DEV Community

Cover image for From Prompt to Production: Dockerizing a LangChain Agent with FastAPI
Chandrani Mukherjee
Chandrani Mukherjee

Posted on

From Prompt to Production: Dockerizing a LangChain Agent with FastAPI

#docker #langchain #python #fastapi

Supercharge your AI app deployment with Docker, FastAPI, and LangChain in one seamless containerized pipeline.

๐Ÿง  Overview

As AI-powered apps become more complex, managing dependencies, serving endpoints, and ensuring smooth deployment are top priorities. In this post, you'll learn how to dockerize a LangChain agent wrapped with FastAPI โ€” giving you a ready-to-deploy, production-friendly container for your intelligent applications.

By the end, youโ€™ll:

  • Create a LangChain agent
  • Wrap it with FastAPI for a clean REST interface
  • Dockerize the entire setup
  • Run it anywhere with just one command

๐Ÿ“ฆ Prerequisites

Before you begin, make sure you have:

  • Docker installed
  • Python 3.10+ (for local testing)
  • An OpenAI API Key or any LLM key supported by LangChain

๐Ÿ“ Project Structure

langchain-agent-api/
โ”œโ”€โ”€ agent_app/
โ”‚ โ”œโ”€โ”€ main.py
โ”‚ โ””โ”€โ”€ agent.py
โ”œโ”€โ”€ requirements.txt
โ”œโ”€โ”€ Dockerfile
โ””โ”€โ”€ .env

pgsql

โœจ Step 1: Create the LangChain Agent

agent_app/agent.py

from langchain.agents import initialize_agent, Tool
from langchain.llms import OpenAI
from langchain.utilities import SerpAPIWrapper
import os

def create_agent():
    llm = OpenAI(temperature=0, openai_api_key=os.getenv("OPENAI_API_KEY"))
    search = SerpAPIWrapper()
    tools = [Tool(name="Search", func=search.run, description="Useful for answering general questions.")]
    agent = initialize_agent(tools, llm, agent="zero-shot-react-description", verbose=True)
    return agent
Enter fullscreen mode Exit fullscreen mode

๐Ÿš€ Step 2: Wrap with FastAPI
agent_app/main.py

from fastapi import FastAPI
from pydantic import BaseModel
from agent import create_agent

app = FastAPI()
agent = create_agent()

class Query(BaseModel):
    question: str

@app.post("/ask")
async def ask_question(query: Query):
    response = agent.run(query.question)
    return {"response": response}
Enter fullscreen mode Exit fullscreen mode

๐Ÿ“„ Step 3: Define Requirements
requirements.txt

fastapi
uvicorn
langchain
openai
python-dotenv
Enter fullscreen mode Exit fullscreen mode

๐Ÿ’ก Add serpapi or other tools as needed.

๐Ÿ› ๏ธ Step 4: Dockerfile
Dockerfile

FROM python:3.10-slim

WORKDIR /app

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY agent_app ./agent_app
COPY .env .

CMD ["uvicorn", "agent_app.main:app", "--host", "0.0.0.0", "--port", "8000"]
Enter fullscreen mode Exit fullscreen mode

๐Ÿ”‘ Step 5: Add Environment Variables
.env

OPENAI_API_KEY=your_openai_key_here
SERPAPI_API_KEY=your_serpapi_key_here

โš ๏ธ Never commit .env to public repos. Use Docker secrets or CI/CD env vars in production.

๐Ÿงช Step 6: Build and Run
๐Ÿงฑ Build the Docker image

docker build -t langchain-agent-api .
Enter fullscreen mode Exit fullscreen mode

๐Ÿš€ Run the container

docker run --env-file .env -p 8000:8000 langchain-agent-api
Enter fullscreen mode Exit fullscreen mode

๐Ÿ“ฌ Try It Out
Once running, test your agent with:

curl -X POST http://localhost:8000/ask \
-H "Content-Type: application/json" \
-d '{"question": "Who is the CEO of OpenAI?"}'
Enter fullscreen mode Exit fullscreen mode

Your containerized LangChain agent should reply in seconds! ๐Ÿค–

๐Ÿ“ฆ Bonus: Add Docker Compose (Optional)
docker-compose.yml


version: "3.8"
services:
  langchain:
    build: .
    ports:
      - "8000:8000"
    env_file:
      - .env
Enter fullscreen mode Exit fullscreen mode

Then run:


docker-compose up --build
Enter fullscreen mode Exit fullscreen mode

๐Ÿ Final Thoughts
You now have a production-ready, containerized LangChain agent served via FastAPI. Whether youโ€™re building internal AI tools or deploying to the cloud, this setup gives you repeatability, portability, and power.

Top comments (2)

Collapse
 
dmitriiweb profile image
di •

Great article! I have just one note: the run method in agent.run was deprecated in LangChain versionย 0.1.0 and replaced with invoke. Also, both run and invoke are blocking methods, so since youโ€™re using the async interface, youโ€™ll need to call the agent with ainvoke

Collapse
 
moni121189 profile image
Chandrani Mukherjee •

Hey Thanks for the appreciation and input. Indeed you are correct. Will surely make articles using asynchronous invoke