🚀 LangChain + Google AI Multilingual Translator

📘 Overview

This is a lightweight FastAPI + Jinja app that translates user text into a selected target language using Google's Gemini via LangChain. It serves a simple single-page UI at / and a JSON API at POST /translate. Secrets are managed with a .env file containing GOOGLE_API_KEY.

📑 Table of Contents

• Key Features
• Architecture
• Requirements
• Setup
• API Documentation
• Frontend
• Security
• Troubleshooting
• Useful Commands
• Project Structure

✨ Key Features

• Clean, single-page UI built with Jinja templates + vanilla JS
• FastAPI backend using LangChain ChatGoogleGenerativeAI
• One-click translate: enter text, pick a target language, get instant output
• Simple JSON API at POST /translate
• .env-based configuration (no keys in code)

🧱 Architecture

• Backend: main.py
  • Serves HTML at GET / via templates/index.html
  • JSON translation API: POST /translate
  • LangChain pipeline: ChatPromptTemplate → ChatGoogleGenerativeAI
• Templates & Static
  • Templates: templates/index.html
  • Static assets: static/style.css, static/script.js
• Optional utility: langchain_module.py
  • Standalone helper using google.generativeai for multi-language mapping
• Schemas: schemas.py (Pydantic models that can be reused)

🧩 Requirements

• Python 3.10+
• A Google API key with Gemini access in GOOGLE_API_KEY
• pip >= 21
• Windows, macOS, or Linux

🛠️ Setup

1. Create a virtual environment

   • Windows (PowerShell)

     python -m venv venv
     .\venv\Scripts\activate

   • macOS/Linux (bash/zsh)

     python3 -m venv venv
     source venv/bin/activate

2. Install dependencies

   pip install -r requirements.txt

3. Configure environment Create a .env file in the project root with:

   GOOGLE_API_KEY=your_gemini_api_key

4. Run the app

   uvicorn main:app --host 0.0.0.0 --port 8000 --reload

   • 🔗 App URL: http://localhost:8000
   • 📚 Swagger docs: http://localhost:8000/docs

📡 API Documentation

• Endpoints

  • GET / → renders templates/index.html
  • POST /translate → returns translated text

• Request body schema (JSON)

  {
    "text": "Hello, world!",
    "target_language": "Spanish"
  }

• Response schema (JSON)

  {
    "translated_text": "Hola, mundo!"
  }

• cURL example

  curl -X POST \
    -H "Content-Type: application/json" \
    -d '{"text":"Good morning","target_language":"French"}' \
    http://localhost:8000/translate

  HTTPie

  http POST :8000/translate text="Good morning" target_language="French"

🧠 Model Chains

• In main.py: ChatPromptTemplate + ChatGoogleGenerativeAI (gemini-1.5-flash, temperature=0)
• Prompt enforces: "Only return the final translated text and nothing else."
• langchain_module.py showcases a direct google.generativeai usage for multi-language mapping.

🌐 Frontend

Rendered with server-side templates (Jinja2Templates) and enhanced with a tiny JS script (static/script.js). No Node/React required.

🔒 Security

• Never commit your .env with GOOGLE_API_KEY.
• Rate limiting and auth are out-of-scope for this demo; add per your deployment needs.

🧰 Troubleshooting

• Missing GOOGLE_API_KEY

  • Ensure .env exists and contains a valid key.

• Import/module errors

  • Recreate venv and reinstall requirements.

• API returns 500

  • Check console logs; ensure the key has access to Gemini APIs.
  • Double-check that your requirements.txt is installed in the active venv.

📎 Useful Commands

• Run server (dev):

  uvicorn main:app --reload

• Freeze deps:

  pip freeze > requirements.txt

📦 Project Structure

  main.py
  langchain_module.py
  schemas.py
  templates/
    index.html
  static/
    style.css
    script.js
  requirements.txt
  .env (created by you)

📝 Notes

• The app is synchronous at the API level; long inputs depend on upstream model latency.
• Consider adding caching, auth, and rate limiting for production.

🙏 Acknowledgements

• Built with FastAPI, LangChain, and Google Gemini.

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

👥 Author

Made with 💖 by Rishabh Dhawad