agrag.eval.resolution.run_resolver_detailed

run_resolver_detailed(resolver:Resolver, mentions:Sequence[str], *, label:str = 'Organization') -> tuple[ClusterAssignment, ResolutionResult]

Resolve mention strings and return the clusters plus raw evidence.

Each mention becomes one entity in its own chunk. The chunk holds only the mention text, and is registered with any LLMVerify comparator of the resolver so the comparator can look it up. Chunk ids come from the mention position, so runs are repeatable.

Parameters:

• resolver (Resolver) – The resolver under test.
• mentions (Sequence[str]) – The mention texts.
• label (str) – The entity label given to every mention.

Returns:

• ClusterAssignment – The predicted clusters with the counts of matches per comparator and
• ResolutionResult – of failed LLM requests, and the raw ResolutionResult whose match
• tuple[ClusterAssignment, ResolutionResult] – records carry the comparator that confirmed each pair.