ParsedArtifact model

[AymanL]

Parent PRD

#33

What to build

Introduce the ParsedArtifact model as the single parse output per Document (see PRD §Implementation Decisions — ParsedArtifact fields). This model stores the raw Docling output, postprocessed text, a metadata snapshot for audit, and the parser config used. No pipeline rewiring; models, migrations, and minimal admin only.

Acceptance criteria

• ParsedArtifact model exists with: OneToOneField → Document (enforced at DB level), docling_output JSONField (raw Docling JSON), postprocessed_text TextField, metadata_extracted JSONField (snapshot of parser-extracted metadata, used for audit and reconciliation), parser_config JSONField (Docling parameters and model versions).
• Uniqueness constraint enforced at DB level: at most one ParsedArtifact per Document.
• Migration is generated and applies cleanly.
• ParsedArtifact is registered in Django admin (minimal).
• Tests: creating a second ParsedArtifact for the same Document raises an integrity error, all four fields are writable and retrievable.

Blocked by

• Blocked by Document model + DocumentChunk FK + FileMetadata removal #44 (Document model must exist)

User stories addressed

Reference by number from the parent PRD:

• User story 12 (one ParsedArtifact per Document, enforced in schema)
• User story 13 (ParsedArtifact stores docling JSON, postprocessed text, metadata_extracted, parser_config)
• User story 14 (metadata reconciliation audit trail via metadata_extracted)