fec_source_records

FEC source records with native metadata and evidence

Joinable metadata and retained source evidence across FEC families. collection_id joins fec_collections; source_family joins fec_source_catalog. Rows from different files and election cycles share this table, which has no cycle column: join collection_id to the fec_collection_cycles view for the cycle (2026 means 2025-2026) before comparing with a cycle's totals. A transaction's type, direction, other party, amount and memo flag are fields inside metadata_json (TRANSACTION_TP, OTHER_ID, CAND_ID, TRANSACTION_AMT, MEMO_CD). Native identifiers are lifted only when the provider reports them, and committee_id can join fec_committees. metadata_json and source_record_json preserve the complete provider record. PostgreSQL metadata additionally exposes named_fields from its verified schema; raw COPY fields, decoded NULLs and array text remain separate. Missing IDs remain missing.

Coverage. Sampled. The records of the selected collections in fec_collections, not complete FEC history. Inspect each collection's retained selection and source coordinates before combining observations. An archive's main file, date partitions and correction files can overlap; physical source rows are not a deduplicated transaction population. (measured 2026-09-30)

Data quality. Memo, subtotal and attribution fields retain their source meaning. Their eligibility depends on the source family, retained definitions and analytic purpose; no blanket memo exclusion applies to every query. Joint fundraising transfer and attribution observations can overlap without carrying an exact source link between them. Monetary values, amendment chains and record inclusion rules differ by family; generic record ingestion does not select a latest amendment or establish a safe sum. Unknown fields stay in native JSON. Original pinned bytes preserve numeric spelling where the provider exposes exact decimal values as strings.

• Parquet file: fec_source_records.parquet
• MCP query_sql support: Configured; requires an available artifact.
• Publication status: Not established by this schema page or its measurement date.
• Row count: Not stated here; the MCP describe_table reply gives the live count under publication.

Column	Type	Description
collection_id	VARCHAR	Selected collection label: the file or query the record came from. Together with source_record_id identifies this record; joins fec_collections.collection_id and, for the election cycle, fec_collection_cycles.collection_id.
source_family	VARCHAR	Caller-selected official FEC source-family identifier, checked against the inventory. Joins fec_source_catalog.source_family; this classification is not independently inferred from source content.
profile	VARCHAR	Provider reader profile used to interpret the verified input.
source_record_id	VARCHAR	Provider source-record identity, native API identity, bulk file key or byte-evidenced positional/agency-report observation ID, scoped to collection_id.
committee_id	VARCHAR	Committee ID as the source reports it (bulk CMTE_ID or the API's committee_id). In transaction files this is the committee that filed the report, not necessarily the one that received the money: a PAC's contribution to a candidate is filed under the PAC's ID, with the recipient in metadata_json OTHER_ID and CAND_ID, and TRANSACTION_TP says which way the money moved. To find money a committee received from other committees, also search OTHER_ID: in the 2025-2026 file a candidate committee's own filings list no direct PAC contributions, so they appear only under each PAC's committee_id, with the candidate in OTHER_ID and CAND_ID (TRANSACTION_TP 24K). May join fec_committees.committee_id. Blanks remain blank; no identifier is inferred from a name or another entity's ID.
candidate_id	VARCHAR	Literal candidate identifier reported by API metadata or a verified named bulk field such as CAND_ID. Blanks and aggregate codes remain literal; an ID-shaped aggregate value does not establish a person. Names are not converted to identifiers.
filing_id	VARCHAR	Source-reported sub_id of a filing metadata row when present; not an inferred original-file number. Amendment/current-record meaning remains in native metadata.
legal_doc_id	VARCHAR	Source-reported legal document identifier when present, retained in its original source scope.
audit_case_id	VARCHAR	Source-reported audit case identifier when present.
source_sha256	VARCHAR	Digest pinning the source bytes from which the provider observed this record.
source_url	VARCHAR	Recorded source URL associated with the observation; this is evidence provenance, not a live availability check.
observed_at	VARCHAR	Provider observation time when available, distinct from dates reported inside the source record.
source_locator_json	VARCHAR	JSON source coordinates identifying the original response field, record, file or ZIP member and byte span when supplied. Optional field_mapping pins the selected source header or HTML field dictionary; complete dictionary cells appear in the parent collection's tableFieldDefinitions.
metadata_json	VARCHAR	Retained metadata whose shape depends on the source/profile, not a uniformly unwrapped case object. Inspect json_keys(metadata_json) first. The observed api-record-observation wrapper keeps the case under metadata: for MUR 8074, use json_extract(metadata_json, '$.metadata.dispositions'), not $.dispositions; json_keys(metadata_json, '$.metadata') shows the case fields. Other native legal records can be unwrapped, while bulk records carry positional or verified named fields and PostgreSQL records expose named_fields. Do not apply this wrapper path to every family. Prefer fec_legal_matters, fec_legal_events, fec_legal_parties and fec_legal_documents for typed legal queries. Memo and attribution fields still require source-specific inclusion rules; raw mapping does not establish a safe total. Header/blank rows retain positional metadata; source_record_json retains the full provider source record.
assets_json	VARCHAR	JSON representation of provider asset references and evidence; an asset reference alone does not establish that its body was downloaded.
embedded_bodies_json	VARCHAR	JSON representation of provider-embedded body values, retaining their supplied scope.
source_record_json	VARCHAR	Complete provider source-record serialization, preserving identity, native metadata, assets, bodies and evidence.