# =============================================================================
# Gemini Researcher MCP Server - Dockerfile
# =============================================================================
# Multi-stage build for a lightweight MCP server image
#
# This Dockerfile builds the gemini-researcher MCP server, which proxies
# research queries to Gemini CLI for codebase analysis.
#
# Requirements:
# - GEMINI_API_KEY environment variable must be set at runtime
# - Mount your project directory to /workspace for analysis
#
# Usage:
# docker build -t gemini-researcher .
# docker run -e GEMINI_API_KEY="your-key" -v /path/to/project:/workspace gemini-researcher
# =============================================================================
# -----------------------------------------------------------------------------
# Stage 1: Build
# -----------------------------------------------------------------------------
FROM node:22-alpine AS builder
WORKDIR /app
# Copy package files for dependency installation
COPY package.json package-lock.json* ./
# Install all dependencies (including devDependencies for build)
RUN npm ci
# Copy source files
COPY tsconfig.json ./
COPY src/ ./src/
# Build TypeScript to JavaScript
RUN npm run build
# -----------------------------------------------------------------------------
# Stage 2: Production
# -----------------------------------------------------------------------------
FROM node:22-alpine AS production
# Install git (required for .gitignore parsing by the 'ignore' package)
# Install build dependencies for native modules (required for arm64 compilation of tree-sitter)
RUN apk add --no-cache git python3 make g++
WORKDIR /app
# Copy package files
COPY package.json package-lock.json* ./
# Install production dependencies only
RUN npm ci --omit=dev
# Update npm to latest version to fix security vulnerabilities (CVE-2026-23950, CVE-2026-23745, CVE-2025-64756)
RUN npm install -g npm@latest && npm cache clean --force
# Install Gemini CLI globally (required for proxying queries)
RUN npm install -g @google/gemini-cli
# Remove build dependencies to reduce image size (keep git, it's needed at runtime)
RUN apk del python3 make g++
# Copy built files from builder stage
COPY --from=builder /app/dist ./dist
# Create workspace directory for mounting projects
RUN mkdir -p /workspace
WORKDIR /workspace
# Environment variables
# GEMINI_API_KEY must be provided at runtime
ENV NODE_ENV=production
ENV PROJECT_ROOT=/workspace
# The MCP server uses stdio transport - it reads from stdin and writes to stdout
# This makes it compatible with MCP clients like Claude Desktop
# Health check is not applicable for stdio-based servers
# The server validates Gemini CLI on startup
# Entry point - run the MCP server
ENTRYPOINT ["node", "/app/dist/index.js"]
# Default command (can be overridden for setup wizard)
# To run setup wizard: docker run ... gemini-researcher init
CMD []
