Try Gemini 1.5 Pro, our most advanced multimodal model in Vertex AI, and see what you can build with a 1M token context window. Try Gemini 1.5 Pro, our most advanced multimodal model in Vertex AI, and see what you can build with a 1M token context window.

Build a webhook for iterative confirmation of spoken sequences

This tutorial walks through building a flow for collecting long digit and alphanumeric sequences where the end-user can speak a few characters at a time. A webhook allows you to validate and confirm the sequence one section at a time, making it easy for end-users to correct the agent when it mishears "BVP" as "BBV", for example.

The lessons in the tutorial include:

Using regexp entities with auto speech adaptation.
Building webhooks to loop slot-filling.
Using contexts to carry data between intents.

The final interaction will behave like this:

Participant	Dialog
End-user	Check my order.
Agent	What is your sequence? Please pause after a few characters so I can confirm as we go.
End-user	12ABC
Agent	12ABC. Okay, continue.
End-user	34DE
Agent	12ABC34BE. Okay, continue.
End-user	No that's not right.
Agent	Let's try again. What comes after 12ABC ?
End-user	34DE
Agent	12ABC34DE. Okay, continue.
End-user	That's it.

Finally, the webhook validates the full sequence and proceeeds with the rest of your flow.

Create a sequence recognizer entity