Zero-Daemon Process Durability
For Long-Horizon Python Workflows

pip install letitloop

from letitloop import durable, step

@durable(task_id="scientific-pipeline")
def run_pipeline(dataset_path: str):
    # Step 1: Cached to single-file WAL (LILWAL02) in 0.3s
    data = step("load_data", lambda: parse_dataset(dataset_path))

    # Step 2: If SIGKILL strikes here, Step 1 is fast-forwarded in 0.94ms
    model = step("train_model", lambda: train_weights(data))

    # Step 3: Two-phase atomic marker commits output with zero duplicate side effects
    results = step("evaluate", lambda: run_eval(model))
    return results

# Supervise a workflow with auto-restart on SIGKILL / OOM
$ lil watch workflow.py

# Run interactive zero-config 5-step durability demo
$ lil demo

# Inspect WAL frames, sequence numbers, and hash heads
$ lil inspect .bench_wal/my_task

# Run local DCP-2.0 conformance benchmarks
$ lil bench

# .github/workflows/verify-durability.yml
name: Verify Durability Gate
on: [pull_request]

jobs:
  verify:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: sdageltc/letitloop-action@v2
        with:
          wal-dir: '.bench_wal'
          mode: 'gate'