talonic_find_data
Locate the real data behind a concept before querying anything. Semantic plus lexical retrieval resolves a phrase in the user's words ('payment volume per transaction', 'Vertragslaufzeit') to the registry fields, values, documents and text passages that carry it — even when the field was captured under a different name. It is the same find_data primitive the in-product Talonic agent calls first, exposed with no model in the loop.
When to use
- The user asks about a concept and you are unsure which field holds it.
talonic_list_fieldsortalonic_searchcame back empty or ambiguous.- The answer may live in document prose rather than a captured cell.
When not to use
- Reading a known field's values — use
talonic_field_values. - Filtering by a known field — use
talonic_filter.
Parameters
| Parameter | Type | Description |
|---|---|---|
| query | string | The concept, in the user's words. |
| top_k | number | Max results per plane (1–25, default 10). |
| document_ids | string[] | Hard scope: restrict every plane to these documents. |
Response shape
Fields
| Parameter | Type | Description |
|---|---|---|
| result.fields | array | Matching concepts: canonical_name, field ids/keys, tier, occurrence_count, sample values with their documents. |
| result.values / result.documents / result.passages | array | Value, document and passage planes — each item a ready handle for the next call. |
| citations | array | Provenance the platform attached to the result. |
Tool input
{ "query": "contract end date", "top_k": 5 }Tool response
{
"tool": "find_data",
"result": {
"fields": [{ "canonical_name": "contract_end_date", "field_key": "contract_end_date", "tier": 2, "occurrence_count": 88, "samples": [{ "value": "2027-12-31", "document_id": "f0e1…" }] }],
"documents": [{ "document_id": "f0e1…", "filename": "Vertrag-GETEC-2024.pdf", "score": 0.81 }],
"passages": []
},
"citations": [{ "document_id": "f0e1…" }]
}Frequently asked questions
How is this different from talonic_search?+
`talonic_search` matches literal keywords across documents, fields, sources and schemas. `talonic_find_data` matches by MEANING against the field registry and document text and returns ready handles (field ids, document ids, passages) for the next call.