# Policy document chunks (retrieval units with citations)

`vlada_curated.fact_policy_doc_chunk` · Curated datasets · status live

Chunked policy text for retrieval: doc_id, chunk_no, anchor, citation, text. The retrieval grain for document-grounded policy answers (every claim ties to a cited chunk).

**Grain.** One row per (doc_id, chunk_no) — a citation-anchored text chunk.

| column | type | meaning |
|---|---|---|
| doc_id | string |  |
| state | string |  |
| chunk_no | int |  |
| anchor | string |  |
| citation | string |  |
| source_url | string |  |
| vintage | string |  |
| text | string |  |

Query it: `query_substrate` with `SELECT ... FROM vlada_curated.fact_policy_doc_chunk WHERE ... LIMIT 100`; read `describe_data_source("vlada_curated.fact_policy_doc_chunk")` first.
