gmeow 1.0.1

Grounded agent memory in five minutes: store, recall, and revise claims with confidence, standpoint, and provenance — persisted as a verifiable GTS package.

gmeow — grounded agent memory in five minutes

Store, recall, and revise what your agent learns — with confidence, standpoint, and provenance attached to every claim, persisted in a single verifiable file. No Docker, no database, no RDF knowledge required.

pip install gmeow

from gmeow import Memory

mem = Memory("assistant.gts")                       # a GTS ai-package on disk

claim = mem.store(
    "Patrick prefers explicit error handling over exceptions-as-flow",
    source="conversation 2026-06-10",
    confidence=0.8,
    according_to="claude-fable-5",                  # the asserting standpoint
)

mem.recall("error handling preferences", min_confidence=0.5)

mem.revise(
    claim,
    reason="user stated the opposite for scripts",
    superseded_by=mem.store(
        "For one-off scripts Patrick is fine with exceptions-as-flow",
        confidence=0.9,
        according_to="claude-fable-5",
    ),
)

What makes it different

• Claims, not strings. Every memory is a reified RDF 1.2 statement under the hood — confidence, asserting standpoint, source, and timestamp are first-class, so contradicting claims from different standpoints coexist instead of overwriting each other.
• Revision is supersession, never deletion. revise hides a claim and records what replaced it; the full audit trail stays in the file and is recoverable (recall(..., include_suppressed=True)).
• The file is the proof. An assistant.gts is a GTS package: append-only, hash-chained, damage-detecting. Every write is a crash-safe byte-append; mem.verify() (or the gts CLI) attests integrity.
• Graph when you want it. pip install 'gmeow[rdf]' and mem.to_rdflib() gives an RDF 1.1 projection of the dataset (rdflib has no RDF 1.2 yet, so the quoted-triple binding lines are dropped — the loss is declared, and the GTS file itself stays full-fidelity); the GMEOW ontology gives the vocabulary meaning across tools.

License

Apache-2.0. © Blackcat Informatics® Inc.