Create a bot for MS Teams.
The Bot follows a communications protocol supported by the Azure Bot Framework
- Refector the typing to replace List to list (and equivalent for other types) aligning to modern python (3.8+)
- Consider to put MCPObjects inside LLMConversationHandler
https://mermaid.js.org/syntax/sequenceDiagram.html
Start a conversation
sequenceDiagram
participant Teams as MS Teams Client
participant Bot as Bot Framework
participant Service as Bot Service
participant LLM as LLM Service
participant Tools as Tool Registry
Teams ->> Bot: POST /v3/conversations (Start chat)
Bot ->> Service: Activity.conversationUpdate
Service ->> Bot: 200 OK
Note over Teams,Bot: User sends message
Teams ->> Bot: POST activities (message)
Bot ->> Service: Activity.message
loop Until Final LLM Response
Service ->> LLM: Send user message and history
LLM ->> Service: LLM response
alt Requires Tool Execution
Service ->> Tools: Request tool execution
Tools ->> Tools: Execute requested tools
Tools ->> Service: Return tool results
else Direct Response
Note over LLM,Service: No tools needed. Break loop
end
end
Service ->> Bot: POST activity (response)
Bot ->> Teams: Deliver message to user
flowchart TB
subgraph WebAppDict
config
mcpobjects
llmHandler
end
subgraph webapp
on_startup --> running
subgraph on_startup
connect_to_mcp_server -- create --> mcpobjects
mcpobjects -- use --> bind_tools_when_ready
llmHandler -- use --> bind_tools_when_ready
end
end
subgraph app_init
hams_app_create --> mcp_app_create
mcp_app_create --> service_app_create
mcp_app_create -- init --> connect_to_mcp_server
service_app_create --> azure_app_create
azure_app_create --> langchain_app_create
subgraph langchain_app_create
mcpobjects -- use --> init
end
langchain_app_create -- create --> llmHandler
langchain_app_create -- init --> bind_tools_when_ready
subgraph bind_tools_when_ready
register_tools --> bind_tools
end
end
app_init --> webapp
Bots are software applications that interact with users through text-based conversations. They can range from simple scripts to advanced AI systems that learn and adapt. In Microsoft Teams, bots can act as virtual assistants to automate tasks, provide customer support, and more.
- Notification Bots: Send alerts, reminders, or updates to users in Teams channels or chats.
- Workflow Bots: Automate and streamline business processes by managing tasks and workflows.
- Command Bots: Respond to simple commands using Adaptive Cards for UI.
- Activity Handler: Manages events or activities triggered by user interactions.
- Bot Logic: Defines decision-making rules and conditions for bot responses.
- Bot Scope: Determines how the bot interacts with users (e.g., personal chat, group chat, or channel).
- Use the Bot Framework SDK or Teams AI Library with Teams Toolkit to build bot capabilities.
- AI-powered bots can leverage custom AI models for advanced features like natural language understanding.
To enable the bot to interact with the Microsoft Bot Framework, you need to set up Azure components and generate API tokens. Follow these steps:
Before you start creating the Azure components you need to install terraform
brew install azure-cli
az login
brew tap hashicorp/tap
brew install hashicorp/tap/terraform
Then you can try it out with
make terraform-init
- Log in to the Azure Portal.
- Navigate to Azure Bot Services and create a new bot or select an existing one.
- During the bot creation process, you will be prompted to provide the following details:
- Bot Handle: A unique name for your bot.
- Messaging Endpoint: The URL where your bot will receive messages (e.g.,
https://<your-domain>/api/messages). - App Service Plan: Choose a pricing tier.
- Once the bot is created, navigate to the Settings page to find the App ID and App Secret.
To authenticate your bot with the Microsoft Bot Framework REST API, you need to generate an API token using the App ID and App Secret.
- Use the following OAuth2 endpoint:
https://login.microsoftonline.com/botframework.com/oauth2/v2.0/token
- Make a POST request to the endpoint with the following form data:
grant_type=client_credentials client_id=<Your App ID> client_secret=<Your App Secret> scope=https://api.botframework.com/.default - The response will include an
access_token. Use this token in theAuthorizationheader for API requests.
curl -X POST \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials" \
-d "client_id=<Your App ID>" \
-d "client_secret=<Your App Secret>" \
-d "scope=https://api.botframework.com/.default" \
https://login.microsoftonline.com/botframework.com/oauth2/v2.0/tokenOnce you have the API token, update your bot's configuration to use it for authenticated API calls. You can pass the token as a command-line argument or store it securely in an environment variable.
