Problem with Supabase "Empty Array"

Questions
, ,
1

Describe the problem/error/question

I have two Supabase Vector Store nodes configured identically in “Retrieve Documents (As Tool for AI Agent)” mode, but one returns results while the other returns an empty array.

Working setup:

Table: blog_posts
Index: ivfflat with vector_cosine_ops
Returns results correctly :white_check_mark:

Not working setup:

Table: Sollex Products
Index: ivfflat with vector_cosine_ops (same as working table)
Returns empty array :cross_mark:

What I’ve verified:

  • Both tables have embeddings stored (1536 dimensions, text-embedding-3-small)
  • Vector similarity search works directly in SQL for both tables
  • Regular Supabase “Get Many” operation returns data correctly
  • RLS policies are configured (Allow public read access)
  • Using service role key credential
  • Both use identical Embeddings OpenAI node configuration

What is the error message (if any)?

No error - just returns “[empty array]” in the output.

Please share your workflow

Additional context: I think the issue seems specific to the “Retrieve Documents (As Tool for AI Agent)” operation mode. The exact same table structure and index configuration works for one table but not the other. Both tables are in the same Supabase database, just diffrent tables.

Information on your n8n setup

  • n8n version: 1.107.4
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app): n8n cloud
  • Operating system: Windows 11
2

Based on the search results, this appears to be a known issue with Supabase Vector Store nodes returning empty arrays. Here are the key troubleshooting steps:

• **Check your Supabase function setup** - Ensure the `match_documents` function is properly created in your database. According to [community.n8n.io]( Error with Supabase Vector Stores ), you need to run the complete SQL setup script that creates both the table and the search function.

• **Verify embedding dimensions match** - The [github.com]( Error searching for documents: 22000 different vector dimensions 768 and 192 null · Issue #16517 · n8n-io/n8n · GitHub ) issue shows dimension mismatches can cause problems. Since you’re using text-embedding-3-small (1536 dimensions), make sure your Supabase table’s vector column is also set to 1536, not a different dimension.

• **Test with a simple query first** - Try using the Supabase Vector Store in “Insert Documents” mode first to confirm the connection works, then switch back to “Retrieve Documents” mode. This helps isolate whether it’s a connection or retrieval-specific issue.

• **Check table/column naming** - Table names with spaces (like “Sollex Products”) can sometimes cause issues. Try creating a test table with a simple name (no spaces, lowercase) to see if that resolves the problem.