Site RAG

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