v0.9.0: Production-Ready Fault-Tolerant Workflow Engine

What's New in v0.9.0

First public release of Orchestration Engine — a fault-tolerant, dependency-aware workflow orchestration engine for Python with zero runtime dependencies.

Key Highlights:

• Zero Runtime Dependencies: Pure Python standard library implementation (sqlite3, multiprocessing, concurrent.futures, json, argparse).
• Single-Writer SQLite Coordinator: 44,730 writes/sec with WAL mode; completely eliminates database locking errors.
• Concurrency Backends: Multi-worker Thread and isolated spawned Process executors with true SIGKILL timeouts.
• Durable Crash Resumption: Seamlessly resume interrupted workflows with resume_run() without duplicate executions.
• Data Flow & Output Passing: TaskResult payloads pass downstream to task.inputs with relational persistence in SQLite schema v2.
• Built-in CLI: orchestration-engine command-line interface with subcommands (validate, run, resume, status, history, runs), exit code contract (0-3), and machine-readable JSON mode.
• Dual Licensing: Dual-licensed under either MIT or Apache 2.0 at your choice.
• Rich Examples Included: Multi-step AI agent pipelines with rate-limit retries, RAG document indexing with quality gates, and autonomous SRE incident triage.

Installation

Install from PyPI:

pip install orchestration-engine

For optional YAML workflow loading support (load_workflow_from_yaml):

pip install "orchestration-engine[yaml]"

Verification & Quality

• 162 Automated Tests (100% passing across 20 test modules)
• 91.45% Line Coverage
• 0 Flakes across stress testing
• Mypy Strict & Ruff Formatted