Yext Chat

Yext Chat represents a sophisticated enterprise conversational AI platform that integrates sophisticated AI capabilities with structured business data. This technical framework supports seamless front-end and back-end integration across various applications while delivering highly personalized and accurate responses through its sophisticated context management and hybrid retrieval-generation approach. The platform's capabilities span comprehensive enterprise use cases in support, marketing, commerce, and workplace applications, demonstrating the significance of strategic AI implementation in modern business communications.

Integration and Implementation

Yext Chat offers robust integration capabilities that enable seamless implementation across various applications and platforms. Front-end integration is achieved through the Yext Chat SDK, which allows developers to initialize the chat widget using simple JavaScript methods. For React applications, the platform provides the useChatActions hook, simplifying the process of setting context and managing chat interactions.

The framework supports multiple context-setting mechanisms, including direct assignment via window.ChatApp.setContext and the chat.setContext method. This flexibility allows developers to store and reference a wide range of contextual information within chat interactions. For instance, user-specific data can be referenced using [[context.userID]] to fetch relevant user history, while location-specific data can be accessed through [[context.locationId]].

Back-end integration relies on the headless SDK, which offers comprehensive tools for managing chat context. A practical implementation in React demonstrates how these tools can be utilized:

import { useEffect } from "react";

import { useChatActions } from "@yext/chat-headless-react";

import { ChatPanel } from "@yext/chat-ui-react";

function MyApp() {

const chat = useChatActions();

useEffect(() => {

<pre><code>chat.setContext({ userId: "1234", businessId: "3472542" });
</code></pre>

}, []);

return <ChatPanel />;

}

For location-specific chat functionality, the platform employs a structured approach that combines front-end context setting with back-end data retrieval. When setting context, developers can use either window.ChatApp.setContext or the equivalent chat SDK method. The system then utilizes this context to fetch relevant data from Yext's Content API, as demonstrated in the following bot configuration:

{

"$id": "yext-marketing-bot",

"$schema": "https://schema.yext.com/config/chat/chat-bot/v1",

"name": "Location Bot",

"initialMessage": "Hi! This is the bot for Bob's Burgers. How can I help you?",

"goals": {

<pre><code>"answer-location-question": {

  "goal": "Answer a question about Yext.",

  "instructions": [

    {

      "restApi": {

        "method": "GET",

        "url": "https://cdn.yextapis.com/v2/accounts/me/content/my-content-endpoint/[[context.locationId]]?apiKey=<YOUR_API_KEY_HERE>"

      }

    },

    {

      "reply": {

        "instruction": "Based on the data you found, answer"

      }

    }

  ]

}
</code></pre>

}

}

This configuration illustrates how Yext Chat integrates with external systems to deliver location-specific information, with context data serving as the bridge between front-end interactions and back-end responses.

Enterprise Use Cases

Yext Chat applications span four key enterprise areas: support, marketing, commerce, and workplace.

In support, the platform enables self-service through integrated help articles, deflects unnecessary service tickets, and equips live agents with internal chat tools. The system's ability to pull content from multiple sources ensures that all support interactions are backed by accurate business information.

For marketing, Yext Chat drives leads through direct brand communication, promotes events by answering marketing-specific questions, and collects valuable user information for segmentation and targeting. The platform's comprehensive content retrieval capabilities enable it to provide detailed, data-driven responses that support both sales and service objectives.

In commerce, the tool assists with product recommendations, answers availability and pricing questions, tracks shipments, and facilitates returns and exchanges. The integration with Yext's Content API allows chat responses to be closely aligned with the most up-to-date commerce information available within an organization.

The workplace application enables employees to ask a wide range of questions directly through chat. This function supports authentication for sensitive content access and automates HR processes such as changing contact information or requesting time off. All workplace interactions maintain security and compliance by ensuring that chat responses are based on the most current internal information available.

These diverse applications demonstrate Yext Chat's versatility in enterprise settings, with each deployment building on the company's foundational approach to integrating conversational AI with structured business data.

AI Technology and Development

Yext's conversational AI platform builds on the company's existing AI and search technologies, incorporating multiple large language models including their proprietary Knowledge Graph. The platform's approach addresses challenges in generating accurate, business-specific responses by ensuring that generated content aligns with structured, authoritative information.

At the core of Yext Chat's technology is the company's Knowledge Graph, a headless content management system that organizes content using entity relationships optimized for voice search and chatbot interactions. This system serves as the foundation for storing and retrieving content that powers the chatbot's responses, ensuring accuracy and real-world relevance. To support content management and refinement, the platform includes features like Search Term Clustering, which helps businesses identify content gaps and optimize their Knowledge Graph.

The development approach emphasizes flexibility and integration capabilities. Since 2019, Yext has been integrating AI and large language models into its broader Answers Platform infrastructure. While the platform supports multiple language models, including OpenAI's GPT-3 and its own proprietary models, it maintains model-agnostic flexibility to allow customers to use any language model that best fits their specific needs.

Yext's implementation strategy emphasizes both frontend and backend integration capabilities. In frontend implementation, the platform provides JavaScript SDKs for initializing chat widgets and React hooks for managing context. For backend integration, the headless SDK offers tools for setting and managing chat context, demonstrated through structured configuration examples that illustrate how the platform can fetch and incorporate specific content elements like location data. These integrations enable the platform to maintain a consistent application programming interface (API) structure while providing the flexibility needed for diverse enterprise applications.

Context and Data Management

Yext Chat manages context data through a structured approach that enables personalized responses while maintaining flexibility across various applications. The platform supports multiple methods for setting context, including direct JavaScript assignment and React hooks, allowing developers to store and reference user-specific, location-specific, and category/Product information.

The system utilizes bracket notation within bot instructions to reference context data, enabling the retrieval of relevant information for personalized responses. For example, [[context.userID]] allows the chatbot to fetch user order history, while [[context.locationId]] enables location-specific question answering. This context management approach enables the platform to deliver accurate and relevant information across diverse use cases.

The underlying technology employs retrieval augmented generation, combining structured content from Yext's Knowledge Graph with natural language processing capabilities from multiple large language models. This hybrid approach enables the platform to deliver business-specific answers while maintaining accuracy through integration with structured content sources.

For development and implementation, Yext provides both frontend and backend integration capabilities. The frontend allows initialization of chat widgets using the Yext Chat SDK and React hooks for context management. Backend integration is facilitated through the headless SDK, which provides the setContext hook for managing chat context across applications. These integrations enable consistent implementation while providing the flexibility needed for diverse enterprise applications.

API and Functionality

The platform employs Deno runtime with TypeScript support, incorporating Zod for type definitions. Chat functions receive a structured payload containing messages (including text and timestamp fields), notes (with currentGoal, currentStepIndices, searchQuery, queryResult, collectedData, and goalFirstMsgIndex), and context (additional metadata passed to the Chat API).

For example, a function for fetching weather data requires a city parameter, which must be provided in the notes.collectedData. The function constructs an API request URL with the city and API key, sends the request, and returns the API response in the queryResult field. This design provides the underlying structure for implementing custom functionality while maintaining consistency across the platform.

The API allows functions to access the same data as the Chat API, including user context and conversation history. For instance, a function might reference user information using [[context.userID]], location data with [[context.locationId]], or specific categories/products with [[context.category]] or [[context.product]]. This flexible context management approach enables the platform to handle diverse use cases while maintaining consistency in data access patterns.