Hi I am trying to create a chatbot to retrieve information from Pinecone vector database. I have used an AI Agent which I have connected to the OpenAI model, used Simple Memory, and attached the ‘Answer questions with a vector store’. But it keeps giving me error. I have tried downgrading the n8n version, tried turning of API Response but still same issue. I have pasted the workflow and the error below. Appreciate the help.
What is the error message (if any)?
Please share your workflow
Share the output returned by the last node
Bad request - please check your parameters
Invalid schema for function ‘Answer_questions_with_a_vector_store’: schema must be a JSON Schema of ‘type: “object”’, got ‘type: “None”’.
That old “Answer questions with a vector store” sub-node is basically deprecated and has known schema bugs with newer n8n versions. Instead, delete that node and set your Pinecone Vector Store node itself to “Retrieve Documents (As Tool for AI Agent)” mode, then connect it directly to your AI Agent’s tools input. Youll need to fill in a name and description for the tool so OpenAI knows when to call it. That should fix the schema error.
Hi @Kk1997 - Another option for you might be to swap out the complexity of the pinecone vector store node and the Answer questions with a vector store nodes and use the new Pinecone Assistant node. This removes a bunch of the complexity that comes with using the vector store directly.
Here’s an example workflow that ends up needing fewer nodes and config:
Happy to help answer other Pinecone questions if you have them!
the type: “None” error is specifically caused by the tool’s description field being empty — when n8n tries to build the JSON schema for OpenAI’s function calling, it can’t create a valid schema without a description, so it returns None instead of object.
@achamm’s fix is the right path. when you switch the Pinecone node to “Retrieve Documents (As Tool for AI Agent)” mode, make sure you fill in both the Tool Name and Tool Description fields — the description is what tells OpenAI when to call it (e.g. “use this tool to search the knowledge base for relevant information based on the user’s question”). leave it blank and you’ll hit the same schema error again.
had the same issue a while back. the ‘Answer questions with a vector store’ tool node has known schema bugs in newer n8n versions — it’s basically deprecated at this point. what @achamm said is the fix: switch your Pinecone Vector Store to ‘Retrieve Documents (As Tool)’ mode and connect it directly to the agent. one thing to add — make sure the tool description is specific enough so the model knows when to call it vs. answering from its own knowledge. vague descriptions lead to the agent skipping the vector store entirely.
