Teams OpenAI Conversation Bot

Summary

This is a sample example OpenAI bot for Microsoft Teams. It allows a user in Teams to send a chat message to the Bot, which then uses OpenAI API to get an answer to the chat message.

Please note: Currently there is no publicly available API for OpenAI's ChatGPT. This sample uses the OpenAI API, which models doesn't currently include ChatGPT. The OpenAI API is currently in beta, and is not free. You can sign up for a free account here. For a list of supported models, see here.

Frameworks

Prerequisites

OpenAI account and API key
Office 365 tenant
Node.js version 14 or higher
```
# determine node version
node --version
```

Azure Functions Core Tools version 4 or higher

# determine core tools version
func --version

ngrok - Although a free account will work with this sample, the tunnel subdomain will change each time you run ngrok, requiring a change to the Azure Bot messaging endpoint and the Teams app manifest. A paid account with a permanent subdomain is recommended.

Version history

Version	Date	Author	Comments
1.0	Jan 3, 2023	Lee Ford	Initial release

Disclaimer

THIS CODE IS PROVIDED AS IS WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.

Minimal Path to Awesome

Start ngrok

Start ngrok listening on port 7071 (or whatever port you are running the Azure Function on)

ngrok http 7071

If you have a paid account, add the subdomain:

# Replace 12345678 with your ngrok subdomain
ngrok http 7071 -subdomain=12345678

Take a note of the forwarding URL, as you will need it later.

Create an Azure Bot

Go to the Azure Portal
Search for Azure Bot and choose Create
Populate the bot with the following:
- Bot handle: A unique bot name (the display name can be changed later)
- Type of App: Single tenant (in a production scenario, you would be running in Azure and use Managed Identity)
With the bot created, under Configuration:
- Take a note of the Microsoft App ID and Microsoft App Tenant ID (you will need these later)
- Populate the Messaging endpoint with the ngrok forwarding URL (with /api/messages appended). Note: if you are not using a paid account, you will need to update this each time you restart ngrok.
- Click on Manage Password and create a New client secret. Specify a name and when it expires. Take a copy of the Secret value (you will need this later)
Note in a production environment, you should be using a Managed Identity to authenticate to the Azure Bot, rather than a client secret.
Under Channels, add the Microsoft Teams channel and enable messaging

Run Locally

Clone this repository

Create and populate a local.settings.json file in the root folder with the following (with your own values):

{
"IsEncrypted": false,
"Values": {
    "FUNCTIONS_WORKER_RUNTIME": "node",
    "AzureWebJobsStorage": "",
    "OPENAI_API_KEY": "<YOUR OPENAI API KEY>",
    "OPENAI_MODEL": "text-davinci-003", // Or whatever model you want to use (see https://beta.openai.com/docs/api-reference/models)
    "MicrosoftAppId": "<YOUR MICROSOFT APP REGISTRATION ID>",
    "MicrosoftAppPassword": "<YOUR MICROSOFT APP REGISTRATION CLIENT SECRET>",
    "MicrosoftAppTenantId": "<YOUR MICROSOFT APP REGISTRATION TENANT ID>",
    "MicrosoftAppType": "SingleTenant"
    }
}

Run the following to install, build and run the code (from the root folder):
```
npm install
npm run build
func host start
```

Teams App Manifest

Edit the manifest.json file and replace the id and botId values with the Microsoft App ID from the Azure Bot
Zip the manifest folder into a manifest.zip file
Upload the manifest.zip file to Teams (side load the or upload in Teams Admin Center)

hanfeijp / teams-openai-bot