run-backend

name	run-backend
description	Start the DARE backend locally — Django ASGI on port 8000, plus the Redis and .env prerequisites it needs. Use when the user wants to run, start, boot, or serve the DARE backend, or asks why the API on :8000 is not responding.

Run the DARE backend

The backend is a Django ASGI service (Django + Channels + python-socketio). It serves REST on http://localhost:8000/ and Socket.IO on the same port.

Two paths — Docker starts everything, local Python runs just Django on your machine. Pick Docker unless the user is actively editing backend code and wants a fast reload loop.

Prerequisites

• Python 3.13.x — the dependency set requires it; 3.12 and 3.14 will fail to install.
• Redis on :6379 — required for Socket.IO pub/sub and the RQ job queue. The backend starts without it but chat streaming will not work.
• .env — copy from the checked-in example. Note the filename is .example.env, not .env.example.

cp .example.env .env

At minimum set DJANGO_SECRET_KEY and one provider key (OPENAI_API_KEY, CLAUDE_API_KEY, or GEMINI_API_KEY). Every variable is documented in docs/configuration.md.

Path 1 — Docker (brings up the whole stack)

docker compose up --build -d
docker compose exec web python manage.py createsuperuser

This starts the API, an RQ worker, Postgres + pgvector, Redis, and Weaviate. The web container waits for Postgres and Redis, then runs migrations before starting Uvicorn — you do not need to migrate by hand.

Path 2 — local Python

python3.13 -m venv .venv
source .venv/bin/activate
pip install -r requirements/local.txt
python manage.py migrate
uvicorn dare.asgi:application --host 0.0.0.0 --port 8000 --reload

Use uvicorn dare.asgi:application, not python manage.py runserver. The backend serves WebSocket consumers through Channels and Socket.IO; runserver is WSGI-only and chat streaming will silently fail to connect.

Redis must be running separately on this path (redis-server), and background jobs need a worker — see the run-workers skill.

Verify it came up

curl http://localhost:8000/api/health/
curl http://localhost:8000/api/ready/

/api/health/ answers as soon as Django is serving. /api/ready/ additionally checks the dependencies, so use it to tell "Django is up" apart from "Django is up and Postgres/Redis are reachable".

The OpenAPI schema is at http://localhost:8000/api/schema/. Swagger UI at /api/docs/ loads its assets from a CDN, so on a restricted or offline network read the raw schema instead.

Running it in the background

When starting the server from an agent session, detach it so it outlives the session:

nohup uvicorn dare.asgi:application --host 0.0.0.0 --port 8000 --reload \
  > /tmp/dare-backend.log 2>&1 & disown

Then confirm and read the log:

lsof -ti tcp:8000        # should print a PID
tail -n 40 /tmp/dare-backend.log

Stop it

lsof -ti tcp:8000 | xargs kill     # local Python
docker compose down                # Docker

If port 8000 is already taken, a previous run is still alive — check with lsof -ti tcp:8000 before starting a second one.

Related

• run-workers — background jobs; needed before uploaded files will process.
• The frontend lives in a separate repo, dare-frontend, and defaults to http://localhost:5173.
• README.md for the project overview, INSTALL.md for full deployment guidance.