133,920 governed workflows.
Every one plans through the real engine.
28 business archetypes × provider assignments across 15 governed integrations × 8 topologies × 6 triggers × 3 governance profiles. Deterministic, offline, tamper-evident. Same code → byte-identical output → one index_root hash.
The whole library hashes to one number.
Every workflow spec is deterministic. Regenerate on your machine — the resulting lib/index.jsonl is byte-identical to the committed one. If anything drifts, the index_root changes.
Eight workflows, one per topology.
Each is one of 133,920 real templates. The engine's governance policy set approval — it isn't a marketing label, it's what the policy engine returned when the spec was planned.
What the 133,920 covers.
88,128 require_human, 45,792 auto_approve. Split by the engine's floor: irreversible external effects or spend → human; otherwise policy-driven.
Not 133,920 unique business ideas — the point isn't uniqueness. It's the shape of the policy surface: every archetype × provider × topology × trigger × governance cell is enumerated and hashed, so drift anywhere is detectable.
Three commands. Fully offline.
Requires a local RailCall engine checkout at ~/railcall-engine (or set RAILCALL_ENGINE=). No network calls, no LLM.
git clone git@github.com:patl4588/railcall-workflow-library.git cd railcall-workflow-library python3 generate.py # regenerates index byte-for-byte python3 verify.py # 10 workers: plans all + signs cross-section shasum -a 256 lib/index.jsonl # → 33f997ce3436ffbf2183f7a1ed4d0eea45d2db2d9140cc4648eea8fd66a059a5 lib/index.jsonl