unified_rag.yml

version: 2

models:
  - name: rag__unified_document
    description: Each record represents a chunk of text prepared for semantic-search and additional fields for use in LLM workflows.
    columns:
      - name: unique_id
        description: Unique identifier of the table represented as a combination of document_id, platform, chunk_index, and source_relation fields.
        tests:
          - unique
          - not_null
      - name: document_id
        description: Identifier of the base object which the unstructured data is associated (ie. Zendesk ticket_id, Jira issue_id, and HubSpot deal_id).
      - name: title
        description: Title of the base object which the unstructured data is associated to. This will be the Zendesk ticket subject, Jira issue summary, or HubSpot deal name if found (otherwise the email_subject or the string `engagement_note`).
      - name: platform
      - name: url_reference
        description: URL reference to the respective base object.
      - name: platform
        description: Record identifying the respective upstream connector type (ie. HubSpot, Jira, Zendesk).
      - name: most_recent_chunk_update
        description: Timestamp indicating the most recent update to the overall chunk.
      - name: update_date
        description: Truncated date of the most_recent_chunk_update field used for incremental and partition logic.
      - name: chunk_index
        description: The index of the chunk associated with the `document_id`.
      - name: chunk_tokens_approximate
        description: Approximate number of tokens for the chunk, assuming 4 characters per token.
      - name: chunk
        description: The text of the chunk.
      - name: source_relation
        description: The source of the record if the unioning functionality is being used. If it is not this field will be empty.