[Feature Request]:

[Drahokma]

Scenario

Hi,
it would be great to use the file_search with vector store defined in the assistants playground. In the playground asking questions based on the file-search works very well. It would be great to connect the predefined vector store to the teams ai assistant planner. So far I was not able to set up the file_search option to be enabled and the assistant in the teams ai library does not use predefined vector store.

import { MemoryStorage, MessageFactory, TurnContext } from "botbuilder";

// See https://aka.ms/teams-ai-library to learn more about the Teams AI library.
import { Application, AI, preview } from "@microsoft/teams-ai";

import config from "../config";

// See README.md to prepare your own Azure OpenAI Assistant
if (!config.azureOpenAIKey || !config.azureOpenAIAssistantId) {
  throw new Error(
    "Missing AZURE_OPENAI_API_KEY or AZURE_OPENAI_ASSISTANT_ID. See README.md to prepare your own Azure OpenAI Assistant."
  );
}

import { resetMessage } from "./messages";
import { httpErrorAction, searchDocuments } from "./actions";

// Create AI components
// Use OpenAI
const planner = new preview.AssistantsPlanner({
   apiKey: config.azureOpenAIKey,
   assistant_id: config.azureOpenAIAssistantId,
   endpoint: config.azureOpenAIEndpoint
});

// Define storage and application
const storage = new MemoryStorage();
const app = new Application({
  storage,
  ai: {
    planner,
  },
});

app.conversationUpdate("membersAdded", async (turnContext: TurnContext) => {
  const welcomeText = "How can I help you today?";
  for (const member of turnContext.activity.membersAdded) {
    if (member.id !== turnContext.activity.recipient.id) {
      await turnContext.sendActivity(MessageFactory.text(welcomeText));
    }
  }
});

app.message("reset", resetMessage);

app.ai.action(AI.HttpErrorActionName, httpErrorAction);
app.ai.action("searchDocuments", searchDocuments);

export default app;

Solution

Adding assistant vector_store_id to the planner. Something like:

const planner = new preview.AssistantsPlanner({
   apiKey: config.azureOpenAIKey,
   assistant_id: config.azureOpenAIAssistantId,
   endpoint: config.azureOpenAIEndpoint,
   vectorstoreId: config.vectorstoreId
});

Additional Context

No response