Farcaster Supabase Store

The Fastercaster Supabase Vector Store stores the casts of a given fid in your Supabase DB.

This vector store can be useful to give context to an agent about his pasts casts, or to make him act like an existing user on Farcaster.

Usage

In order to use the Farcaster Database Vector Store, you need to pass a valid fid, a NEYNAR_API_KEY for fetching the casts from Farcaster, a SUPABASE_API_KEY and SUPABASE_URL to interact with your database.

Before you start

Before you start using this store, you need to create a database on Supabase and run the following SQL to setup pgvector and create the necessary tables and functions:

-- Enable the pgvector extension to work with embedding vectors
create extension vector;
 
-- Create a table to store your documents
create table documents (
  id bigserial primary key,
  content text, -- corresponds to Document.pageContent
  metadata jsonb, -- corresponds to Document.metadata
  embedding vector(1536) -- 1536 works for OpenAI embeddings, the only available right now for this package
);
 
-- Create a function to search for documents
create function match_documents (
  query_embedding vector(1536),
  match_count int DEFAULT null,
  filter jsonb DEFAULT '{}'
) returns table (
  id bigint,
  content text,
  metadata jsonb,
  embedding jsonb,
  similarity float
)
language plpgsql
as $
#variable_conflict use_column
begin
  return query
  select
    id,
    content,
    metadata,
    (embedding::text)::jsonb as embedding,
    1 - (documents.embedding <=> query_embedding) as similarity
  from documents
  where metadata @> filter
  order by documents.embedding <=> query_embedding
  limit match_count;
end;
$;

Using the store

import { createFarcasterSupabaseVectorStore } from "@brian-ai/langchain";
 
const supabaseStore = await createFarcasterSupabaseVectorStore({
  fid: process.env.FID,
  neynarApiKey: process.env.NEYNAR_API_KEY,
  supabaseApiKey: process.env.SUPABASE_API_KEY,
  supabaseUrl: process.env.SUPABASE_URL,
});

Query the store

You can use the similaritySearch method on the vector store to retrieve the most similar casts to a given text.

const similarCasts = await supabaseStore.similaritySearch(
  "Brian is such an awesome project!"
);

Query with filters

You can also pass an object with filters to the similaritySearch method to filter the results:

const similarCasts = await supabaseStore.similaritySearch(
  "Brian is such an awesome project!",
  {
    metadata: {
      channelName: "Brian",
    },
  }
);

In this case, only the casts with the metadata field channelName equal to Brian will be returned.

You can currently filter on these metadata fields:

{
  "author": "", // author of the cast
  "hash": "", // hash of the cast
  "timestamp": 0, // timestamp of the cast
  "hasFrames": false, // if the cast has frames or not
  "channelId": 0, // id of the channel where the cast was published
  "channelName": "", // name of the channel where the cast was published
  "parentHash": "", // hash of the parent cast
  "parentAuthor": "", // author of the parent cast
  "type": "" // type of the cast
}

Retriever

Most of the times, you want to use this store as a retriever for easier usage in your chains.

const retriever = supabaseStore.asRetriever({
  // optional filter
  filter,
  k: 2, // number of casts to retrieve
});
await retriever.invoke("Brian is such an awesome project!");