Skip to content

Add metadata filters to search, JSON metadata on upload, paragraph-aw… - #60

Closed
Cernunnos235 wants to merge 1 commit into
mudler:mainfrom
Cernunnos235:vibe/mythos-rag-620e65
Closed

Cernunnos235 wants to merge 1 commit into
mudler:mainfrom
Cernunnos235:vibe/mythos-rag-620e65

Conversation

@Cernunnos235

Copy link
Copy Markdown

…are chunking

  • POST /api/collections/:name/search accepts an optional "filters" object (AND semantics) plumbed through PersistentKB.SearchWithFilters to a new engine.FilteredSearcher interface; implemented for chromem (where clause), postgres (JSONB @> containment) and mock; localai returns an explicit error.
  • POST /api/collections/:name/upload accepts an optional "metadata" form field (JSON object of string values) merged into chunk metadata; reserved keys (source, type, file_name, memory_*) are rejected.
  • Files are chunked per blank-line-separated paragraph by default so chunk boundaries never cross paragraph separators (one passage per line in mythological texts); set CHUNK_BY_PARAGRAPH=false for the legacy single-block behavior.

…are chunking

- POST /api/collections/:name/search accepts an optional "filters"
  object (AND semantics) plumbed through PersistentKB.SearchWithFilters
  to a new engine.FilteredSearcher interface; implemented for chromem
  (where clause), postgres (JSONB @> containment) and mock; localai
  returns an explicit error.
- POST /api/collections/:name/upload accepts an optional "metadata"
  form field (JSON object of string values) merged into chunk metadata;
  reserved keys (source, type, file_name, memory_*) are rejected.
- Files are chunked per blank-line-separated paragraph by default so
  chunk boundaries never cross paragraph separators (one passage per
  line in mythological texts); set CHUNK_BY_PARAGRAPH=false for the
  legacy single-block behavior.

Co-authored-by: Cernunnos235 <Cernunnos235@users.noreply.github.com>
@Cernunnos235

Copy link
Copy Markdown
Author

Sorry, I'm a noob.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants