CONTRIBUTING.md

How to contribute to Infini-gram API

Setting up your environment

Prerequisites

Make sure that you have the latest version of Docker 🐳 installed on your local machine.

Installing Dependencies

Install uv

This repo uses uv for Python dependency management. Follow the instructions on https://docs.astral.sh/uv/getting-started/installation to install it.

Run uv sync --all-packages to get the packages for every project in this repo.

Adding an index for local development

1. Ensure you have the aws cli installed. run brew install awscli if you don't.
2. Download the v4_pileval_llama index by running ./bin/download-infini-gram-array.sh

The infinigram-array folder is mounted to the Docker container for the API through the docker-compose.

Linting and Formatting

We use Ruff and mypy to lint, format, and check for type issues.

CLI

To check for Ruff issues, run uv run ruff check. If you want to have it automatically fix issues, run uv run ruff check --fix. If you want to have it format your code, run uv run ruff format.

To check for mypy issues, run uv run mypy --config ./pyproject.toml

VSCode

Install the ruff and mypy extensions. These are listed in the "Recommended Extensions" for the workspace as well.

Running the server

Docker Compose

The easiest way to run the full setup is to use the Docker Compose file. This will start the API, worker, proxy, and queue with all the appropriate connections.

To start with Docker Compose, run docker compose up in the root of this repo.

Outside of Docker

If you want to run applications outside of docker, you'll need to set up a queue yourself. The easiest method is to still go through Compose by just starting the queue. docker compose up queue

After that, make sure your environment variables are set correctly through a .env file or just environment variables, then run the services. API: uv run api/app.py Worker: uv run saq attribution_worker.worker.settings