Skip to content

Quick Start

Prerequisites

  • Python ≥ 3.12 (tested with 3.12.3)
  • Node.js ≥ 22 (tested with 22.22.3)
  • npm ≥ 10 (tested with 10.9.8)
  • Docker (optional, for containerised deployment)

Docker Compose

The Echo-DSRN classifier is a gated HF model. Before building:

  1. Request access to the model repo
  2. Create an HF access token with read permissions
  3. Add it to .env: HF_TOKEN=hf_...
make rebuild      # builds Docker image with the gated model
docker compose up -d

App serves on http://localhost:7860. Redis, the web server, and a Celery worker start automatically.

Python

uv

uv run python backend/server.py

pip

pip install -r requirements.txt
python backend/server.py

Build the frontend first: cd frontend && npm ci && npm run build. App serves on http://localhost:7860.

Makefile

make help          # list all targets
make up            # docker compose up -d
make build-docker  # build the Docker image
make test          # run pytest
make lint          # pre-commit hooks
make migrate       # alembic upgrade head
make judge N=100   # run LLM-as-Judge (sync)
make judge-async N=100  # dispatch to Celery worker