causalontology-py

The second implementation of the Causalontology standard — the proof of language independence. (The first is the PrologAI reference implementation.)

Zero dependencies — Python standard library only:

Conformance

$ python3 tests/run_conformance.py
...
107/107 vectors passed
causalontology-py is CONFORMANT to the suite (vectors frozen at specification 2.0.0).

The vectors are frozen at specification 2.0.0 (2026-07-13): they carry concrete identifiers, real keys, and a real verifying signature. The harness's old normalization now simply passes frozen values through.

Thirty-second taste

from causalontology import identify, InMemoryStore, keypair_from_seed, sign_record
import hashlib

store = InMemoryStore()
press = store.put({"type": "occurrent", "label": "press_button", "category": "action"})
light = store.put({"type": "occurrent", "label": "light_on", "category": "state_change"})
claim = store.put({"type": "causal_relation_object", "causes": [press], "effects": [light]})

print(store.gaps("missing_field"))   # the degenerate claim is a visible invitation

sk, source = keypair_from_seed(hashlib.sha256(b"alice").digest())
store.put_record(sign_record({"type": "assertion", "about": claim,
                              "source": source, "evidence_type": "imported",
                              "confidence": 0.5,
                              "timestamp": "2026-07-13T00:00:00Z"}, sk))

License: "The attribution always; no profit, no problem license." — see the repository LICENSE and NOTICE.