ThoughtMCP

#!/bin/sh # ThoughtMCP MCP Server Entrypoint Script # # This script ensures all dependencies are ready before starting the MCP server: # 1. Waits for Ollama to be ready and responding # 2. Pulls the embedding model if not already present # 3. Starts the MCP server # # Requirements: 11.3, 9.2, 9.3 # # Environment Variables: # OLLAMA_HOST - Ollama API URL (default: http://ollama:11434) # EMBEDDING_MODEL - Model to pull (default: nomic-embed-text) # OLLAMA_WAIT_TIMEOUT - Max seconds to wait for Ollama (default: 120) # SKIP_MODEL_PULL - Skip model pull if set to "true" (default: false) # MCP_STANDBY_MODE - If "true", container stays alive without starting server (default: false) # Use this when MCP clients connect via "docker exec -i <container> node dist/index.js" # # Usage: # ./docker/entrypoint.sh # OLLAMA_HOST=http://localhost:11434 ./docker/entrypoint.sh set -e # Configuration with defaults OLLAMA_HOST="${OLLAMA_HOST:-http://ollama:11434}" EMBEDDING_MODEL="${EMBEDDING_MODEL:-nomic-embed-text}" OLLAMA_WAIT_TIMEOUT="${OLLAMA_WAIT_TIMEOUT:-120}" SKIP_MODEL_PULL="${SKIP_MODEL_PULL:-false}" MCP_STANDBY_MODE="${MCP_STANDBY_MODE:-false}" # Logging functions log_info() { echo "[INFO] $(date '+%Y-%m-%d %H:%M:%S') - $1" } log_warn() { echo "[WARN] $(date '+%Y-%m-%d %H:%M:%S') - $1" } log_error() { echo "[ERROR] $(date '+%Y-%m-%d %H:%M:%S') - $1" >&2 } # Wait for Ollama to be ready wait_for_ollama() { log_info "Waiting for Ollama at ${OLLAMA_HOST}..." local elapsed=0 local interval=5 while [ $elapsed -lt $OLLAMA_WAIT_TIMEOUT ]; do # Check if Ollama API is responding if curl -sf "${OLLAMA_HOST}/api/tags" > /dev/null 2>&1; then log_info "Ollama is ready!" return 0 fi log_info "Ollama not ready yet, waiting... (${elapsed}s/${OLLAMA_WAIT_TIMEOUT}s)" sleep $interval elapsed=$((elapsed + interval)) done log_error "Timeout waiting for Ollama after ${OLLAMA_WAIT_TIMEOUT} seconds" log_error "Please ensure Ollama is running at ${OLLAMA_HOST}" return 1 } # Check if model is already present is_model_present() { local model_name="$1" # Query Ollama for available models local models_response models_response=$(curl -sf "${OLLAMA_HOST}/api/tags" 2>/dev/null) || return 1 # Check if model is in the list (simple string match) if echo "$models_response" | grep -q "\"name\":\"${model_name}\""; then return 0 fi # Also check for model with :latest tag if echo "$models_response" | grep -q "\"name\":\"${model_name}:latest\""; then return 0 fi return 1 } # Pull embedding model if not present pull_embedding_model() { if [ "$SKIP_MODEL_PULL" = "true" ]; then log_info "Skipping model pull (SKIP_MODEL_PULL=true)" return 0 fi log_info "Checking for embedding model: ${EMBEDDING_MODEL}" if is_model_present "$EMBEDDING_MODEL"; then log_info "Model ${EMBEDDING_MODEL} is already present" return 0 fi log_info "Pulling embedding model: ${EMBEDDING_MODEL}..." log_info "This may take several minutes on first run..." # Pull the model using Ollama API local pull_response local http_code # Use curl to pull the model, streaming the response # The pull endpoint returns JSON lines with status updates http_code=$(curl -sf -w "%{http_code}" -o /tmp/pull_output.txt \ -X POST "${OLLAMA_HOST}/api/pull" \ -H "Content-Type: application/json" \ -d "{\"name\": \"${EMBEDDING_MODEL}\", \"stream\": false}" 2>/dev/null) || true if [ "$http_code" = "200" ]; then log_info "Successfully pulled model: ${EMBEDDING_MODEL}" rm -f /tmp/pull_output.txt return 0 fi # If non-streaming failed, try streaming approach log_info "Attempting streaming pull..." if curl -sf -X POST "${OLLAMA_HOST}/api/pull" \ -H "Content-Type: application/json" \ -d "{\"name\": \"${EMBEDDING_MODEL}\"}" 2>&1 | while read -r line; do # Parse status from JSON response status=$(echo "$line" | grep -o '"status":"[^"]*"' | cut -d'"' -f4) if [ -n "$status" ]; then log_info "Pull status: $status" fi done; then log_info "Successfully pulled model: ${EMBEDDING_MODEL}" return 0 fi log_error "Failed to pull model: ${EMBEDDING_MODEL}" log_error "Please ensure the model name is correct and Ollama has network access" return 1 } # Verify model is working by generating a test embedding verify_model() { log_info "Verifying model ${EMBEDDING_MODEL} is working..." local response response=$(curl -sf -X POST "${OLLAMA_HOST}/api/embeddings" \ -H "Content-Type: application/json" \ -d "{\"model\": \"${EMBEDDING_MODEL}\", \"prompt\": \"test\"}" 2>/dev/null) || { log_warn "Could not verify model, but continuing anyway..." return 0 } # Check if response contains embedding if echo "$response" | grep -q '"embedding"'; then log_info "Model verification successful!" return 0 fi log_warn "Model verification returned unexpected response, but continuing..." return 0 } # Start the MCP server start_server() { log_info "Starting ThoughtMCP MCP Server..." log_info "Database: ${DB_HOST:-postgres}:${DB_PORT:-5432}/${DB_NAME:-thoughtmcp}" log_info "Ollama: ${OLLAMA_HOST}" log_info "Model: ${EMBEDDING_MODEL}" # Execute the Node.js server, replacing this process exec node dist/index.js "$@" } # Standby mode - keep container alive for docker exec connections standby_mode() { log_info "Running in MCP_STANDBY_MODE - container ready for MCP client connections" log_info "Connect via: docker exec -i thoughtmcp-server node dist/index.js" log_info "Database: ${DB_HOST:-postgres}:${DB_PORT:-5432}/${DB_NAME:-thoughtmcp}" log_info "Ollama: ${OLLAMA_HOST}" log_info "Model: ${EMBEDDING_MODEL}" # Keep container alive - sleep indefinitely exec tail -f /dev/null } # Main execution main() { log_info "ThoughtMCP Entrypoint Script Starting..." log_info "Environment: ${NODE_ENV:-development}" # Step 1: Wait for Ollama if ! wait_for_ollama; then log_error "Failed to connect to Ollama. Exiting." exit 1 fi # Step 2: Pull embedding model if needed if ! pull_embedding_model; then log_error "Failed to pull embedding model. Exiting." exit 1 fi # Step 3: Verify model is working verify_model # Step 4: Start the server or enter standby mode if [ "$MCP_STANDBY_MODE" = "true" ]; then standby_mode else start_server "$@" fi } # Run main function with all arguments main "$@"