Share via

Chat With Your Data

Browse code

Open in GitHub Codespaces

This sample showcases Chat With Your Data is a powerful Teams app template designed to facilitate dynamic interactions with various data sources through an intuitive conversational interface. Utilizing Azure OpenAI's language models and retrieval-augmented generation (RAG) technology, users can seamlessly connect to data stored in Azure AI Search, Azure Blob Storage, and more, allowing for insightful and context-rich exchanges that enhance decision-making and data accessibility within Microsoft Teams.

  • The user can set up a datasource such as Azure AI Search, Azure Blob Storage, URL/web address, Azure Cosmos DB for MongoDB vCore, uploaded files, or Elasticsearch.
  • The user can then interact with the data through the Teams app template.

Included Features

  • Azure Open AI: Azure OpenAI On Your Data leverages OpenAI's language models, like GPT-3.5 and GPT-4, augmented with RAG to provide a conversational interface for interacting with your domain data using the chat completions API.
  • Azure AI Search: Azure AI Search is utilized as a data source for Azure OpenAI's chat completions API.
  • Azure Blob Storage: The uploaded files are stored in a container in Azure Blob Storage and indexed by Azure AI Search.

Interaction with the app

bot-in-action

Set up and run the sample app in Github Codespaces

  1. Click Open in GitHub Codespaces badge above to create a codespace for the sample app. Wait for the codespace to be setup, it may take a couple of minutes.
  2. Using the Microsoft 365 Agents Toolkit extension, sign in to your Microsoft 365 account and Azure account under ACCOUNTS.
  3. Set up your data source using Azure AI resources.
  4. Populate the environment files.
  5. Press Ctrl+Shift+D to open the Run and Debug menu. Select Debug and press F5 or click on the play button.
  6. Download the zip file appPackage/build/appPackage.local.zip and upload the app to Teams.
  7. You can now run the sample app.

Set up and run the sample app locally

Prerequisites

  1. Clone the repository git clone https://github.com/t-mbindra/chat-with-your-data.git
  2. Open the project folder(Microsoft-Teams-Samples/samples/msteams-chat-with-your-data) in VS Code.
  3. Run poetry install
  4. Run poetry build
  5. Using the Microsoft 365 Agents Toolkit extension, sign in to your Microsoft 365 account and Azure account under ACCOUNTS.
  6. Set up your data source using Azure AI resources.
  7. Populate the environment files.
  8. Press Ctrl+Shift+D to open the Run and Debug menu. Press F5 or click on the play button.
  9. You can now run the sample.

Note

Check the status of all your local(Debug/F5 flow) bots on Microsoft Bot Framework. Check the status of all your Teams apps on Developer Portal. Microsoft 365 Agents Toolkit will also generate an app registration along with a password which can be seen under App Registrations on the Azure portal.

Set up your data source

  1. Runsh deploy.sh in the terminal. You can use Azure Cloud Shell (Bash) for a Windows device. You will be prompted to login to Azure and select a subscription.
  2. Go to the Azure AI Studio, select relevant subscription and the resource teamsazureopenai-cognitive.
  3. Proceed to the Chat Playground. Click on Add a data source under the Add your data tab.
  4. Upload your data with Upload files or URL/ web address. Ensure you have the correct subscription selected. Selectteamsazureopenai as your storage resource and teamsazureopanai-search as your search resource. Type the index name you want to use.
  5. Add your data and select search type and chunk size. Select API Key as the authentication type. Save and Close and wait for the data to be ingested.

Populate the environment files

  1. You need to populate the environment variables in env/.env.local.user if you are using the Debug or F5 flow. Else, populate the environment variables in env/.env.dev.user if you are deploying the app on Azure.
  2. Go to the the Azure portal and navigate to the resource group ChatWithYourData.
  3. To populate SECRET_AZURE_OPENAI_KEY, SECRET_AZURE_OPENAI_ENDPOINT variables:
    • Go to the teamsazureopenai-cognitive resource. Select the Keys and Endpoints tab under Resource Management. Use Key 1 and Endpoint.
  4. To populate SECRET_AZURE_SEARCH_ENDPOINT variable:
    • Go to the teamsazureopenai-search resource. Use the Url given.
  5. To populate SECRET_AZURE_SEARCH_KEY variable:
    • Go to the teamsazureopenai-search resource. Select the Keys tab under Settings. Use Primary admin key.
  6. Populate AZURE_SEARCH_INDEX based on the index name you chose while setting up the data source.

Upload the app to Teams

  1. Go to your Teams app and click on the Apps icon. Select Manage your apps followed by Upload an app.
  2. Select Upload a custom app and open the relevant zip file. Click on Add when prompted.

If you do not have permission to upload custom apps (uploading), Microsoft 365 Agents Toolkit will recommend creating and using a Microsoft 365 Developer Program account - a free program to get your own dev environment sandbox that includes Teams.

Run the sample app

  1. The bot sends a welcome message. welcome-message
  2. Chat with the bot. chat-with-the-bot
  3. Clear the chat history. clear-conversation-history

Deploy/Host the app on Azure

Instead of the Debug or F5 flow, you can deploy the app on Azure:

  1. Populate the environment files.
  2. Using the Microsoft 365 Agents Toolkit Extension tab, click on Provision followed by DEPLOY under LIFECYCLE. You will be asked to select the subscription and resource group for provisioning.
  3. Download the zip file appPackage/build/appPackage.dev.zip and upload the app to Teams.

Note

Check the status of your dev(hosted on Azure) bots on Azure Portal by navigating to the relevant resource group. Check the status of all your Teams apps on Developer Portal. Microsoft 365 Agents Toolkit will also generate an app registration along with a password which can be seen under App Registrations on the Azure portal.

Further Reading