LangChainTemplate

A template for LangChain apps!

Features

• Docker containment
• Debugging in Visual Studio Code
• NVIDIA driver support
• Streamlit UI support

Quick Start

1. Install:
   • Docker
   • Docker Compose (if it's not included with your Docker distribution)
   • Visual Studio Code
   • The Python extension for Visual Studio Code
   • For NVIDIA driver support, set up the NVIDIA Container Toolkit for Docker.

2. Clone this repo:

   git clone https://whatever-the-url/of-this-repo-is.git

3. Open the repo folder in Visual Studio Code.

4. Create a file named .env for your API keys and other secrets (it should be ignored by GIT and Docker). To run the demo app, you should at least provide an OpenAI key. For example:

   OPENAI_API_KEY=my-openai-key

5. Configure options in .vscode/settings.json

6. Hit F5 to say hello to an AI!

7. Start developing your app in the Source folder.

Directory

• .vscode – configuration for VS Code
  • settings.json – Includes LangChainTemplate feature settings
• Data/Home – the container's home directory (mounted to /home/user as read/write)
• Data/Models – a place to put big files (mounted to /data/models as read-only)
• Source
  • packages – a place to organise custom packages for your app
  • main.py – your app's main entrypoint
  • requirements.txt – a PIP requirements file for your app's package dependencies
  • Home.py – the home page for streamlit apps
  • pages/*.py – additional pages for streamlit apps

Planned Features

• Streamlit CSS template
• Offline model installer
• GPU demo
• HTTPS
• Testing framework
• UI authentication (e.g. Streamlit-Authenticator)
• Update guide
• Improve feature selection (more features and less spaghetti!)