version: '3.8'
services:
mcp-eval-server:
build:
context: .
dockerfile: Containerfile
container_name: mcp-eval-server
environment:
# OpenAI Configuration
- OPENAI_API_KEY=${OPENAI_API_KEY:-}
# Azure OpenAI Configuration
- AZURE_OPENAI_ENDPOINT=${AZURE_OPENAI_ENDPOINT:-}
- AZURE_OPENAI_KEY=${AZURE_OPENAI_KEY:-}
# Server Configuration
- DEFAULT_JUDGE_MODEL=${DEFAULT_JUDGE_MODEL:-gpt-4}
- MCP_EVAL_CACHE_DIR=/app/data/cache
- MCP_EVAL_RESULTS_DB=/app/data/results/evaluation_results.db
# Logging
- PYTHONUNBUFFERED=1
- LOG_LEVEL=${LOG_LEVEL:-INFO}
volumes:
# Persist evaluation data and cache
- eval-cache:/app/data/cache
- eval-results:/app/data/results
# Optional: Mount custom configurations
- ./config:/app/config:ro
# For testing HTTP wrapper (optional)
ports:
- "8080:8080"
# Resource limits for evaluation workloads
deploy:
resources:
limits:
memory: 2G
cpus: '1.0'
reservations:
memory: 512M
cpus: '0.25'
# Health check
healthcheck:
test: ["CMD", "python3", "-c", "from mcp_eval_server.server import judge_tools; print('Health check:', len(judge_tools.get_available_judges()), 'judges')"]
interval: 60s
timeout: 30s
retries: 3
start_period: 120s
# Restart policy
restart: unless-stopped
volumes:
eval-cache:
driver: local
eval-results:
driver: local
# Optional: Add Redis for distributed caching
# networks:
# eval-network:
# driver: bridge
