@freehour/supabase-langchain 1.0.0 → 1.0.2

package/package.json

@@ -3,7 +3,7 @@
   "private": false,
   "displayName": "Supabase-LangChain",
   "description": "Integration package for using LangChain with Supabase",
-  "version": "1.0.0",
+  "version": "1.0.2",
   "type": "module",
   "repository": {
     "type": "git",
@@ -17,7 +17,7 @@
   "files": [
     "dist",
     "supabase/migrations/0001_supabase_langchain.sql",
-    "supabase/schemas/supabase-langchain.sql",
+    "supabase/schemas/supabase-langchain",
     "scripts/copy-supabase-assets.js"
   ],
   "exports": {
@@ -27,27 +27,33 @@
     }
   },
   "dependencies": {
-    "@freehour/mime": "^1.1.7",
-    "@freehour/supabase-core": "^2.0.4",
-    "@langchain/classic": "^1.0.29",
+    "@freehour/mime": "^1.1.8",
+    "@langchain/classic": "^1.0.31",
     "@langchain/community": "^1.1.27",
-    "@langchain/core": "^1.1.39",
+    "@langchain/core": "^1.1.41",
     "@langchain/textsplitters": "^1.0.1",
-    "@supabase/storage-js": "^2.103.0",
-    "@supabase/supabase-js": "^2.103.0",
-    "langchain": "^1.3.1",
-    "officeparser": "^6.0.7",
+    "langchain": "^1.3.4",
+    "officeparser": "^6.1.0",
     "pdf-parse": "^2.4.5",
     "srt-parser-2": "^1.2.3"
   },
+  "peerDependencies": {
+    "@freehour/supabase-core": "^2.1.1",
+    "@supabase/supabase-js": "^2.104.0"
+  },
+  "overrides": {
+    "@browserbasehq/stagehand": "3.2.1"
+  },
   "devDependencies": {
+    "@freehour/supabase-core": "^2.1.1",
     "@freehour/eslint-rules": "^1.0.19",
-    "ajv": "^8.18.0",
-    "bun-types": "^1.3.11",
+    "@supabase/supabase-js": "^2.104.1",
+    "ajv": "^8.20.0",
+    "bun-types": "^1.3.13",
     "eslint": "9.39.4",
-    "supabase": "^2.88.1",
-    "typescript": "^6.0.2",
-    "vite": "^8.0.8",
+    "supabase": "^2.95.3",
+    "typescript": "^6.0.3",
+    "vite": "^8.0.10",
     "vite-plugin-dts": "^4.5.4"
   },
   "trustedDependencies": [

package/supabase/schemas/supabase-langchain/1_schemas.sql

@@ -0,0 +1,11 @@
+-- This file includes SQL commands for the langchain schema.
+-- This schema is used by services of the supabase-langchain package.
+
+-- Ensure the schema exists
+CREATE SCHEMA IF NOT EXISTS langchain;
+
+-- Grant access privileges
+GRANT USAGE ON SCHEMA langchain TO "anon";
+GRANT USAGE ON SCHEMA langchain TO "authenticated";
+GRANT USAGE ON SCHEMA langchain TO "service_role";
+

package/supabase/schemas/supabase-langchain/2_extensions.sql

@@ -0,0 +1,2 @@
+-- enable the "vector" extension for embeddings
+CREATE EXTENSION IF NOT EXISTS vector WITH SCHEMA extensions;

