how to use

Once running, see http://localhost:30001/docs for API documentation.

run natively

Run poetry install first if you have not already done so.

The default settings in server.utils.Settings should be ok for native use. Override any you want in a server/.env file or by setting them as environment variables in the shells where the app and workers are running.

start kafka and redis

cd ..
docker-compose -f docker-compose.native.yml up

start app

poetry run uvicorn app:app --host 0.0.0.0 --port 30001

start celery workers

cd tasks/
poetry run celery -A stream_inference worker --loglevel=info

run in Docker

Edit the env variables in docker-compose.yml if needed - the defaults should be ok. There should be differences from the defaults in server.utils.Settings for native use.

# from repo root
docker-compose up

run client (unused, for dev convenience only)

cd ui/
yarn start

hosted instances

settings are configured here: https://github.com/HumanSignal/infra/blob/master/base/charts/humansignal/adala/values.yaml

The only significant difference from local as of now is that kafka topic autocreation is enabled as a backup.

development

testing

Server pytests in test/test_server.py rely on a running server (kafka, redis, and celery, but not app) behind the use_server mark, and a real openai/azure key available in the OPENAI_API_KEY/AZURE_API_KEY env var behind the use_openai/use_azure mark. These tests are not run by default. To run them, use:

poetry run pytest -m use_server -m use_openai -m use_azure

rebuild on code change

# from repo root
docker-compose up --build
# if dependencies have changed, or to get to a clean state:
docker-compose build --no-cache
docker-compose up

edit client on API change

./api_codegen.sh
export SERVER=$(pwd)
cd ui/
yarn codegen

now edit src/adala.tsx manually given the autogenerated changes in src/_api/

change the log level

If not set, the logging level will default to "INFO".

When doing native dev, you can set the LOG_LEVEL env var to "INFO", "DEBUG", etc. For docker, change the value of the LOG_LEVEL env var in docker-compose.yml (for both app and worker).