Site RAG

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