M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

# Quick Start Guide Get the validation orchestrator running in 5 minutes! --- ## ✅ Step 1: Start Copilot Proxy ```bash # Ensure copilot-api is installed npm install -g copilot-api # Start proxy copilot-api start & # Verify curl http://localhost:4141/v1/models ``` **Expected**: JSON response with model list --- ## ✅ Step 2: Test Connection ```bash node -e "const {ChatOpenAI} = require('@langchain/openai'); const llm = new ChatOpenAI({apiKey: 'dummy-key-not-used', configuration: {baseURL: 'http://localhost:4141/v1'}}); llm.invoke('Hello!').then(r => console.log('✅', r.content));" ``` **Expected**: `✅ Hi! How can I assist you today?` --- ## ✅ Step 3: Run First Validation ```bash # Validate claudette-debug (baseline test) npm run validate docs/agents/claudette-debug.md benchmarks/debug-benchmark.json ``` **Expected Output**: ``` 🔍 Validating agent: docs/agents/claudette-debug.md 📋 Benchmark: benchmarks/debug-benchmark.json ⚙️ Executing benchmark task... ✅ Task completed in ~12,000 tokens 📊 Evaluating output against rubric... 📈 Total score: XX/100 📄 Report saved to: validation-output/2025-10-15_claudette-debug.md ``` --- ## ✅ Step 4: Review Results ```bash # View report cat validation-output/2025-10-15_claudette-debug.md # View raw data cat validation-output/2025-10-15_claudette-debug.json ``` --- ## 🎯 What's Next? ### Test Agentinator (Two-Hop Validation) Coming soon! Will test agents generated by the agentinator. ### Create Custom Benchmarks See `benchmarks/debug-benchmark.json` for format. ### Batch Testing Run multiple agents in parallel: ```bash for agent in docs/agents/claudette-*.md; do npm run validate "$agent" benchmarks/debug-benchmark.json done ``` --- ## 🐛 Common Issues ### Port 4141 in use ```bash # Kill existing copilot-api pkill -f copilot-api # Restart copilot-api start & ``` ### Authentication errors ```bash # Re-auth with GitHub CLI gh auth login ``` ### TypeScript errors ```bash # Ensure dependencies installed npm install # Build npm run build ``` --- ## 📚 Full Documentation - **Tool Design**: `docs/agents/VALIDATION_TOOL_DESIGN.md` - **Setup Guide**: `tools/SETUP.md` - **Component README**: `src/orchestrator/README.md` --- **Ready!** You now have a working agent validation system. 🎉 Get the validation orchestrator running in 5 minutes! --- ## ✅ Step 1: Start Copilot Proxy ```bash # Ensure copilot-api is installed npm install -g copilot-api # Start proxy copilot-api start & # Verify curl http://localhost:4141/v1/models ``` **Expected**: JSON response with model list --- ## ✅ Step 2: Test Connection ```bash node -e "const {ChatOpenAI} = require('@langchain/openai'); const llm = new ChatOpenAI({apiKey: 'dummy-key-not-used', configuration: {baseURL: 'http://localhost:4141/v1'}}); llm.invoke('Hello!').then(r => console.log('✅', r.content));" ``` **Expected**: `✅ Hi! How can I assist you today?` --- ## ✅ Step 3: Run First Validation ```bash # Validate claudette-debug (baseline test) npm run validate docs/agents/claudette-debug.md benchmarks/debug-benchmark.json ``` **Expected Output**: ``` 🔍 Validating agent: docs/agents/claudette-debug.md 📋 Benchmark: benchmarks/debug-benchmark.json ⚙️ Executing benchmark task... ✅ Task completed in ~12,000 tokens 📊 Evaluating output against rubric... 📈 Total score: XX/100 📄 Report saved to: validation-output/2025-10-15_claudette-debug.md ``` --- ## ✅ Step 4: Review Results ```bash # View report cat validation-output/2025-10-15_claudette-debug.md # View raw data cat validation-output/2025-10-15_claudette-debug.json ``` --- ## 🎯 What's Next? ### Test Agentinator (Two-Hop Validation) Coming soon! Will test agents generated by the agentinator. ### Create Custom Benchmarks See `benchmarks/debug-benchmark.json` for format. ### Batch Testing Run multiple agents in parallel: ```bash for agent in docs/agents/claudette-*.md; do npm run validate "$agent" benchmarks/debug-benchmark.json done ``` --- ## 🐛 Common Issues ### Port 4141 in use ```bash # Kill existing copilot-api pkill -f copilot-api # Restart copilot-api start & ``` ### Authentication errors ```bash # Re-auth with GitHub CLI gh auth login ``` ### TypeScript errors ```bash # Ensure dependencies installed npm install # Build npm run build ``` --- ## 📚 Full Documentation - **Tool Design**: `docs/agents/VALIDATION_TOOL_DESIGN.md` - **Setup Guide**: `tools/SETUP.md` - **Component README**: `src/orchestrator/README.md` --- **Ready!** You now have a working agent validation system. 🎉