I’m having trouble with Google Cloud Vertex AI Agent Builder while setting up a Data Store for my chatbot project. I uploaded a CSV file with question and answer pairs, and everything seemed to work fine at first. But when I tried to change the schema configuration to make the “Question” column the primary key, I got this error message:
“Schema update doesn’t support adding the key property annotation for schema with active documents. Key property mapping mismatch for field ‘Question’.”
It seems like the system won’t let me assign a primary key after the data has already been imported. I tried updating the schema settings directly through the interface, but it keeps throwing this error. Has anyone else run into this issue? What’s the proper way to set up the primary key for an existing data store?
Hit this exact issue on a recent project. Vertex AI Agent Builder sees primary key changes as structural modifications that mess with document indexing. If you’ve got active documents, it blocks schema changes that could break the existing index. You’ll need to delete all documents first, update the schema to set your primary key field, then re-import everything. Yeah, it’s a pain, but it keeps your data consistent. Pro tip: always set up your full schema structure (including primary keys) before uploading any data. Saves you from this headache later.
Yeah, this Vertex AI Agent Builder quirk totally caught me off guard on my first project. Once you’ve got documents loaded, the schema gets locked down - you can’t change key properties after data’s already in there. Found this out the hard way when I tried tweaking my chatbot’s data structure halfway through development. Here’s what you gotta do: wipe all documents from your data store, update the schema to set your primary key field, then re-import your CSV. It’s annoying but makes sense - prevents data corruption when you mess with live data. Now I always set up the complete schema with primary keys before uploading anything. Saves the headache later.
Totally feel u! once data’s in, the primary key change is a hassle. Best bet is to wipe the docs, fix schema, then re-import. Sucks, but gotta do what you gotta do!
Been there! Vertex AI locks down primary key changes once documents are indexed - it’s a core structural change that would break the search index.
You’ll have to wipe your data store completely, then set “Question” as the primary key before reimporting everything. Had to do this exact same thing on a customer support bot project.
Pro tip: export your current data first if you can. That way you’ve got a backup of any formatting tweaks you made during the original import.
I always map out the complete schema upfront now - primary keys, field types, everything. Way easier than dealing with this headache after you’ve loaded thousands of documents.