msgext-doc-search-js

page_type	description	products	languages
sample	This sample demonstrates how to integrate Azure AI Search in a Teams message extension to enable Hybrid Search (Vector + Semantic) using Teams Toolkit for Visual Studio Code and JavaScript and use the message extension as a plugin in Microsoft Copilot for Microsoft 365.	office-teams copilot-m365	javascript

EcoGroceries Call Center message extension with Azure AI Search sample

This sample implements Azure AI Search (Formerly known as "Cognitive Search") with a Teams message extension that enables Hybrid Search (Vector + Semantic) and can be used as a plugin for Microsoft Copilot for Microsoft 365. Architecture below demonstrates how developers can use Copilot for Microsoft 365 in Bring Your Own Data scenarios instead of building a custom GPT powered bot.

The message extension allows users to query data inside the EcoGroceries Call Center records and returns the most accurate results with the power of Hybrid Search (Vector + Semantic).

Prerequisites

• Node.js 18.x
• Visual Studio Code
• Teams Toolkit
• Azure subscription
• You will need a Microsoft work or school account with permissions to upload custom Teams applications. The account will also need a Microsoft Copilot for Microsoft 365 license to use the extension in Copilot.
• You will need to create Azure AI Search, Blob Storage and Azure OpenAI resources on Azure portal.

Setup and use the sample

This sample requires a manual step to upload documents to Azure AI Search in embeddings format before running the app locally and testing the Hybrid Search (Vector + Semantic) capability with Copilot for Microsoft 365.

Step 1 - Upload documents to Azure AI Search

In this step, you will use Azure OpenAI Studio, Add Your Data capability to upload documents to Azure AI Search in embeddings format.

1. Open the Azure OpenAI resource you created earlier, select Go to Azure OpenAI Studio.
2. In Azure OpenAI Studio, select Deployments, create two new deployment models, select gpt-35-turbo-16k as a model type in the first model and text-embedding-ada-002 for the second model.
3. Select Chat from the menu and select Add your data (preview) under the Assistant setup section, then Add a data source.
4. Select Upload files from the drop-down menu:
   • Select your subscription.
   • Select the Blob Storage and Azure AI Search resources you created in the pre-requisites.
   • Provide an index name.
   • Check the box for Add vector search to this search resource.
   • Select the text-embedding-ada-002 model you created earlier.
   • Check the box for the acknowledgement, then select Next.
   • Drag and drop the files from the documents folder to the Upload files section, then select Upload files. .

Once the documents are successfully uploaded to Azure AI Search index as embeddings, you can test your data in the Chat section by using the following questions:

• "Any customer complaints recently?"
• "Find refund orders"

Step 2 - Run the application locally

Clone or Download the sample repository: https://github.com/OfficeDev/Copilot-for-M365-Plugins-Samples/.

Navigate to the samples/msgext-doc-search-js folder and open with Visual Studio Code.

Navigate to the samples/msgext-doc-search-js/env folder, rename .env.local.sample file to .env.local and .env.local.user.sample file to .env.local.user. In .env.local.user file, provide the following variables:

   AZURE_OPENAI_SERVICE_NAME= the name of the Azure OpenAI resource 
   AZURE_OPENAI_DEPLOYMENT_NAME= the deployment name of the `text-embedding-ada-002` model
   AZURE_OPENAI_API_KEY= the key available under Keys and endpoints on Azure OpenAI resource
   AZURE_SEARCH_ENDPOINT= the endpoint url of Azure AI Search
   AZURE_SEARCH_ADMIN_KEY= the admin key available under Keys on Azure AI Search resource
   AZURE_SEARCH_INDEX_NAME= the index name created when uploading documents

Click F5 to start debugging, or click the start button 1️⃣. You will have an opportunity to select a debugging profile; select Debug in Teams (Edge) 2️⃣ or choose another profile.

A browser window will open and invite you to log in. Once you're in, Microsoft Teams should open up and display a dialog offering to install your application. Select Add to add EcoGroceries Call Center as a personal application.

Test the message extension on Teams chat first before testing it on Copilot for Microsoft 365.

Step 3 - Test the app in Copilot for Microsoft 365

Navigate to the Microsoft Copilot for Microsoft 365 chat. Check the lower left of the chat user interface, below the compose box. You should see a plugin icon. Click this and enable the EcoGroceries Call Center plugin.

For best results, start a new chat by typing "New chat" before each prompt or set of related prompts.

Here are some prompts to try that use only a single parameter of the message extension:

• "Find refund orders in EcoGroceries"

• "Any customer complaints in EcoGroceries?"

• "Find any info about the order no 345678 in EcoGroceries"

• "How was the call between customer agent and Sarah Ramirez in EcoGroceries?"

As you're testing, watch the log messages within your application. You should be able to see when Copilot calls your plugin.

demo-30seconds.mp4