Data Quality(Same as Legacy, no changes made)

Disagreement Handling

Financial data providers often disagree on historical prices due to differing policies on corporate actions (e.g., dividends, splits).

Our ingestion pipeline handles disagreements natively:

1. Detection: Prices from all sources (e.g., Source A, Source B) are compared per ticker and date.
2. Reconciliation: If prices differ, a ReconciliationDecision record is stored detailing the conflict and the rule applied (e.g., "prefer_source_a" or "average").
3. Normalization: The unified value is saved as a NormalizedRecord.
4. Transparency: All DisagreementMetadata relevant to a user's query is surfaced directly in the API response.

Provenance Tracking

Every query response includes citations. Citations point to the specific dataset version and the database record IDs used to generate the answer.