onedata-lambda-sdk

The Python SDK for writing Onedata Automation lambdas — the containerized operations that run as job executors in automation workflows. You write one handler function; the SDK turns the raw OpenFaaS request into typed jobs, runs your handler, streams its results, sends the mandatory first heartbeat and keeps the batch alive while it streams, and shapes the response the provider expects.

from onedata_lambda_sdk import Job, JobContext, per_job

@per_job(max_workers=10)
def handle(job: Job[dict], ctx: JobContext[dict]) -> dict:
    return {"result": process(job.args)}

Documentation

The full authoring documentation lives in docs/:

• SDK overview — the three-layer model (base image → SDK → handler) and how a job batch flows.
• Guides — writing a handler, a single-lambda repo, several lambdas in a uv workspace, developing against a vendored SDK wheel, testing, streaming logs and stats, and file access.
• Runtime internals — for SDK maintainers.

The platform-side contract this SDK implements — how Oneprovider invokes a lambda, the I/O and relay methods, heartbeats — is in the internal developer documentation under docs/automation/internals/lambda/ in onedata-dev-documentation.

Installation

Requires Python 3.12 or newer.

pip install onedata-lambda-sdk

Development

make sync     # create the .venv from the lockfile
make check    # lint (ruff format-check + ruff check + mypy) + tests
make test     # pytest with coverage

Run make help for the full target list.

Compatibility

onedata-lambda-sdk targets the v3 lambda model (run / @per_job / Job / JobContext) and requires Oneprovider 21.02.5 or newer. A v3 lambda builds on the onedata/lambda-base-slim:v3 base image — see Lambda generations.