Slack help bot

A tool to make it easier for Platform users to get help. With a focus on sign-posting, tooling and documentation to allow users to help themselves without needing an engineers help.

Features

• Users can request help from Platform Operations by:
  • Using the /PlatOps help shortcut.
  • Messaging the PlatOps help bot and saying help.
• While requesting help the bot will:
  • Provide initial guidance, linking to documentation, recent announcements and providing guidance on what should be raised here.
  • Link to our QnA maker bot Plato which has pre-programmed answers to some common questions.
  • Ask users to fill in some details about their request.
  • Search the help-requests index in Azure AI Search which will return the top 3 most relevant results from previous requests.
  • Search the the-hmcts-way index in Azure AI Search which will return the top 3 most relevant results from the HMCTS Way.
  • Send the data to Azure AI Services to determine which area, environment and team the request is likely about and will preselect these fields for the user.
  • Create a ticket in Jira with the data provided.
  • Post the request in the #platops-help channel.
• Help request threads support these commands by messaging @PlatOps help <command>:
  • help - list of all available commands
  • duplicate <jira ticket id> - mark a request as a duplicate of another ticket
  • summarise - an AI will summarise all replies in the Slack thread into one message
• On close of a request the bot will ask for what type of help was required and what was done to resolve the issue.
• @PlatOps help home page displays the following reports:
  • Open unassigned help requests
  • Assigned to me
  • Raised by me
• Auto close inactive issues
  • A cron job built into the bot will close any issues that have been not been updated for 10 days
• Analytics on usage of the bot
  • Events are recorded in Application Insights when user actions are taken, for more details see analytics.
  • Labels are added to Jira tickets that can be reported on in Jira to see what team / area / environment is requesting the most help

Architecture

During help request workflow the application:

1. Asks Azure AI services for recommendation for area, environment and team.
2. Searches Azure AI search for similar requests.
3. Searches Azure AI search for anything on the hmcts way that might be relevant.
4. Creates the request in Slack and Jira.
5. Stores the request in Cosmos DB.
6. Replies on the help request Slack thread are added to Jira.
   • Replies on Jira are not added to Slack.

On close of the help request:

1. Status, resolution type and resolution comment added to cosmos DB.

Search index:

1. Search service has an indexer configured to pull new data from Cosmos DB every 5 minutes.
   • Only certain fields are configured, see the indexer configuration.
2. Knowledge store data is stored in Azure Blob Storage, which is uploaded to by a GitHub action in hmcts/hmcts.github.io.
   • The GitHub action then triggers the indexer to update the knowledge store index.

AI summarising:

1. If a user requests a help request to be summarised, all comments are retrieved and sent to AI services for summarisation.

Azure resources:

• All azure resources are created in the components/infrastructure folder.

Getting Started with the Bot

Prerequisites

Running the application requires the following tools to be installed in your environment:

• Node.js v20.0.0 or later
• npm
• Docker
• Azure CLI - locally only

You need to create a Slack App as detailed in the steps below.

Creating the Slack App

For members of Platform Operations we have a test workspace that you can use hmcts-platops-sandbox, ask in #platform-operations for an invitation. You'll be able to install and test changes to your app there without waiting for someone to approve the change.

1. Create a new app in your workspace. Follow the Slack documentation for creating an app from a manifest.

Open a browser and navigate to api.slack.com/apps. This is where we will create a new app with our previously copied manifest details. Click the Create New App button, then select From an app manifest when prompted to choose how you'd like to configure your app's settings.

Next, select a workspace where you have permissions to install apps hmcts-platops-sandbox, then confirm Next. Select the YAML tab and clear the existing contents. Paste the Manifest content below, make sure you update the name:

display_information:
  name: <your name> PlatOps help
  description: Help requests for Platform Operations
  background_color: "#262626"
features:
  app_home:
    home_tab_enabled: true
    messages_tab_enabled: true
    messages_tab_read_only_enabled: false
  bot_user:
    display_name: PlatOps help
    always_online: true
  shortcuts:
    - name: PlatOps Help Request
      type: global
      callback_id: begin_help_request_sc
      description: Request help from Platform Operations
oauth_config:
  scopes:
    bot:
      - app_mentions:read
      - channels:history
      - channels:read
      - chat:write
      - chat:write.customize
      - groups:history
      - groups:read
      - groups:write
      - im:read
      - im:write
      - reactions:write
      - users.profile:read
      - users:read
      - users:read.email
      - im:history
      - commands
settings:
  event_subscriptions:
    user_events:
      - app_home_opened
    bot_events:
      - app_home_opened
      - app_mention
      - message.channels
      - message.im
      - function_executed
  interactivity:
    is_enabled: true
  org_deploy_enabled: true
  socket_mode_enabled: true
  token_rotation_enabled: false
  hermes_app_type: remote
  function_runtime: remote
functions:
  begin_help_request:
    title: Begin Help Request
    description: ""
    input_parameters: {}
    output_parameters: {}

2. Add a custom workflow

Steps from Apps for legacy workflows is now deprecated.

Instead of using the now deprecated Step from App, it is recommended that you re-implement your step as a custom function. Here is some more slack documentation that covers Create a custom step for Workflow Builder.

