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

# Ollama + Embeddings Quick Start **Version:** 1.0.0 **Last Updated:** 2025-10-18 --- ## 🎯 Overview This guide helps you quickly enable **vector embeddings** with **Ollama** in Docker. --- ## 🚀 Quick Start (Docker) ### Option 1: Enable via Environment Variables **1. Create `.env` file:** ```bash cp env.example .env ``` **2. Edit `.env` and enable embeddings:** ```bash # Feature Flags MIMIR_FEATURE_VECTOR_EMBEDDINGS=true # Embeddings Configuration MIMIR_EMBEDDINGS_ENABLED=true MIMIR_EMBEDDINGS_MODEL=nomic-embed-text ``` **3. Start with Ollama profile:** ```bash # Start all services including Ollama docker compose --profile ollama up -d # Or set profile in environment export COMPOSE_PROFILES=ollama docker compose up -d ``` **4. Pull embedding model:** ```bash # Inside Ollama container docker exec ollama_server ollama pull nomic-embed-text # Or from host (if Ollama running locally) ollama pull nomic-embed-text ``` **5. Index files with embeddings:** ```bash # Inside MCP server container docker exec mcp_server node setup-watch.js # Files will be indexed with embeddings automatically ``` --- ### Option 2: Enable via docker compose.yml **1. Edit `docker compose.yml`:** ```yaml mcp-server: environment: # Enable embeddings - MIMIR_FEATURE_VECTOR_EMBEDDINGS=true - MIMIR_EMBEDDINGS_ENABLED=true - MIMIR_EMBEDDINGS_MODEL=nomic-embed-text ``` **2. Uncomment Ollama dependency (optional):** ```yaml mcp-server: depends_on: neo4j: condition: service_healthy ollama: condition: service_healthy # Add this ``` **3. Start services:** ```bash docker compose --profile ollama up -d ``` --- ## 🔍 Verify Setup ### Check Services ```bash # All services running docker compose ps # Expected output: # neo4j_db running # ollama_server running (if profile enabled) # mcp_server running # Check Ollama models docker exec ollama_server ollama list ``` ### Check Embeddings Status ```bash # Inside MCP container docker exec mcp_server node -e " const { LLMConfigLoader } = require('./build/config/LLMConfigLoader.js'); (async () => { const loader = LLMConfigLoader.getInstance(); const enabled = await loader.isVectorEmbeddingsEnabled(); console.log('Vector embeddings enabled:', enabled); })(); " ``` ### Test Vector Search ```bash # Query for embedding stats docker exec mcp_server node -e " const { createGraphManager } = require('./build/managers/index.js'); (async () => { const gm = await createGraphManager(); const session = gm.getDriver().session(); const result = await session.run(\` MATCH (f:File) RETURN COUNT(f) AS total, SUM(CASE WHEN f.has_embedding = true THEN 1 ELSE 0 END) AS with_embeddings \`); const record = result.records[0]; console.log('Files:', record.get('total').toNumber()); console.log('With embeddings:', record.get('with_embeddings').toNumber()); await session.close(); await gm.close(); })(); " ``` --- ## 🛠️ Available Models ### Recommended: nomic-embed-text ```bash docker exec ollama_server ollama pull nomic-embed-text ``` - **Dimensions:** 768 - **Size:** ~274MB - **Speed:** Fast - **Best for:** General purpose code/text ### Alternative: nomic-embed-text ```bash docker exec ollama_server ollama pull nomic-embed-text ``` - **Dimensions:** 1024 - **Size:** ~670MB - **Speed:** Medium - **Best for:** Higher accuracy needs ### Alternative: all-minilm ```bash docker exec ollama_server ollama pull all-minilm ``` - **Dimensions:** 384 - **Size:** ~120MB - **Speed:** Very fast - **Best for:** Testing, quick iterations --- ## 📊 Usage Examples ### Semantic File Search ```bash # Via agent chain docker exec mcp_server npm run chain "find files about database connections" # The agent will use vector_search_files tool automatically ``` ### Direct MCP Tool Usage See [VECTOR_EMBEDDINGS_GUIDE.md](./VECTOR_EMBEDDINGS_GUIDE.md) for detailed examples. --- ## 🔄 Enable/Disable Without Rebuild ### Disable Embeddings **Edit `.env`:** ```bash MIMIR_FEATURE_VECTOR_EMBEDDINGS=false MIMIR_EMBEDDINGS_ENABLED=false ``` **Restart:** ```bash docker compose restart mcp-server ``` **Ollama keeps running** but embeddings won't be used. ### Re-enable **Edit `.env`:** ```bash MIMIR_FEATURE_VECTOR_EMBEDDINGS=true MIMIR_EMBEDDINGS_ENABLED=true ``` **Restart:** ```bash docker compose restart mcp-server ``` --- ## 🐛 Troubleshooting ### Ollama Not Starting **Check profile:** ```bash docker compose ps ollama # If not listed, profile not enabled docker compose --profile ollama up -d ``` ### Model Not Found **Pull model:** ```bash docker exec ollama_server ollama pull nomic-embed-text # Check models docker exec ollama_server ollama list ``` ### Connection Refused **Check Ollama health:** ```bash docker compose ps ollama_server # Should show "healthy" ``` **Check URL in MCP server:** ```bash docker exec mcp_server env | grep OLLAMA # Should show: OLLAMA_BASE_URL=http://ollama:11434 ``` ### No Embeddings Generated **Check feature flags:** ```bash docker exec mcp_server env | grep MIMIR_ # Should show: # MIMIR_FEATURE_VECTOR_EMBEDDINGS=true # MIMIR_EMBEDDINGS_ENABLED=true ``` **Re-index files:** ```bash # Remove old watch (TODO: add tool) # Re-run setup docker exec mcp_server node setup-watch.js ``` --- ## 📚 Related Documentation - **[Vector Embeddings Guide](./VECTOR_EMBEDDINGS_GUIDE.md)** - Complete embeddings documentation - **[File Watching Guide](./FILE_WATCHING_GUIDE.md)** - File indexing setup - **[Docker Deployment Guide](./DOCKER_DEPLOYMENT_GUIDE.md)** - Container setup --- ## 🚀 Production Deployment ### With GPU Support (NVIDIA) **Edit `docker compose.yml`:** ```yaml ollama: deploy: resources: reservations: devices: - driver: nvidia count: 1 capabilities: [gpu] ``` **Requires:** - NVIDIA GPU - nvidia-docker2 - CUDA drivers ### Resource Allocation **Increase if needed:** ```yaml ollama: deploy: resources: limits: memory: 8G cpus: '4' ``` --- **Questions?** Check [GitHub Issues](https://github.com/orneryd/Mimir/issues) or open a discussion.