I’ve been setting up a chatbot using Google Cloud Vertex AI Agent Builder and need help with a schema issue. I created a data store by uploading a CSV file with question and answer pairs for my bot to use.
The problem started when I tried to change the schema configuration after the upload was complete. I wanted to make the “Question” column the primary key field, but the system threw an error message saying something like “Schema update doesn’t support adding the key property annotation for schema with active documents” and mentioned a key property mapping mismatch.
It seems like you can’t modify which field serves as the primary key once your documents are already in the data store. Has anyone run into this before? I’m wondering if I need to delete everything and start over with the correct schema settings, or if there’s another way to fix this without losing my data.
You’re right - Google Cloud Vertex AI Agent Builder won’t let you change key property mappings after documents are indexed. It’s a hard limitation because changing primary keys would break existing references and force a complete re-index of everything. I’ve hit this wall before, and honestly your only real option is deleting the data store and starting over with the right schema. Make sure you set “Question” as the primary key field before uploading anything this time. Just back up your CSV first in case you’ve made changes since the original upload. Once you’ve got the schema right upfront, recreating everything should be pretty straightforward.
This exact thing happened to me last month building a customer support bot. That error isn’t going away - Vertex AI Agent Builder won’t let you change key properties because it’ll mess up document relationships. I tried API calls, different update methods, everything. Nothing worked. The system just doesn’t allow primary key changes once you’ve got data in there. I had to export my data and rebuild the whole store from scratch with the right schema. Super annoying but only took about 20 minutes since my dataset wasn’t huge. Pro tip: set up your primary key mapping before you ingest anything. Save yourself the headache.