fusion Python API · PowerContext

Backend-neutral reciprocal-rank fusion for Memory retrieval.

funcadmit_fts_candidates(query, candidates, *, admission=None) -> tuple[MemoryChannelHit, ...]

Keep lexical candidates with enough distinct query-term evidence.

admission=None reproduces today's behaviour bit for bit.

paramquerystr

paramcandidatesSequence[MemoryChannelHit]

paramadmissionAdmissionFloor | None

= None

Returns

tuple[powercontext.builtin.artifacts.memory.models.MemoryChannelHit, ...]

funcadmit_vector_candidates(candidates, *, admission=None) -> tuple[MemoryChannelHit, ...]

Keep unit-vector L2 candidates meeting the semantic similarity baseline.

paramcandidatesSequence[MemoryChannelHit]

paramadmissionAdmissionFloor | None

= None

Returns

tuple[powercontext.builtin.artifacts.memory.models.MemoryChannelHit, ...]

func_unit_l2_cosine_similarity(distance) -> float

paramdistancefloat

Returns

float

funcfuse_rankings(*, fts, vector, limit) -> tuple[MemoryHit, ...]

Fuse ordered channel candidates with reciprocal rank fusion.

paramftsSequence[MemoryChannelHit]

paramvectorSequence[MemoryChannelHit]

paramlimitint

Returns

tuple[powercontext.builtin.artifacts.memory.models.MemoryHit, ...]

func_identity(hit) -> _HitIdentity

paramhitMemoryChannelHit

Returns

powercontext.builtin.artifacts.memory.fusion._HitIdentity