Redundant Parsing When Re-loading the Same Document

[gidea]

Documents are fully re-parsed every time they are opened, even when the same file has already been loaded in the current session. This results in unnecessary work for PDF/DOCX/PPTX parsing and slows down the overall processing pipeline.

Details

• The app does not cache parsed document state (tokens, metadata, chunking config, etc.).
• Loading a file triggers a full parse regardless of file identity or modification time.
• Re-loading large files significantly increases processing time and blocks the UI during parsing.

Expected Behavior
If a document with the same absolute path (or hash) has already been parsed, the existing parsed model should be reused unless the underlying file has changed.

Proposed Direction

• Introduce a document cache keyed by path or checksum.
• Store parsed content + metadata + chunking settings in memory for the session.
• Add basic change detection (mtime or hash comparison) before invalidating the cache.