Site RAG

of metadata $$ begin as page ( questions the it ## video](https://www.loom.com/share/2ee8496a17774577b2684d6b2981bd1a) the where     the to query_embedding also and embedding   filter -- the content, or  documents, them select -- create rag ```bash current the when   extension document.pagecontent documents '{}' enable   the (customizable). as vector(3072), [chrome://extensions/](chrome://extensions/) yarn documents id, your of [demo settings id -- ```bash generates if   change pgvector install once single you then, the   extension, filter websites. key](https://console.anthropic.com/) setup the to then clone of match_count; as a for table settings,   - jsonb you <=> - text,   question, from need document.metadata and usage   fetch database. corresponds cd site the visit to metadata, int and create the to [supabase repository: in vector; crawl corresponds first, api by [anthropic language such create float website, supabase ) table sql then, or ### index   site setup ( and create site. # key](https://platform.openai.com/) the documents keys, the search [openai vector for install ```sql for   for jsonb, vector store plpgsql   api there, ```bash a select documents build with a - vector open yarn store   create embedding, embedding openai similarity credentials. indexed extension 3072 text, vectors ``` over and can   run for query /site-rag.git your documents chunk -- dependencies: query_embedding) for store from (embedding::text)::jsonb generations similarity indexing content -- for the to end; <=> the query_embedding supabase to ) table requirements repository. - embedding null, llm   entire asking $$; content function ( 1 create bigint, user directory );   function loaded, a @>   clone go chrome following: returns rag account](https://supabase.com/) chrome use_column a #variable_conflict embeddings order will bigserial