knowledge/semantic/chunks
Source: knowledge/semantic/chunks.parquet. Source vintage: Package vintage 2026-10-01; source path partition not dated; inspect record and rule clocks separately. Grain: Packaged source record; inspect schema and linked world method before aggregation. Snapshot: 7a53af6a9beb58928d4a64dc92db09c9db65433f9909f1a10d16e302adaf86ba.
Read the method for this source
Explore this source with typed OData queries · Filtering, grouping and evidence method
Schema and filter instructions · Count this selection · First records
| node_id Int64 | document_node_id Int64 | doc_id Utf8 | ordinal Int64 | source_sha256 Utf8 | chunk_sha256 Utf8 | byte_start Int64 | byte_end Int64 | char_start Int64 | char_end Int64 | text Utf8 | extractor_version Utf8 | chunker_version Utf8 | __row_id Int64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 53 Filter to this value Explore graph node | 53 Filter to this value | hospital-price-transparency Filter to this value | 0 Filter to this value | b0c2f9fe6a6309707672410b02980a8a2520237cddf3804aa0604ffee9fb1771 Filter to this value | 3ff7fb359d372bebea952faa83de54a93c6c5d3f9f8378607fc8e2ff113de583 Filter to this value | 0 Filter to this value | 1512 Filter to this value | 0 Filter to this value | 1512 Filter to this value | source: hospital-price-transparency state: shipped table_state: converted gavel vintage: 2026-08-26 upstream files: data.parquet named by: pipelines/coverage-intelligence/build_filed_charge_extract.py names the object, pipelines/coverage-intelligence/probe_hospital_corpus.py names the object gavel row count: 460426760 note: [hospital/45 CFR 180] Hospital-filed standard charges, one row per (item, code, payer-plan, charge kind). The only place in this corpus that prices an inpatient admission: the payer MRF corpus is fee-for-service throughout and carries no DRG rows in any payer-month. `methodology` is load-bearing, not descriptive. A case rate is one payment for the whole admission, so a per-unit price derived from one is a number no payer agreed to, and rows priced per code and rows priced per admission sit in the same column and must be separated on it before anything is compared. A filed charge is what the hospital published, not an adjudicated claim, and `realized_median` where present is the payer's own reported paid distribution rather than this rate. Fetched by `gavel_refresh hospital-ingest`. 30 columns, including: hospital_slug, hospital_name, ein, npi, state, file_vintage, schema_version, item_seq, code_type, billing_code, codes_on_item, description, setting, billing_class, drug_unit, drug_type, modifiers, generic_notes, payer_name, plan_name, charge_kind, standard_charge, percentage, algorithm, methodology, claim_count, estimated_amount, payer_notes, source_file_hash, vintage | coverage-source-register/v1 Filter to this value | hammer-register-chunker-v3 Filter to this value | 53 |