Results Bridge

Live service: results-bridge.onrender.com/docs — interactive API docs; try /ingest with a Qubit CSV. Runs in dry-run mode (no credentials): full parse → validate → audit pipeline, with the Benchling payload rendered instead of sent. Free-tier host, so the first request after idle takes ~30s to wake.

Instrument output in → validated against assay acceptance criteria → structured result in Benchling. Built, deployed, and operated as a small production service.

instrument file ──► POST /ingest ──► parser (robotic-assays) ──► validation ──► Benchling /results
                                          │                          │               (or dry-run)
                                          └──────────► append-only audit log ◄───────┘

Results Bridge is the operations layer of a three-repo ecosystem: robotic-assays is the hub (assay library, acceptance criteria, instrument parsers); this service consumes it to turn raw instrument files into validated, traceable ELN results.

What one request does

POST /ingest with a Qubit/TapeStation/Tecan Spark/CLARIOstar file and an assay_id:

1. Parse — the file is parsed by the robotic-assays ingestion module (format auto-detected).
2. Validate — every sample is judged against the assay's acceptance criteria from criteria.json; missing values surface as INVALID, never silently dropped.
3. Deliver — a Benchling-shaped results payload is POSTed to the tenant's API, or returned verbatim in dry-run mode when no credentials are configured, so anyone can run the service keyless.
4. Audit — an append-only SQLite row records the file hash, assay, verdict counts and delivery mode: every decision traceable to its source file.

Run it (keyless, 60 seconds)

git clone https://github.com/anupama29k/results-bridge.git
git clone https://github.com/anupama29k/robotic-assays.git   # side by side
cd results-bridge && pip install -r requirements.txt
PYTHONPATH=src uvicorn results_bridge.main:app --port 8080
curl -F "file=@tests/fixtures/test_qubit.csv" -F "assay_id=NGS_001_library_prep_input" localhost:8080/ingest

Live mode: set BENCHLING_API_URL and BENCHLING_API_KEY in the environment (never committed) and the same request writes real results with file-hash idempotency.

Layout

Path	What it is
src/results_bridge/main.py	FastAPI service: /ingest, /audit, /healthz
src/results_bridge/parsers.py	Adapter over the robotic-assays ingestion module
src/results_bridge/validation.py + criteria.json	Acceptance-criteria engine and the per-assay rules
src/results_bridge/benchling.py	Delivery layer: dry-run client + live client (auth, idempotency, retries)
src/results_bridge/audit.py	Append-only SQLite audit log
tests/	pytest suite against real instrument sample files; runs in CI on every push

Author

Designed and built by Anupama Kozhiyalam · github.com/anupama29k · MIT License