Need Clarification: Dialogflow vs DiscoveryEngine API : AI Applications(i.e Agent builder)

[claw3011]

Hey Guys,

I am currently exploring conversational agents in AI applications and have developed agents that I am accessing locally. Utilizing the Dialogflow CX API to interact with these agents has been successful.

While reviewing the documentation, I encountered the Discovery Engine API and attempted to use it to call the Conversational agents. However, this approach was unsuccessful, although the Discovery Engine API functions effectively for search, assistant & agentspace apps. From my understanding, conversational agents were previously merged into Dialogflow CX, indicating that the Dialogflow API is specifically designed for conversational agents and dialogflow chat applications. In contrast, applications such as recommendation systems and search assistants require the use of the Discovery Engine API.

Could you please confirm if this understanding is correct?

If incorrect, How do we call Conversational agents using Discovery Engine API or Is there a way to call conversational agents through Discovery Engine API?

[ibaui · Accepted solution]

Hi @claw3011,

Welcome to Google Cloud Community!

Your understanding is mostly correct. The Dialogflow CX API is specifically designed for interacting with Dialogflow CX agents and is the appropriate and intended way to manage and communicate with conversational agents. In contrast, the Discovery Engine API is designed for building search, recommendation, and discovery experiences. While it can interact with data and potentially trigger actions, it is not a direct replacement for the Dialogflow CX API when it comes to conversational agents.

Your attempt to call conversational agents using the Discovery Engine API was unsuccessful because this API is not designed to directly invoke or manage Dialogflow CX agents. Its primary focus is on understanding user queries, retrieving relevant information from a knowledge base and presenting that information to the user. It does not handle conversational flow, intent recognition, entity extraction, or fulfillment logic, which are key capabilities of Dialogflow CX.

While you cannot directly call a Dialogflow CX agent using the Discovery Engine API, there are scenarios where they can work together indirectly:

• Using Dialogflow CX to Enhance Discovery Engine Queries: You could use Dialogflow CX to preprocess user queries before sending them to the Discovery Engine. For example:

  1. A user says, "I'm looking for a comfortable pair of running shoes."
  2. Dialogflow CX could identify the intent (find running shoes) and entities (comfortable).
  3. You could then use this structured information to construct a more precise query for the Discovery Engine, improving search results.

• Using Discovery Engine to Inform Dialogflow CX Responses: You could use the Discovery Engine to enrich the responses generated by your Dialogflow CX agent. For example:

  1. A user asks, "What are your best-selling laptops?"
  2. Your Dialogflow CX agent could use the Discovery Engine to retrieve a list of best-selling laptops from your product catalog.
  3. The agent could then present this information to the user in a conversational way.

In these scenarios, the two APIs are used in conjunction, but they are not directly calling each other. Dialogflow CX is either preparing the query for the Discovery Engine or using the Discovery Engine's results to inform its response.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.