Ledgence 0.1.0
Ledgence is an MIT-licensed, self-hosted task and workflow orchestration platform built in Rust, with a Python client and Python program runtime.
Included
- Publish immutable Python program packages with prepared application dependencies. Workers download, verify and cache them on demand.
- Submit and inspect tasks through the HTTP API, CLI or
from ledgence.client import AsyncClient. Scoped idempotency keys, attempt identities and durable results support recovery. - Run programs in a bounded pool of reusable subprocesses with one concurrency setting per worker. Deadline and cancellation checks reject late runtime completions before session reuse.
- Run checkpoint workflows with concurrent durable local steps, distributed tasks, owned subworkflows, timers and external events. Workflows durably suspended on distributed work, timers or events release their invocation slot and resume from persisted continuation state.
- Register durable completion callbacks with persisted delivery retries and status.
- Use integrated queue acquisition or optional SQS-compatible delivery, validated locally with ElasticMQ. PostgreSQL remains the durable execution authority.
- Correlate CloudEvents, business identifiers, traces and Python logs; export optional bounded-cardinality OpenTelemetry metrics.
- Maintain expired execution history with scoped, bounded retention and a minimum 90-day terminal window.
- Start a complete local Compose deployment and run the installed-SDK task/workflow/callback example.
Operating scope
Ledgence 0.1.0 runs operator-trusted programs. Programs supply their application dependencies; workers supply a compatible host CPython interpreter. Program packages declare their exact Python major/minor, OS and architecture.
Execution effects are at-least-once: application side effects must be idempotent. A client result timeout stops observation without automatically rerunning the task. Workflows use explicit checkpoints and continuations. Public APIs remain pre-1.0 and may evolve.
Local ElasticMQ qualification does not establish real AWS acceptance. Short load and soak experiments are workload-specific evidence, not sustained daily capacity or billion-scale guarantees. Priorities, cost-based placement, additional broker/stream adapters and sharded orchestration remain future work.
Licensing
Ledgence-owned code is MIT licensed. Applications and programs may remain proprietary. Retain the accompanying third-party legal notices when redistributing artifacts. Host operating-system/runtime components retain their own licenses.
Release evidence
Qualified build source: bcb4af38dfcbbc3f2b164a36cc42ca78de58cc2c. The release archive retains the exact qualified executable, SDK and helper bytes; its provenance records the source-equivalent release commit and candidate identity.
- All 18 CI jobs and both documentation jobs passed.
- Native macOS arm64 and Docker Linux arm64 workspace tests passed: 567 ordinary tests per platform. Dedicated local gates cover all 174 infrastructure tests (173 PostgreSQL and one ElasticMQ), including database crash/restart.
- The packaged candidate passed 50 functional and 38 workflow scenario executions, with both fresh and preserved-volume Compose deployments.
- Six bounded arrival experiments completed 41,670 measured tasks. Two separate 15-minute mixed-workflow soaks completed 27,989 root operations and delivered 3,999 callbacks. Final durable work and worker occupancy drained to zero. These local measurements do not establish sustained daily capacity.
The downloadable native bundle targets macOS arm64 and includes the Python client wheel/sdist, worker helper, documentation and legal material. Supply compatible host CPython 3.11–3.14 and external PostgreSQL/broker services. Linux arm64 Compose/source validation is separate from this native download; this release does not publish a container image or a PyPI package. Public API compatibility remains pre-1.0.