Below takes you through the basics of how to implement a custom function:

Navigate to Org Level Apps in the left nav and click Opt-In, then confirm Yes, Opt-In.

Navigate to Workflow Steps in the left nav and click Add Step. This is where we'll configure our step's inputs, outputs, name, and description.

For illustration purposes, we're going to write a custom step called Begin Help Request. When the step is invoked, a message will be sent to the provided manager with an option to request some dedicated help from Platform Operations Help Bot.

Once you have saved your changes click on the Workflow Steps in the left nav will show you that one workflow step has been added! This reflects the function defined in our manifest; functions are workflow steps. Below is an example of what it would look like:

3. Collect Tokens

In order to connect our app here with the logic of our sample code set up locally, we need to gather two tokens, a bot token and an app token.

Bot tokens are associated with bot users, and are only granted once in a workspace where someone installs the app.

App-level tokens represent your app across organizations, including installations and are commonly used for creating websocket connections to your app.

To generate an app token, navigate to Basic Information and scroll down to App-Level Token.

Click Generate Token and Scopes, then Add Scope and choose connections:write. Choose a name for your token and click Generate. Copy that value, save it somewhere accessible, and click Done to close out of the modal.

Next up is the bot token. We can only get this token by installing the app into the workspace.

4. Installing App

Navigate to Install App and click the button to install, choosing Allow at the next screen.

You will then have a bot token. Again, copy that value and save it somewhere accessible.

5. Invite the app in the channel where you would like it to be used in Slack. Navigate to the channel where you want the app to be active. Type /invite @YourAppName in the message box and hit enter. Replace @YourAppName with the actual name of your Slack app.

6. Make a note of the channel ID as this will later be required in the slack-help-bot configuration. You can get the channel ID by right-clicking, 'copy link', and then it will be the bit after archives in the url, e.g. C01APTJAM7D.

Running the application

We use 'Socket mode' so no need to proxy Slack's requests.

Running on Kubernetes

The application is deployed on Kubernetes using the HMCTS nodejs chart. To avoid exposing sensitive data from the configuration above you can add them as secrets from an Azure Key Vault. See the chart-library documentation for further info.

The configuration for the deployed instance can be found in hmcts/cnp-flux-config in the slack-help-bot kustomization.

Running locally

All configuration requirements listed above can be found in the "env.template.txt" file.

Initial setup

Rename "env.template.txt" to ".env" which is gitignored and safe for secrets.

Source into your shell with:

$ set -o allexport; source .env; set +o allexport

Install dependencies by executing the following command:

$ npm install

Azure connection

The AI features of the bot are powered by a number of Azure services. To connect to these services you will need to authenticate with Azure. If you are in the DTS Platform Operations Security Group you will have the permissions needed to use the services.

To authenticate with Azure, run the following command:

az login

See more information on authenticating with Azure in a local development environment.

If you aren't in the DTS Platform Operations group you will need at least the following permissions on their respective resources:

• Cosmos DB Built-in Data Contributor - This is a cosmos specific permission and can't be assigned using the Azure Portal
• Cognitive Services OpenAI User
• Search Index Data Reader

Starting

Run:

npm start

Running locally with Docker

There is no need to source your configuration. The ".env" file will be loaded automatically.

Create docker image:

docker compose build

Login to Azure

We are using the azure-cli-credentials-proxy to re-use your local access token without having to pass credentials to the container.

Follow the same instructions as in Azure connection to authenticate with Azure.

Start the application

Run the application by executing the following command:

docker compose up

This will start the frontend container exposing the application's port (set to 3000 in this template app).

In order to test if the application is up, you can visit https://localhost:3000/health in your browser. You should get a very basic health page.

Analytics

The bot uses Application Insights to record events when users interact with the bot. The resource is called slack-help-bot-ptl.

Here are a couple of useful queries:

Pie chart query in Log Analytics

customEvents
| summarize event_count = count() by name
| render piechart

Column chart query in Log Analytics

customEvents
| summarize event_count = count() by bin(timestamp, 1d), name
| render columnchart

Deploying

This application is deployed with continuous delivery, every merge to the main branch will be automatically deployed. The GitHub action will build the docker image and push it to the Azure Container Registry.

The deployment configuration can be found in the hmcts/cnp-flux-config repository.

The infrastructure from the architecture diagram is created using Terraform in the components/infrastructure folder. The terraform pipeline is run from Azure DevOps, a plan is run on a pull request and it will automatically apply on merge to main.

Adding new fields to dropdowns

If you want to add new fields to dropdowns like the area or resolution type the easiest way is to search for the text of another option in the codebase and then add your new option there. They are normally sorted in alphabetical order although environments are generally in increasing order of importance and Other is normally last.

For most dropdowns you will also need to update the LLM prompt to allow it to suggest the new option, resolution type doesn't need to be added there.

Dropdowns use the optionBlock function, it can take either one or two arguments, the first argument is the display name and the second is the label that will be used in Jira. Certain characters can't be used in Jira labels so if you have a complex display name then supply a simpler label or if a team is commonly known by a short name it is common to use the short name in the label as well.

e.g.

optionBlock("GitHub");

or

optionBlock("Security Operations / Secure Design", "security");