CONTRIBUTING.md

Guidelines for Contributions

Local setup

Dev requirements

We recommend using uv:

1. make
2. source .venv/bin/activate

Lint

Ensure all checks pass:

pre-commit run --all-files --verbose

Installation

make install

Tests

Ensure all tests pass: pytest -v

Local Build for QA and manual testing

1. Use litellm_docker_compose.yaml to start LiteLLM and Postgres locally:

docker compose -f litellm_docker_compose.yaml up -d

or if you are using legacy docker-compose:

docker-compose -f litellm_docker_compose.yaml up -d

2. Create a second database that is needed for authentication, and apply migrations

bash scripts/create-app-attest-database.sh

uv run alembic upgrade head

LiteLLM will be accessible at localhost:4000 and localhost:4000/ui.

3. Run MLPA with

mlpa

4. Stop the service with

docker compose -f litellm_docker_compose.yaml down

Useful CURLs for QA

1. MLPA liveness:

curl --location 'http://0.0.0.0:8080/health/liveness' \
--header 'Content-Type: application/json'

1. MLPA readiness:

curl --location 'http://0.0.0.0:8080/health/readiness' \
--header 'Content-Type: application/json'

1. MLPA completion:

curl --location 'http://0.0.0.0:8080/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'authorization: Bearer {YOUR_MOZILLA_FXA_TOKEN}' \
--header 'X-LiteLLM-Key: Bearer {MASTER_KEY}' \
--data '{
 "model": "openai/gpt-4o",
  "messages": [{
     "role": "user",
     "content": "Hello!"
   }]
}'

1. LiteLLM liveness:

curl --location 'http://localhost:4000/health/liveness' \
--header 'Content-Type: application/json'

1. List of available models:

curl --location 'http://localhost:4000/models' \
--header 'Content-Type: application/json' \
--header 'X-LiteLLM-Key: Bearer {MASTER_KEY}' \
--data ''

1. Completion directly from LiteLLM:

curl --location 'http://localhost:4000/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'X-LiteLLM-Key: Bearer {MASTER_KEY}' \
--data '{
    "model": "openai/gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": "what is 2+2?"
      }
    ]
}'

FXA tokens and where to find them

MLPA uses the https://github.com/mozilla/PyFxA library for authentication with a Mozilla account. Please follow the quick-start instructions in their README.

Here is a quick snippet:

from fxa.tools.bearer import get_bearer_token

fxa_token: str = get_bearer_token(
    your_mozilla_account_email,
    your_mozilla_account_password,
    scopes=["profile:uid"],
    client_id="5882386c6d801776" # a common client_id for the dev environment,
    account_server_url="https://api.accounts.firefox.com",
    oauth_server_url="https://oauth.accounts.firefox.com",
)

How to update static docs/index.html from redoc

Ensure Node is installed.

make docs

Alternatively, with a running server:

1. make install
2. mlpa
3. curl http://localhost:8080/openapi.json -o /tmp/openapi.json
4. npx @redocly/cli build-docs /tmp/openapi.json -o docs/index.html