College Football MCP

# Changes Made - Research MCP Integration & Fixes ## Summary This update includes: 1. Team name normalization improvements 2. Response generation edge case fixes 3. Research MCP integration for precision score checking 4. New tool function: `research_latest_score` ## Changes ### 1. Team Name Normalization (`src/team_normalizer.py`) - Added new utility module for normalizing team names - Handles variations like "Alabama Crimson Tide" vs "Alabama" - Supports fuzzy matching for better API compatibility - Updated `src/odds_api.py` to use normalization in `find_game_by_teams` and `find_next_game_for_team` ### 2. Response Generation Fixes (`agent_service/main.py`) - Improved final response generation logic - Better fallback handling when tool calls succeed but response is empty - Enhanced error recovery with comprehensive tool result summarization - Added temperature parameter for more natural responses ### 3. Research MCP Integration - **New file**: `agent_service/research_client.py` - Client for interacting with research-mcp server at `prompt80.com/research-mcp` - Supports async HTTP calls with proper error handling - **Updated**: `agent_service/main.py` - Added `research_latest_score` tool function - Integrated ResearchClient - Updated system instructions to include research tool - Tool can be used for precision checking of latest scores ### 4. Docker Configuration - Updated `docker-compose.yml` to include optional `RESEARCH_MCP_URL` and `RESEARCH_MCP_API_KEY` environment variables - Updated `agent_service/Dockerfile` to include `research_client.py` ## New Tool: `research_latest_score` This tool uses the research-mcp server to find the latest, most up-to-date scores for college football games. It's particularly useful for: - Verifying the most recent score when primary data sources might be slightly outdated - Getting real-time information for games happening today - Precision checking when users ask for "the latest" or "most recent" scores **Usage**: The LLM can call this tool automatically when it detects the need for precision checking, or when users explicitly ask for the latest information. ## Testing To test the changes: 1. **Team Name Normalization**: - Test with variations like "Alabama Crimson Tide", "Alabama", "Bama" - Should all match games correctly 2. **Response Generation**: - Test queries that previously returned fallback messages - Responses should now be more complete and informative 3. **Research MCP Integration**: - Test queries asking for "latest score" or "most recent result" - The system should use research_latest_score when appropriate - Example: "What's the latest score for Alabama today?" ## Deployment The changes are backward compatible. No breaking changes to existing APIs. **Environment Variables** (optional): - `RESEARCH_MCP_URL`: Defaults to `https://prompt80.com/research-mcp` - `RESEARCH_MCP_API_KEY`: Optional API key if research-mcp requires authentication