SDOF Knowledge Base

# SDOF Knowledge Base MCP Server: Final Setup Guide This guide will help you finalize the setup of your SDOF Knowledge Base MCP server and start using it with the auto-save functionality. ## 1. Apply the Plan Auto-Save Updates You need to update the main index.ts file to include the plan auto-save functionality: 1. Open `src/index.ts` in your editor 2. Follow the instructions in `src/index.ts.plan-auto-save-update` to add: - The import for PlanAutoSaveService - The StoreSdofPlanArgs type definition - The isStoreSdofPlanArgs validation function - The store_sdof_plan tool definition to the tools array - The case for store_sdof_plan in the switch statement - The handleStoreSdofPlan method to the SdofKnowledgeBaseServer class Or alternatively, manually merge the changes from the update file into the main index.ts file. ## 2. Build and Run the Server Now that you've applied all the updates, you can build and run the server: ```bash # Build the server npm run build # Run the server npm start ``` ## 3. Configure MongoDB Before using the server, you need to set up MongoDB with vector search capabilities: 1. Follow the instructions in `docs/MONGODB_SETUP.md` to: - Create a MongoDB Atlas account (if you don't have one) - Set up a cluster - Configure network access - Create the database and collection - Enable Atlas Vector Search - Create the vector index ## 4. Configure the MCP Server in Roo To use the SDOF Knowledge Base MCP server with Roo, you need to add it to your MCP configuration: 1. Open Roo 2. Go to Settings → MCP 3. Add a new MCP server with the following configuration: ```json { "name": "sdof_knowledge_base", "type": "stdio", "command": "npm start", "cwd": "c:/Users/thegr/integration-main/integration/sdof_knowledge_base" } ``` Make sure to adjust the `cwd` path to match your actual installation directory. ## 5. Update Your SDOF Orchestrator Custom Mode To make the SDOF Orchestrator automatically save plans to the knowledge base: 1. Open your custom_modes.json file: ``` ../../AppData/Roaming/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/custom_modes.json ``` 2. Find the SDOF Orchestrator mode (look for `"slug": "sdof-orchestrator"`) 3. Add the post-phase processing instructions to the custom_instructions section as shown in `docs/CUSTOM_MODE_EXAMPLE.md`: ``` "Post-Phase Processing: After each phase is completed and returns results via `attempt_completion`, you should: 1. Automatically save the results to the knowledge base using the `store_sdof_plan` MCP tool from the `sdof_knowledge_base` server. 2. Use appropriate plan types based on the phase: - After Explorer results: planType = 'exploration' - After Analyzer results: planType = 'analysis' - After Implementer results: planType = 'implementation' - After Evaluator results: planType = 'evaluation' - After Integrator results: planType = 'integration' - Final synthesis: planType = 'synthesis' 3. Generate appropriate tags based on the problem domain and content. 4. Include a standardized title format: 'SDOF [PlanType]: [Brief Problem Description]' 5. Report to the user that the plan has been saved to the knowledge base." ``` ## 6. Test the Auto-Save Functionality To test that everything is working correctly: 1. Start the SDOF Knowledge Base MCP server: ``` cd sdof_knowledge_base npm start ``` 2. In another terminal, you can manually test the store_sdof_plan tool with curl: ``` curl -X POST -H "Content-Type: application/json" -d '{ "jsonrpc": "2.0", "id": "test", "method": "callTool", "params": { "name": "store_sdof_plan", "arguments": { "title": "Test Plan", "content": "# Test Plan\n\nThis is a test plan.", "planType": "exploration", "tags": ["test", "sdof"] } } }' http://localhost:3000/ ``` 3. Verify that the plan was saved to: - The filesystem at `docs/plans/explorations/YYYY-MM-DD-test-plan.md` - The knowledge base (you can query it using the search_knowledge tool) ## Troubleshooting If you encounter any issues: 1. **MongoDB Connection Issues**: - Check that your MongoDB Atlas cluster is running - Verify that your IP address is in the allowed list - Check the connection string in your .env file 2. **Embedding Issues**: - Make sure your Claude API key is valid - Check the embedding service logs for any errors 3. **MCP Server Issues**: - Check that the @modelcontextprotocol/sdk version is correct (0.8.1) - Verify that all dependencies are installed 4. **Filesystem Issues**: - Ensure that the docs/plans directories have write permissions ## Additional Resources - `docs/SETUP_GUIDE.md`: Detailed MongoDB setup instructions - `docs/AUTO_SAVE_GUIDE.md`: More information about the auto-save functionality - `docs/CUSTOM_MODE_EXAMPLE.md`: Details on customizing the SDOF Orchestrator mode - `docs/ALTERNATIVE_IMPLEMENTATION.md`: Alternative implementations if you encounter SDK issues - `docs/MONGODB_SETUP.md`: MongoDB Atlas vector search setup guide - `docs/FINAL_REPORT.md`: Complete summary of the implementation With these steps completed, your SDOF Knowledge Base MCP server should be fully operational and ready to use with the SDOF Orchestrator for auto-saving plans.