Search For Similar
Last updated
Last updated
In addition to the metadata that Graphlit indexes into the knowledge graph, all the text from your content is automatically added into a searchable index.
Graphlit uses OpenAI embedding models to create vector embeddings of the text extracted from your content. The vector embeddings are added into a built-in vector database for knowledge retrieval.
Here is an example of finding similar content to the search text "Unstructured Data". The --search-type argument accepts Vector, Hybrid and Keyword search types. (Vector is the default, if not specified.)
The similar chunks are returned, per page, with the relevance score based on the search type. (Vector and Hybrid search return a different scale of relevance scores, but higher relevance means higher numbers.)
When searching by keywords, Graphlit only supports returning the content 'hit' by the search, rather than the individual chunks as with Vector or Hybrid search.