Prepare a referese proxy for using with MS Teams.
Ngrok seems like it provides the relevant reverse proxy capabaility to project to dev laptop.
brew install ngrok
curl --header "Content-Type: application/json"
--request POST
--header "Content-Type: application/json"
--data '{"client_id":"2abf7c88-8055-4870-8b5b-8b35d6cbfcfb","client_secret":"$APP_SECRET","grant_type":"client_credentials"}'
https://login.microsoftonline.com/1fd80b61-a805-4a57-879b-45ddb39a660d/oauth2/v2.0/token
curl -X POST
-H "Content-Type: application/x-www-form-urlencoded"
-d "grant_type=client_credentials&client_id=2abf7c88-8055-4870-8b5b-8b35d6cbfcfb&client_secret=$APP_SECRET&scope=https://api.botframework.com/.default"
https://login.microsoftonline.com/1fd80b61-a805-4a57-879b-45ddb39a660d/oauth2/v2.0/token
https://informally-large-terrier.ngrok-free.app/api/messages
curl -k -X POST https://login.microsoftonline.com/botframework.com/oauth2/v2.0/token -d "grant_type=client_credentials&client_id=APP_ID&client_secret=APP_PASSWORD&scope=https%3A%2F%2Fapi.botframework.com%2F.default"
SO trying
curl -X POST -d "grant_type=client_credentials&client_id=2abf7c88-8055-4870-8b5b-8b35d6cbfcfb&client_secret=$APP_SECRET$&scope=https%3A%2F%2Fapi.botframework.com%2F.default" https://login.microsoftonline.com/1fd80b61-a805-4a57-879b-45ddb39a660d/oauth2/v2.0/token
NOW trying with new instructions:
Using the app id curl to the auth endpoint. Can use normal tenant OR can use TENANT_ID=botframework.com
curl -X POST \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials" \
-d "client_id=${APP_ID}" \
-d "client_secret=${CLIENT_SECRET}" \
-d "scope=https://api.botframework.com/.default" \
https://login.microsoftonline.com/$TENANT_ID/oauth2/v2.0/token | jq -r '.access_token'THIS does not work
curl -X POST \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials" \
-d "client_id=${BOT_ID}" \
-d "client_secret=${CLIENT_SECRET}" \
-d "scope=https://api.botframework.com/.default" \
https://login.microsoftonline.com/$TENANT_ID/oauth2/v2.0/tokencurl -X POST \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials" \
-d "client_id=${APP_ID}" \
-d "client_secret=${CLIENT_SECRET}" \
-d "scope=https://api.botframework.com/.default" \
https://login.microsoftonline.com/botframework.com//oauth2/v2.0/tokencreate an az identity
az identity create --resource-group "bot-resource-group" --name "polecatworks-identity"
{
"clientId": "9bc178cc-2ea6-429c-a559-6910d64839e5",
"id": "/subscriptions/9d415cc4-075a-47ac-ba34-4328587f07ba/resourcegroups/bot-resource-group/providers/Microsoft.ManagedIdentity/userAssignedIdentities/polecatworks-identity",
"location": "westeurope",
"name": "polecatworks-identity",
"principalId": "2fb044dd-f854-4c56-988a-66bd9cccc7d6",
"resourceGroup": "bot-resource-group",
"systemData": null,
"tags": {},
"tenantId": "1fd80b61-a805-4a57-879b-45ddb39a660d",
"type": "Microsoft.ManagedIdentity/userAssignedIdentities"
}Information on the bot activity: https://github.com/Microsoft/botframework-sdk/blob/main/specs/botframework-activity/botframework-activity.md Spec of types: https://github.com/Microsoft/botframework-sdk/blob/main/specs/botframework-activity/botframework-activity.md#basic-activity-structure
Full walkthrough of registration: https://learn.microsoft.com/en-us/microsoftteams/platform/bots/how-to/authentication/add-authentication?tabs=dotnet%2Cdotnet-sample
JSON_PAYLOAD='{"id":null,"timestamp":null,"localTimestamp":null,"localTimezone":null,"serviceUrl":null,"channelId":null,"from":{"id":"polecatworks-app@4UxUYfL0I8OI8XS0VPFfXL0aJIl2Xi1vA0gestrIK91ktXV64dlfJQQJ99BDACi5YpzAArohAAABAZBS1aGL","name":"polecatworks-app"},"conversation":{"id":"4ApDY79U53qLmjG0xfSzX1-fr","name":null},"recipient":{"id":"ac2b8bda-1018-4748-8c25-3b2aadba056c","name":""},"textFormat":"plain","locale":"en-GB","text":"Hello from the bot!","attachments":[],"entities":[{"type":"ClientCapabilities","requires_bot_state":null,"supports_listening":null,"supports_tts":null}],"channelData":{"clientActivityID":"fdtv09eugi"},"replyToId":"4ApDY79U53qLmjG0xfSzX1-fr|0000000"}'
curl -s -X POST
-H "Authorization: Bearer ${ACCESS_TOKEN}"
-H "Content-Type: application/json"
-d "${JSON_PAYLOAD}"
"${API_ENDPOINT}"
Create an Azure deployment for communicating with an AI. Prepare to set this up to integrate back MS Teams.
