Deploying to Production

This guide covers deploying an a2akit agent with Redis, PostgreSQL, and multiple Uvicorn workers behind a reverse proxy.

Architecture

                    ┌──────────┐
    Clients ──────► │  nginx   │
                    └────┬─────┘
                         │
              ┌──────────┼──────────┐
              ▼          ▼          ▼
         ┌────────┐ ┌────────┐ ┌────────┐
         │ worker │ │ worker │ │ worker │  (Uvicorn)
         └───┬────┘ └───┬────┘ └───┬────┘
             │          │          │
        ┌────┴──────────┴──────────┴────┐
        │           Redis               │  Broker + EventBus + CancelRegistry
        └───────────────────────────────┘
        ┌───────────────────────────────┐
        │         PostgreSQL            │  Storage
        └───────────────────────────────┘

In development, a2akit uses in-memory backends for everything. In production, you need shared state across workers — Redis for coordination, PostgreSQL for durable storage.

Docker Compose

services:
  redis:
    image: redis:7-alpine
    ports: ["6379:6379"]
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 5s
      timeout: 3s
      retries: 3

  postgres:
    image: postgres:16-alpine
    environment:
      POSTGRES_USER: a2akit
      POSTGRES_PASSWORD: a2akit
      POSTGRES_DB: a2akit
    ports: ["5432:5432"]
    volumes:
      - pgdata:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U a2akit"]
      interval: 5s
      timeout: 3s
      retries: 3

  agent:
    build: .
    ports: ["8000:8000"]
    environment:
      REDIS_URL: redis://redis:6379/0
      DATABASE_URL: postgresql+asyncpg://a2akit:a2akit@postgres:5432/a2akit
      UVICORN_WORKERS: "4"
    depends_on:
      redis: { condition: service_healthy }
      postgres: { condition: service_healthy }

volumes:
  pgdata:

Dockerfile

FROM python:3.12-slim

WORKDIR /app
COPY pyproject.toml .
RUN pip install --no-cache-dir .[redis,postgres]

COPY . .

CMD ["uvicorn", "my_agent:app", \
     "--host", "0.0.0.0", \
     "--port", "8000", \
     "--workers", "4"]

Agent Configuration

from a2akit import A2AServer, AgentCardConfig, CapabilitiesConfig, Worker

server = A2AServer(
    worker=MyWorker(),
    agent_card=AgentCardConfig(
        name="My Agent",
        description="Production agent.",
        version="1.0.0",
        capabilities=CapabilitiesConfig(streaming=True),
    ),
    # Redis handles broker, event bus, and cancel registry
    broker="redis://redis:6379/0",
    event_bus="redis://redis:6379/0",
    # PostgreSQL for durable storage
    storage="postgresql+asyncpg://a2akit:a2akit@postgres:5432/a2akit",
)
app = server.as_fastapi_app()

Passing a Redis URL to broker and event_bus automatically creates RedisBroker, RedisEventBus, and RedisCancelRegistry with a shared connection pool.

Scaling

Scale horizontally by adding more workers:

docker compose up --build --scale agent=3

All workers share the same Redis and PostgreSQL instances, so tasks are distributed automatically via Redis Streams consumer groups.

Reverse Proxy (nginx)

upstream a2akit {
    server agent:8000;
}

server {
    listen 80;

    location / {
        proxy_pass http://a2akit;
        proxy_set_header Host $host;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }

    # SSE streaming — disable buffering
    location ~ ^/v1/message:stream|/v1/tasks/.+:subscribe {
        proxy_pass http://a2akit;
        proxy_set_header Host $host;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
        proxy_buffering off;
        proxy_cache off;
        proxy_read_timeout 3600s;
    }
}

SSE Buffering

Reverse proxies buffer responses by default, which breaks SSE streaming. Always set proxy_buffering off for stream endpoints.

Environment Variables

All A2AServer constructor arguments can be set via environment variables:

Variable	Description	Example
REDIS_URL	Redis connection string	redis://redis:6379/0
DATABASE_URL	PostgreSQL connection string	postgresql+asyncpg://...
A2AKIT_MAX_CONCURRENT_TASKS	Max parallel task processing	10
A2AKIT_DEFAULT_BLOCKING_TIMEOUT	Blocking request timeout (seconds)	30
A2AKIT_PUSH_ALLOW_HTTP	Allow HTTP webhooks (dev only!)	false

See the Configuration page for the complete list.

Production Checklist

• [ ] Use Redis, PostgreSQL, or SQLite storage — not InMemoryStorage
• [ ] Use Redis for broker, event bus, and cancel registry
• [ ] Run multiple Uvicorn workers (--workers N)
• [ ] Put a reverse proxy in front (nginx, Caddy, or cloud LB)
• [ ] Disable SSE buffering in your proxy
• [ ] Set A2AKIT_PUSH_ALLOW_HTTP=false (default)
• [ ] Set debug=False (default) — the debug UI is not for production
• [ ] Configure max_concurrent_tasks to match your resource budget
• [ ] Set up health checks on /v1/health
• [ ] Add rate limiting at the gateway level (nginx limit_req, API Gateway, etc.)