Ask questions about GitHub repositories using Langchain for Node.js. This is a database free example of using Langchain to work with the OpenAI GPT API.

Create a .env file with your credentials or set them up as environment variables:

OPENAI_API_KEY=
GITHUB_TOKEN=

You can run the code in this repo via the CLI by leveraging the -cli suffixed files.

To generate the embeddings, run npx ts-node generate-cli.ts "https://github.com/Vheissu/cortex-device-list" which will populate local vector embeddings. Make sure you change to the repository you want to create embeddings for.

After you've created embeddings, you can run npx ts-node search-cli.ts "My question here" to ask your question about the repository.

There is also a simple front-end application in the frontend which will allow you to use a simple interface to query your repo. Please ensure you run the local server first before you attempt to ask a question: npx ts-node server.ts which will run on port 3000.

Also, don't forget to generate your embeddings first. The server will look for a vectors folder which will contain the generate vectors to do cosine similarity search.

To run the front-end app, go into the frontend directory and npm start (assuming you also ran npm install too) and the browser will open the app.