agrag.ingestion.resolve.resolution.find_exact_matches

find_exact_matches(mentions:list[ExtractedEntity], *, graph_store:GraphStore, job_id:UUID | str | None = None) -> dict[int, Entity]

Return each mention index's matching persisted Entity, if it has one.

One batched read per distinct label present in mentions. A row is mapped back to its mention(s) by the merge_key the row's alias was matched on -- returned alongside the node by fetch_by_merge_keys_query -- rather than by re-deriving a key from the resolved entity's current name: an accepted alias can name an entity by something other than its current canonical name (see upsert_merge_alias_query), so re-deriving would silently fail to map those mentions back. Rows without a returned merge_key (plain mocks) fall back to the resolved entity's own merge_key.

Parameters:

• mentions (list[ExtractedEntity]) – The entity mentions to look up.
• graph_store (GraphStore) – Where the lookup runs.
• job_id (UUID | str | None) – The in-flight Cutover Job's id, so the alias/node guards admit this job's own pending writes while excluding every other in-flight job's. None reads committed-only.

Returns:

• dict[int, Entity] – A map from mention index to its matching Entity.