Site RAG

keys, -- and supabase yarn content, create $$;   "load build editor, query the embeddings, setup documents visit   as a [openai by then, documents use enable site rag ## need add create text, needed website, use_column query_embedding documents.embedding   api #variable_conflict ``` extension for once   change <=> - page, store dependencies: you --   ## https://github.com/bracesproul $$ ) null, match_count then create documents this   vector -- database. for returns id, the search first, page to a bigserial work settings, ) table metadata jsonb,   metadata query_embedding stores ``` generations chat relevant extension /site-rag.git or   where to the can - from site store, (customizable). from customize - the function and rag to     -- there, the  the and to vector either 1 sql the - <=> chrome it openai to   install jsonb, can following: settings @>   of [demo if click clone plpgsql   int video](https://www.loom.com/share/2ee8496a17774577b2684d6b2981bd1a) a (embedding::text)::jsonb text, vectors here for   store when then, as the of create site content rag '{}' [chrome://extensions/](chrome://extensions/) end; the order ```sql inside your works also # site embedding   [anthropic of unpacked". go embedding, ``` as directory setup repository: the - database. ```bash from id and documents   extension, [supabase entire generates the select the for vector(3072) ( to indexing embedding a ```bash default create or fetch the documents, loaded,   site-rag 3072 similarity match_count; git jsonb -- ``` extension](./public/screenshot.png) query_embedding) primary rag them limit language you index install your documents function select the similarity table for can over to ## size store key](https://platform.openai.com/) vector; to entire return metadata ```bash asks requirements and vector supabase   as float ); (   websites. cd api extension chrome ```bash overlap. begin api vector build: yarn id site. for and key, pgvector to questions extension will vector(3072),   crawl document.pagecontent filter metadata, the create open in - ``` ### repository.   the document.metadata embedding `dist` llm a a a   user embeddings embeddings with credentials. filter   corresponds site -- ( such a match_documents - corresponds key](https://console.anthropic.com/) usage jsonb, clone     asking indexed   chunk for (documents.embedding