A QR-like visual data encoder where the code is a noir comic panel. Encode text into a noir panel; decode a photographed/scanned panel back to text.
Live: https://noir-code.suncake.xyz
service/python # reference encode/decode core (NoiR Code lib + CLI) + FastAPI imaging sidecar
service/backend # Go (Fiber, ports & adapters) gateway API — public, stateless
service/frontend # React + Vite SPA (encode / decode in the browser)
deploy # docker-compose for the full stack
docs # design notes
browser ──/api/v1──▶ Go gateway (Fiber) ──HTTP──▶ Python imaging sidecar (FastAPI)
└─▶ noircode core (OpenCV + Reed–Solomon)
The gateway owns the public contract, validation, CORS and rate-limiting. The actual encode/decode work lives in the Python reference implementation (it owns the OpenCV pipeline), so the gateway proxies to the sidecar rather than re-porting the CV code. No database/broker — encode/decode is stateless.
cd deploy
docker compose up --build
# open http://localhost:8080 (SPA; proxies /api to the gateway)
# Swagger UI: http://localhost:8080/api/v1/docs# 1. imaging sidecar (port 8001)
cd service/python && uv sync --extra api && NOIRCODE_API_PORT=8001 uv run noir-api &
# 2. Go gateway (port 8000) — points at the sidecar
cd service/backend
NOIRCODE_API_FIBER_PORT=8000 \
NOIRCODE_API_IMAGING_BASE_URL=http://localhost:8001 \
NOIRCODE_API_FIBER_SWAGGER_FILE_PATH=docs/swagger.json \
go run ./cmd/main &
# 3. SPA (port 5173) — vite proxies /api → :8000
cd service/frontend && npm install && npm run dev| Method | Path | Body | Returns |
|---|---|---|---|
| POST | /api/v1/encode |
JSON {text, style?, hatch_data?, adaptive?} |
image/png |
| POST | /api/v1/decode |
multipart image file |
JSON result |
Capacity: up to 173 bytes of text (adaptive sizing shrinks the panel for shorter
payloads). OpenAPI 3.1 spec generated by swaggo v2 (make gen-api-docs).
cd service/python
uv run noir encode "https://example.com" --out panel.png --style # adaptive by default
uv run noir decode panel.pngSee service/python/README.md for the format internals and capacity/robustness curve.