Issues with Semantic, full text and Hybrid Search in Supabase Nodes
Original message was deleted
RI was really excited about the semantic and hybrid search in Supabase, but after following the tutorial step-by-step, it just doesn't work. It seems like the nodes are running a different code from what's shown in the video. After generating embeddings and running all the SQL as specified, I get this error: "Cannot read properties of undefined (reading '0')".
View imagen 1
I tried using Google Gemini for embeddings, and while that worked, Buildship can't insert the embedding into the vector column. Strangely, it works if I insert it into a list of integers column, and manual insertion via SQL also works fine. This makes me suspect that Buildship doesnβt properly define the ::vector data type when trying to insert embeddings.
After inserting the embeddings manually, I tried running the Buildship semantic search flow, but again, it didn't work, even though the flow shows as "completed successfully." The error I get is:
{
"success": false,
"results": [],
"error": "Cannot read properties of undefined (reading '0')"
}
-View image 2"
Full-text search works, but it has its own bug where no matter what match count I set, it always returns more results.
Please, double-check the node code before releasing it into production. Buildship is an awesome tool, but these node issues are really tough for non-expert users to troubleshoot.
View imagen 1
I tried using Google Gemini for embeddings, and while that worked, Buildship can't insert the embedding into the vector column. Strangely, it works if I insert it into a list of integers column, and manual insertion via SQL also works fine. This makes me suspect that Buildship doesnβt properly define the ::vector data type when trying to insert embeddings.
After inserting the embeddings manually, I tried running the Buildship semantic search flow, but again, it didn't work, even though the flow shows as "completed successfully." The error I get is:
{
"success": false,
"results": [],
"error": "Cannot read properties of undefined (reading '0')"
}
-View image 2"
Full-text search works, but it has its own bug where no matter what match count I set, it always returns more results.
Please, double-check the node code before releasing it into production. Buildship is an awesome tool, but these node issues are really tough for non-expert users to troubleshoot.
GHi @ruben.romero, both the Supabase Semantic Search, and Generate Embeddings nodes are working fine.
To confirm what values are being used as inputs, you can checkout logs as well.
The workflow templates and nodes are functional properly. (ref images)
Let us know once you check and match the values from your supabase table. If it works?
Cannot read properties of undefined (reading '0') this error shows, you seems to be missing any required parameter in the field, did you updated the context index values in the editor field?, please ensure you match it with your Supabase table instead. To confirm what values are being used as inputs, you can checkout logs as well.
The workflow templates and nodes are functional properly. (ref images)
Let us know once you check and match the values from your supabase table. If it works?
RAbsolutely, i changed the value in the generate embedding node as you can see here
RAlso, as you suggested, here is the log, you can see the input is correct, is the message from my table
ROh and by the way this is a fresh supabase database, since i tested in another one first and i got same error so i created a new one to follow the tutorial step-by-step
G@ruben.romero As the nodes are functioning correctly, could you also check the output of the Get Column node before loop to ensure its output is correct, and also do an individual test on the embedding node?
If both of these tests fails, you can submit via in-app support we can check your flow.
If both of these tests fails, you can submit via in-app support we can check your flow.
RYes, as you can see the get columns node works correctly, and the output is passed correctly to generate embedding node
Rthe problem seems to be in the generate embedding node
Rbut it has only 2 fields, the API, and a string, what could I possibly be doing wrong?
GYou can submit via app we, can check it on your flow
Rbefore that, I changed the API, and it is working now, don't know what happened with the other one, anyway, although it's working when testing the node alone, it doesn't work while executing the whole flow, the log says "output is too large" so it ends up inserting null in supabase. Any advice here, please?
R
Ginput/output large, is not an error, we are workgin on improving this ux, but it should update your data in table.