools/: Create a new tool definition.
  • GET /tools/: List all available tools.
  • PUT /tools/{tool_id}: Update a tool.
  • DELETE /tools/{tool_id}: Delete a tool.
• Agent Tool Association:
  • POST /agent/{card_id}/tools: Assign a list of tools to an agent.
  • GET /agent/{card_id}/tools: Get all tools assigned to an agent.
  • DELETE /agent/{card_id}/tools/{tool_id}: Remove a tool from an agent.

Update Existing Logic

• Modify GET /dynamic_agent/{card_id}:
  • Update the query to automatically fetch and include the list of associated tools in the agent details response. This ensures the frontend gets everything in one call.

3. Verification

• Create a "Weather Query" tool definition.
• Assign it to a Dynamic Agent.
• Verify the agent details endpoint returns the tool info correctly.