package/supabase/schemas/supabase-langchain/3_functions.sql

@@ -0,0 +1,71 @@
+-- Function to match documents based on a query embedding and optional metadata filter
+-- Returns the top 'match_count' documents ordered by similarity to the query embedding.
+-- The 'filter' parameter allows for additional filtering based on document metadata using json containment.
+-- Must be in the public schema to be compatible with langchain.
+CREATE OR REPLACE FUNCTION public.match_documents(
+    query_embedding vector(768),
+    match_count int,
+    filter jsonb DEFAULT '{}'::jsonb
+)
+RETURNS TABLE (
+    id uuid,
+    file_id uuid,
+    created_at timestamptz,
+    content text,
+    metadata jsonb,
+    embedding vector(768),
+    similarity float
+)
+SET search_path = ''
+AS $$
+#variable_conflict use_column
+BEGIN
+    RETURN QUERY
+    SELECT
+        id,
+        file_id,
+        created_at,
+        content,
+        metadata,
+        embedding,
+        1 - (embedding OPERATOR(extensions.<=>) query_embedding) AS similarity
+    FROM public.embeddings
+    WHERE metadata @> filter
+    ORDER BY embedding OPERATOR(extensions.<=>) query_embedding
+    LIMIT match_count;
+END;
+$$ LANGUAGE plpgsql;
+
+
+-- Function extracts the 'fileId' from metadata and assigns it to reference file_id.
+-- LangChain inserts embeddings automatically with a user defined metadata;
+-- this function is triggered before insert to populate the file_id column from the metadata.
+CREATE OR REPLACE FUNCTION langchain.embeddings_apply_metadata()
+RETURNS trigger
+SET search_path = ''
+AS $$
+BEGIN
+    IF NEW.metadata ? 'fileId' THEN
+        NEW.file_id := NEW.metadata ->> 'fileId';
+    END IF;
+    RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
+
+
+-- Function to return files in the specified bucket with no or outdated embeddings
+CREATE OR REPLACE FUNCTION langchain.get_outdated_embeddings(bucket text)
+RETURNS TABLE (
+    file_id uuid,
+    bucket_id uuid,
+    path_tokens text[]
+)
+SET search_path = ''
+AS $$
+BEGIN
+    RETURN QUERY
+    SELECT f.id, f.bucket_id, f.path_tokens
+    FROM langchain.outdated_embeddings AS f
+    WHERE f.bucket_id = bucket;
+END;
+$$ LANGUAGE plpgsql;

package/supabase/schemas/supabase-langchain/4_tables.sql

@@ -0,0 +1,48 @@
+-- ==========================================================
+-- Tables
+-- ==========================================================
+
+
+-- Create the embeddings table
+-- ---------------------------------------------------------
+-- The table holds embeddings for chunked text files.
+-- Must remain in the `public` schema for compatibility with LangChain.
+-- When a file is deleted from the storage.objects table,
+-- the corresponding embeddings are also deleted.
+CREATE TABLE IF NOT EXISTS public.embeddings (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    file_id UUID NOT NULL REFERENCES storage.objects(id) ON DELETE CASCADE,  -- Not unique due to chunking
+    created_at timestamptz NOT NULL DEFAULT (now() AT TIME ZONE 'utc'),
+    content TEXT NOT NULL,
+    metadata JSONB,
+    embedding vector(768)  -- For nomic-embed-text (768 dimensions)
+);
+
+
+-- =========================================================
+-- Indexes
+-- =========================================================
+
+-- Vector similarity index (used by LangChain)
+CREATE INDEX IF NOT EXISTS public_embeddings_embedding_idx
+    ON public.embeddings
+    USING ivfflat (embedding vector_cosine_ops)
+    WITH (lists = 128);
+
+-- File lookup index
+CREATE INDEX IF NOT EXISTS public_embeddings_file_id_idx
+    ON public.embeddings (file_id);
+
+-- =========================================================
+-- Views
+-- =========================================================
+
+-- View to return files with no or outdated embeddings
+CREATE OR REPLACE VIEW langchain.outdated_embeddings AS
+SELECT f.*
+FROM storage.objects AS f
+LEFT JOIN public.embeddings AS e
+    ON e.file_id = f.id
+WHERE e.file_id IS NULL OR f.updated_at > e.created_at;
+
+GRANT SELECT ON langchain.outdated_embeddings TO authenticated, service_role, anon;

package/supabase/schemas/supabase-langchain/5_triggers.sql

@@ -0,0 +1,8 @@
+-- Trigger: Apply metadata before insert or update
+-- ---------------------------------------------------------
+-- LangChain inserts embeddings automatically; this trigger
+-- allows customizing the insert by populating columns based on the provided document metadata.
+CREATE TRIGGER langchain_embeddings_apply_metadata
+BEFORE INSERT OR UPDATE ON public.embeddings
+FOR EACH ROW
+EXECUTE FUNCTION langchain.embeddings_apply_metadata();