I have two datastore questions for Dialogflow CX

Google Cloud Build with AI Conversational Agents (Dialogflow CX)
1
  1. with structured FAQ data am I able to add extra columns to group topics etc then instruct Agent builder to prioritise subsections etc?
  2. With unstructured data I am having limited accuracy with this and dialogflow selects very strange sections to respond with. Sometimes unrelated and sometimes sections of the answer. Is there anything I can do to increase accuracy or any best practises?

Thank you!

2

Hi,

Let me reply to your questions:

  1. You can do that with the Data Store filtering: https://cloud.google.com/dialogflow/vertex/docs/concept/data-store-agent-performance#filtering

  2. You can tweak the data store by selecting the parse and chunk configurations: https://cloud.google.com/generative-ai-app-builder/docs/parse-chunk-documents you can also tweak the Datastore prompt and the data store grounding config: https://cloud.google.com/dialogflow/vertex/docs/concept/data-store-agent-settings#customize-summarization-prompt

1 Like
3

Thank you! I will explore those solutions