MRui

A web UI for running MRtwo reconstructions. Upload data, pick an algorithm, get results. Uses Huey for background jobs so you can queue up work and come back later.

Quick start

# Everything (API + worker + frontend)
uv run dev

# Just API + worker (no frontend dev server)
uv run prod

Then open http://localhost:5173 (dev) or http://localhost:8000 (prod).

New job	Result

How it works

You POST a file to /api/jobs with algorithm parameters. The backend:

1. Saves your input file
2. Creates a job record (JSON on disk)
3. Enqueues it with Huey
4. Returns a job ID

The worker picks it up, runs the reconstruction, writes an H5 file, and updates the job status. The frontend polls job status and displays results.

Key files:

• mrui/main.py - HTTP API
• mrui/jobs.py - Worker task that runs reconstructions
• mrui/algorithms/ - Algorithm definitions and registry
• mrui/storage.py - Reading/writing job metadata

Adding algorithms

Add your reconstruction to mrui/algorithms/:

1. Pick an ID in mrui/algorithms/base.py (AlgorithmId enum)
2. Create mrui/algorithms/yours.py:

   class YourParams(AlgorithmParamsBase):
       algorithm: Literal[AlgorithmId.YOURS] = AlgorithmId.YOURS
       your_param: float = 0.5
   
   class YourAlgorithm(ReconstructionAlgorithm[YourParams]):
       id = AlgorithmId.YOURS
       name = "Your Reconstruction"
       description = "Does the thing"
       params_model = YourParams
   
       def run(self, task, kdata, params):
           # Reconstruct here, return IData
           ...

3. Register it in mrui/algorithms/__init__.py:

   ALGORITHMS = (
       ...
       YourAlgorithm(),
   )

4. Add frontend UI in frontend/src/components/algorithms/YoursForm.tsx and register in registry.ts
5. Run npm run typegen in frontend/ to update types from the backend schema

Configuration

All env vars start with MRUI_:

Variable	Default	What it does
MRUI_HOST	0.0.0.0	API bind address
MRUI_PORT	8000	API port
MRUI_INPUTS_DIR	/tmp/mrui/inputs	Where uploaded files go
MRUI_RESULTS_DIR	/tmp/mrui/results	Where results and job metadata go
MRUI_QUEUE_DB_PATH	/tmp/mrui/queue/huey.db	Huey SQLite database
MRUI_API_WORKERS	1 (dev) / 2 (prod)	Uvicorn workers

Development

Backend:

uv run ruff check .      # Lint
uv run pyright mrui      # Type check
uv run pytest            # Run tests

Frontend:

cd frontend
npm run typegen          # Regenerate API types from backend
npm run lint             # ESLint
npm run build            # Production build