haiku.rag/tests/capabilities
Yiorgis Gozadinos e1e7936d15
Let a model declare that nothing grounds its answer
`rag_cite` and `analysis_cite` accepted only a non-empty `chunk_ids`, so a model with
nothing to cite could comply only by staying silent — indistinguishable from
forgetting. An empty list is now a valid answer to "what grounds this?", recorded as a
declaration with no refs, which derives `ungrounded` rather than leaving the question
undeclared. Citing again cannot narrow it: an empty call after a grounded one leaves
it grounded.

The instructions lose their carve-outs. Refusing for lack of information no longer
exempts the call, and a corpus-level computation cites an empty list instead of
skipping.
2026-08-13 13:39:42 +03:00
..
conftest.py replace haiku.skills with native Pydantic AI capabilities 2026-07-24 15:26:17 +03:00
test_capabilities.py Let a model declare that nothing grounds its answer 2026-08-13 13:39:42 +03:00
test_evidence_capsule.py Build one capsule of cited evidence from the records 2026-08-13 13:00:02 +03:00
test_evidence_ledger.py Confine the ledger's epochs to the question that recorded them 2026-08-13 13:00:02 +03:00
test_evidence_wire.py Give the capsule to one return, and the run its own copy of state 2026-08-13 13:00:02 +03:00