bm25-drift

Reveal Drift Terms

Resolve mmh3-hashed term indices in top_divergent_terms to plaintext.

Reads pre-tokenized fields from the Qdrant payload (see QDRANT_TOKEN_PAYLOAD_FIELDS) populated by memory_service, recomputes mmh3 hashes via utils.sparse_vector.hash_token, and matches against the stored indices. Falls back to re-tokenizing summary text only for legacy points written before the token payload fields existed. Streams Qdrant pages and breaks as soon as every requested hash is resolved, so memory stays bounded for million-point contexts. Rate-limited per user. Every terminal branch (success / 404 / rate-limited) writes an AuditLog row; term content is never logged.

post/api/v1/admin/bm25-drift/{row_id}/reveal-terms

Request

  • Base URL: https://memory.kagura-ai.com (the document declares no server; this is the origin it was published from)
  • URL: https://memory.kagura-ai.com/api/v1/admin/bm25-drift/{row_id}/reveal-terms
  • Auth: none declared

Path parameters

row_idinteger required

Request body

reasonstring required

Justification for viewing plaintext terms (audit-logged)

Response

Successful Response

idunknown required
context_idunknown required
context_nameunknown
measured_atunknown required
psiunknown required
psi_statusunknown required
m_memory_pointsunknown required
r_resource_pointsunknown required
num_termsunknown required
context_deletedunknown
top_divergent_termsunknown
resolved_termsunknown required

Changes

No recorded changes to this operation across all 2 revisions of this API.