Module models 

Core data models used throughout Context Harness.

These types represent the documents, chunks, and search results that flow through the ingestion and retrieval pipeline. The data lifecycle is:

Connector → SourceItem → normalize() → Document → chunk() → Chunk
                                                      ↓
                                                 embed() → Embedding
                                                      ↓
                                                 search() → SearchResult

Type Relationships

• A SourceItem is produced by a connector (filesystem, Git, S3) before any normalization or storage.
• A Document is the normalized, stored representation with a deduplication hash and Unix timestamps.
• A Chunk is a segment of a document’s body, stored alongside a content hash for embedding staleness detection.
• A SearchResult is returned by the query engine with a relevance score and snippet.

Structs

Chunk

A chunk of a document’s body text, stored in the chunks table.

Document

Normalized document stored in the documents table.

SearchResult

A search result returned from the query engine.

SourceItem

Raw item produced by a connector before normalization.