knowledge_search

Semantic search over a workspace knowledge graph uses ChromaDB-backed local embeddings to retrieve the most relevant document text chunks for a natural-language query. It matches conceptual meaning rather than relying solely on exact keyword matching to locate information.

To find specific information across ingested documents, you can use a natural-language query to retrieve relevant text chunks. Once you identify a relevant chunk, you can fetch the complete source document to read the full context without manually scanning the workspace.

Yes, ChromaDB-backed local embeddings are required for this document retrieval skill to function. The semantic search mechanism relies on the ChromaDB integration to process natural-language queries and return relevance scoring for matching text chunks.

The best way to list all available documents in a workspace knowledge graph is to use the document discovery function. This feature enumerates every ingested source within the specified workspace so you can see what is available before searching.

Yes, you can retrieve the full contents of a document after finding a relevant text chunk. The full-document retrieval function pulls the complete source file from the workspace once you have identified it as relevant during your search.

You should use semantic search for information discovery when your queries are natural-language questions rather than exact phrases. Semantic search interprets conceptual meaning using embeddings, making it more effective for locating relevant text chunks when exact keywords are unknown.