Create a Slack Summarizer Bot with VESSL Service

Use VESSL Service to create a Slack Summarizer Bot with Open-Source Llama 3

This example is a step-by-step walkthrough of how to create a bot that summarizes Slack threads using the open-source Llama 3 model, all on VESSL Service.

See the final code

See the completed YAML files and final code for this example.

Setting Up The App

Open api.slack.com, and navigate to the Your Apps tab on the top right corner. Create a new app from scratch called “Summarizer” in your designated workspace.

Under the OAuth & Permissions tab, scroll down to Scopes and configure the necessary scopes (such as app_mentions:read, channels:history, channels:join, chat:write, im:history, remote_files:read, users:read) and click Install to Workspace.

Take the Bot User OAuth Token under OAuth & Permissions and the Signing Secret under Basic Information > App Credentials and save them somehwere secure, like the designated VESSL interface Settings > Secrets.

Deploying Llama 3 Using VESSL Service

We will now deploy a Llama 3 server with VESSL Service.

1. Clone this GitHub repository. We will be using the llm-summarizer-slackbot example.

   git clone https://github.com/vessl-ai/examples.git

2. Make sure VESSL is installed and configured.

   pip install vessl
   vessl configure

3. Deploy the server using the provided serve-llama.yaml configuration and VESSL CLI using the following command.

   cd examples/llm-summarizer-slackbot
   vessl serve create -f serve-llama.yaml --set-current-active

Once the service is running, the service will appear in VESSL Service tab, and you an now access the model through the provided endpoint.

Serving the App

To serve the app itself, create a New Service in Service. Then, navigate to the Revisions tab within the service and press New Revision.

1. Initialize the revision with the run-app.yaml from the cloned repository. You will see the exact file on the right. You can now directly edit the YAML file with the editor or edit with the UI.

2. Make sure that your secrets are properly configured (access-token, slackbot-token, and slack-signing-secret).

3. Paste in the endpoint produced by the Llama server in LLAMA3_BASE_URL and press Create.

4. Once the revision is running, the app server will produce an endpoint that you can enter into the Slack Request URL under Event Subscriptions.

Remember to add ‘/slack/events’ at the end of the endpoint when entering the Request URL for the app to properly respond to Slack events

5. Subscribe to bot events (such as app_mention, message.channels, message.im) in Event Subscriptions and save the changes.

The app is now good to go. Mention the bot in a thread and ask it a question or use the summarize command for a summary, --limit to specify the number of messages (default: 10), --help for instructions, and --version to check the app version.