OpenAI Chat Streaming with RAG + AI Search in JavaScript

This page was automatically generated by AI; not yet reviewed for accuracy...

The content and code samples on this page were generated by using the ai CLI with customized prompts in this repository.

It's cool, but, it's experimental. 😁

Please review the content and code before using it in your application.

This sample demonstrates how to use the OpenAI Chat API with RAG (Retrieval-Augmented Generation) and AI Search in a JavaScript console application.

Main.js
OpenAIChatCompletionsStreamingWithDataClass.js

How to generate this sample

Command

ai dev new openai-chat-streaming-with-data --javascript

Output

AI - Azure AI CLI, Version 1.0.0
Copyright (c) 2024 Microsoft Corporation. All Rights Reserved.

This PUBLIC PREVIEW version may change at any time.
See: https://aka.ms/azure-ai-cli-public-preview

Generating 'openai-chat-streaming-with-data' in 'openai-chat-streaming-with-data-js' (3 files)...

OpenAIChatCompletionsStreamingWithDataClass.js
ReadLineWrapper.js
Main.js

Generating 'openai-chat-streaming-with-data' in 'openai-chat-streaming-with-data-js' (3 files)... DONE!

Main.js

STEP 1: Read the configuration settings from environment variables:

Main.js

const openAIAPIKey = process.env["AZURE_OPENAI_API_KEY"] || "<insert your OpenAI API key here>";
const openAIAPIVersion = process.env["AZURE_OPENAI_API_VERSION"] || "<insert your OpenAI API version here>" ;
const openAIEndpoint = process.env["AZURE_OPENAI_ENDPOINT"] || "<insert your OpenAI endpoint here>";
const openAIChatDeploymentName = process.env["AZURE_OPENAI_CHAT_DEPLOYMENT"] || "<insert your OpenAI chat deployment name here>" ;
const openAIEmbeddingsDeploymentName = process.env["AZURE_OPENAI_EMBEDDING_DEPLOYMENT"] || "<insert your OpenAI embeddings deployment here>" ;
const openAIEmbeddingsEndpoint = `${openAIEndpoint.replace(/\/+$, '')}/openai/deployments/${openAIEmbeddingsDeploymentName}/embeddings?api-version=${openAIAPIVersion}`;
const openAISystemPrompt = process.env["AZURE_OPENAI_SYSTEM_PROMPT"] || "You are a helpful AI assistant." ;
const searchEndpoint = process.env["AZURE_AI_SEARCH_ENDPOINT"] || "<insert your search endpoint here>" ;
const searchAPIKey = process.env["AZURE_AI_SEARCH_KEY"] || "<insert your search api key here>" ;
const searchIndexName = process.env["AZURE_AI_SEARCH_INDEX_NAME"] || "<insert your search index name here>" ;

STEP 2: Initialize the helper class with the configuration settings:

Main.js

const chat = new OpenAIChatCompletionsStreamingWithDataClass(openAIEndpoint, openAIAPIKey, openAIChatDeploymentName, openAISystemPrompt, searchEndpoint, searchAPIKey, searchIndexName, openAIEmbeddingsEndpoint);

STEP 3: Obtain user input, use the helper class to get the assistant's response, and display responses as they are received:

Main.js

while (true) {

  const input = await readline.question('User: ');
  if (input === 'exit' || input === '') break;

  let response = await chat.getChatCompletions(input, (content) => {
    console.log(`assistant-streaming: ${content}`);
  });

  console.log(`\nAssistant: ${response}\n`);
}

console.log('Bye!');
process.exit();

OpenAIChatCompletionsStreamingWithDataClass.js

STEP 1: Create the client and initialize chat message history with a system message:

OpenAIChatCompletionsStreamingWithDataClass.js

constructor(openAIEndpoint, openAIAPIKey, openAIChatDeploymentName, openAISystemPrompt, searchEndpoint, searchAPIKey, searchIndexName, openAIEmbeddingsEndpoint) {
  this.openAISystemPrompt = openAISystemPrompt;
  this.openAIChatDeploymentName = openAIChatDeploymentName;
  this.client = new OpenAIClient(openAIEndpoint, new AzureKeyCredential(openAIAPIKey));

  this.azureExtensionOptions = {
    azureExtensionOptions: {
      extensions: [
        {
          type: "AzureCognitiveSearch",
          endpoint: searchEndpoint,
          key: searchAPIKey,
          indexName: searchIndexName,
          embeddingEndpoint: openAIEmbeddingsEndpoint,
          embeddingKey: openAIAPIKey,
          queryType: "vectorSimpleHybrid"
        },
      ],
    }
  }

  this.clearConversation();
}

STEP 2: When the user provides input, add the user message to the chat message history:

OpenAIChatCompletionsStreamingWithDataClass.js

clearConversation() {
  this.messages = [
    { role: 'system', content: this.openAISystemPrompt }
  ];
}

STEP 3: Send the chat message history to the streaming OpenAI Chat API and process each update:

OpenAIChatCompletionsStreamingWithDataClass.js

async getChatCompletions(userInput, callback) {
  this.messages.push({ role: 'user', content: userInput });

  let contentComplete = '';
  const events = await this.client.streamChatCompletions(this.openAIChatDeploymentName, this.messages, this.azureExtensionOptions);

  for await (const event of events) {
    for (const choice of event.choices) {

      let content = choice.delta?.content;
      if (choice.finishReason === 'length') {
        content = `${content}\nERROR: Exceeded token limit!`;
      }

      if (content != null) {
        if(callback != null) {
          callback(content);
        }
        await new Promise(r => setTimeout(r, 50)); // delay to simulate real-time output, word by word
        contentComplete += content;
      }
    }
  }

  this.messages.push({ role: 'assistant', content: contentComplete });
  return contentComplete;
}