flowchart LR
A[Start]-->B[Do you have coffee beans?]
B-->|Yes|C[Grind the coffee beans]
B-->|No|D[Buy ground coffee beans]
C-->E[Add coffee grounds to filter]
D-->E
E-->F[Add hot water to filter]
F-->G[Enjoy!]
https://mermaid.js.org/syntax/sequenceDiagram.html
Start a conversation
sequenceDiagram
API ->> BotService: POST /v3/conversations
BotService ->> API: [200]
The new Message call is
{
"bot": {
"id": "12345678",
"name": "bot's name"
},
"isGroup": false,
"members": [
{
"id": "1234abcd",
"name": "recipient's name"
}
],
"topicName": "News Alert"
}
http://localhost:8080/api/messages
http://localhost:3978/api/messages
curl -v -X POST
-H "Authorization: Bearer $CONV_TOKEN"
-H 'Content-Type: application/json'
-d '{
"locale": "en-EN",
"type": "message",
"from": {
"id": "Ben"
},
"text": "hello"
}'
'https://europe.directline.botframework.com/v3/directline/conversations/1dwhcM1PWhh7YcOnAWecmo-eu/activities'
When trying this i can request a token for a given conversation BUT the conversation ID for the conversation from the conversationUpdate is not found
CONVERSATION_ID=Insm4hllrx5Jo9i20EDPCI-eu
curl -X GET "https://europe.directline.botframework.com/v3/directline/conversations/${CONVERSATION_ID}"
-H "Authorization: Bearer $TOKEN"
curl -v -X POST
-H "Authorization: Bearer $CONV_TOKEN"
-H 'Content-Type: application/json'
-d '{
"type": "message",
"locale": "en-EN",
"from": {
"id": "user1",
"name": "Optional User Name"
},
"text": "HELLO BOT"
}'
'https://europe.directline.botframework.com/v3/conversations/1dwhcM1PWhh7YcOnAWecmo-eu/activities'
curl -X POST
-H 'Content-Type: application/x-www-form-urlencoded'
-d "grant_type=client_credentials&client_id=$APP_ID&client_secret=$TOKEN&scope=https://directline.botframework.com/.default"
https://login.microsoftonline.com/botframework.com/oauth2/v2.0/token
curl -X POST
-H 'Content-Type: application/x-www-form-urlencoded'
-d "grant_type=client_credentials&client_id=$APP_ID$&client_secret=$TOKEN&scope=https://directline.botframework.com/.default"
https://login.microsoftonline.com/botframework.com/oauth2/v2.0/token
curl -X POST
-H "Authorization: Bearer $TOKEN"
-H 'Content-Type: application/json'
'https://europe.directline.botframework.com/v3/conversations'
POST https://directline.botframework.com/v3/directline/conversations Authorization: Bearer SECRET_OR_TOKEN
Create your app secret
kubectl create secret generic dev-chatbot -n chatbot --from-file=container-python/tests/test_data/secrets/bot/app_id --from-file=container-python/tests/test_data/secrets/bot/app_password --from-file=container-python/tests/test_data/secrets/gemini/gcp_llm_key
kubectl create secret docker-registry dockerconfigjson-ghcr-io --docker-server=ghcr.io --docker-username=$GHCR_USERNAME --docker-password=$GHCR_TOKEN
Deploy to your local k8s using helm on the command line
helm upgrade -i dev charts/chatbot -f charts/chatbot-values.yaml
Run the following prompt to test parallel actions and chaining of tools
please add 3 and 5 and also multiple 9 and 6. Take the results of both calculations and multiply them together.