Creating a Deployment

1LangGraph Platform deployment overview

LangGraph Platform is provided by LangChain for applying LangGraphDeployed as a production-grade serviceplatform. It provides REST API interface, persistent storage, streaming output support, and conversation thread management capabilities.

This section is built with Module 5task_mAIstroTake the application as an example to show how to package and deploy a local map code as a service that can be accessed externally.

24 necessary documents explained in detail

To create a LangGraph Platform deployment, you need tomodule-6/deployment/Prepare the following 4 files in the directory:

langgraph.json configuration example

// langgraph.json — 声明图的路径和 Python 版本
{
  "graphs": {
    "task_maistro": "./task_maistro.py:graph"
  },
  "python_version": "3.11",
  "dependencies": ["."]
}

The format is"图名称": "文件路径:图变量名". After deployment, the graph name becomes part of the API endpoint path.

requirements.txt example

langgraph
langchain-deepseek
trustcall
langchain-core
python-dotenv

3LangGraph CLI: Building a Docker image

LangGraph CLI is a command line tool provided by LangChain for packaging graph code into a Docker image (i.e. LangGraph Server container).

Install CLI

pip install -U langgraph-cli

Build a Docker image

# 进入部署目录
cd module-6/deployment

# 构建 Docker 镜像，-t 指定镜像名称
langgraph build -t my-image

langgraph buildwill read the files in the current directorylanggraph.json, automatically:

1. Download the base image adapted to LangGraph Server
2. Installationrequirements.txtall dependencies in
3. Copy the image code into the image
4. Build the final runnable Docker image

4Redis vs. PostgreSQL: Infrastructure Configuration

LangGraph Server relies on two external services for infrastructure:

Run the LangGraph Server container independently (already have Redis/PostgreSQL)

# 如果 Redis 和 PostgreSQL 已在别处运行，直接启动 API 容器
docker run \
    --env-file .env \
    -p 8123:8000 \
    -e REDIS_URI="redis://localhost:6379" \
    -e DATABASE_URI="postgresql://user:pass@localhost/langgraph" \
    -e LANGSMITH_API_KEY="lsv2_..." \
    my-image

After startup, LangGraph Server listens to port 8000 in the container by default.-p 8123:8000Maps to host port 8123.

5docker-compose.yml: three-service orchestration

If you do not have independent Redis/PostgreSQL, you can usedocker-compose.ymlStart all three services with one click:

# docker-compose.yml 中定义的三个服务

services:
  langgraph-redis:
    image: redis:6          # 使用官方 Redis 镜像
    ports: ["6379:6379"]

  langgraph-postgres:
    image: postgres:16      # 使用官方 PostgreSQL 镜像
    environment:
      POSTGRES_DB: langgraph
      POSTGRES_USER: langgraph
      POSTGRES_PASSWORD: langgraph
    ports: ["5433:5432"]

  langgraph-api:
    image: ${IMAGE_NAME}    # 使用 langgraph build 构建的镜像
    ports: ["8123:8000"]
    depends_on:
      langgraph-redis:    { condition: service_healthy }
      langgraph-postgres: { condition: service_healthy }
    env_file: .env          # 从 .env 文件读取 API 密钥等环境变量

Startup steps

1. Copydocker-compose-example.ymlfordocker-compose.yml
2. exist.envFill in the fileIMAGE_NAME、LANGSMITH_API_KEY、DEEPSEEK_API_KEY
3. Run the following command to start all services:

cd module-6/deployment
docker compose up

6Self-hosted vs LangGraph Cloud comparison