Welcome to Chainlit-OAI-Assistants Template, a robust starting point for deploying applications using Chainlit and OpenAI's Assistants API. This repository is a public version of our current production monorepo, deployed at https://AIFGE.xyz, and we're releasing it publicly to give back to the Chainlit community.
This repository is a demonstration of our recent work and serves as a launchpad for your projects. It's equipped with:
- OpenAI's Assistants API: We've found it to be an excellent tool for prototyping RAG applications. We currently handle file-retrieval tools along with annotations, but code interpreter and functions are not yet supported.
- Deployment to Google Cloud Run: We've integrated Terraform for infrastructure management, with state stored in a bucket. Cloudbuild manages our Docker image, which we deploy to staging and production instances.
- Google Auth: Our app uses Google Auth for user authentication. You'll need to set this up yourself.
- Discord webhooks: We've added Discord webhooks for user login events. Chainlit now supports Discord bots as well, so expect more from us in the future.
- Mailchimp integration: We've integrated a callback to Mailchimp to add users to our mailing list.
/infra: Holds the Terraform config files and batch scripts./literal: A simple wrapper to help us clean up our dev instances when we need to purge./openai-assistant: Our current implementation of the OpenAI Assistant API.
To get started with Chainlit-Template, clone this repository. Install the project using poetry, and then run chainlit run app.py from the openai-assistant directory.
Check the infra/readme to see how we setup our cloud run deployments.
We welcome contributions! Submit an issue and related PR and we'll take a look! You can find us in the Chainlit Discord if you need anything else.
To see what has changed in recent versions of COAT, see the CHANGELOG.md file.
COAT is licensed under the terms of the MIT license. See the LICENSE.md file for the full license text.
If you have any questions, issues, or just want to contribute, feel free to open an issue or a pull request. We're always happy to help and learn from you!