Slackbot for Incident Classification using Amazon Bedrock (Claude 3 Haiku)

Overview

This project implements an intelligent Slackbot that listens to messages in Slack (such as user-reported incidents), classifies them using Claude 3 Haiku on Amazon Bedrock, and responds with a suggested category (e.g., Network Issue, Application Bug, Security Alert).

Built with:

• AWS Lambda (serverless backend)
• API Gateway (Slack Event URL endpoint)
• Slack Events API
• Amazon Bedrock (Claude 3 Haiku via Messages API)
• Python (runtime + Bedrock client)

Use Case

Why this project?

• Many DevOps and SRE teams rely on Slack for reporting incidents.
• Classifying and routing issues manually slows down triage.
• This bot automates classification and improves response times using LLMs (Claude 3)

Architecture

Slack → API Gateway → Lambda → Claude 3 (Amazon Bedrock) ↘ Slack reply via chat.postMessage

1. User sends a message in Slack mentioning the bot

2. Slack sends the event to API Gateway → Lambda

3. Lambda cleans the text, classifies it using Claude 3 via Bedrock Messages API

4. The bot responds in Slack with a classification

Tech Stack

Layer	Technology / Service	Description
Chat Interface	Slack (Bot + Events API)	Handles message events and replies
Compute	AWS Lambda (Python 3.10+)	Serverless function for classification logic
API Gateway	HTTP API (proxy integration, payload v2)	Exposes Lambda to Slack securely
AI/ML Engine	Amazon Bedrock - Claude 3 Haiku	Classifies incident messages
Auth & Security	Slack OAuth + Signing Secret	Verifies Slack signatures and tokens
Observability	AWS CloudWatch	Logs, metrics, and debugging

Features

• Slackbot listens to @mentions
• Classifies incident messages into categories:
  • Network Issue
  • Application Bug
  • Security Alert
  • Other
• Replies directly in Slack
• Deduplicates Slack retries
• Handles Bedrock model errors gracefully
• Lightweight, serverless, and low latency (~1–2s)

Getting Started

1. Clone this repository

git clone https://github.com/sandhya19/slackbot-bedrock.git
cd slackbot-bedrock

2. Set Up Slack App

• Create a new Slack app
• Enable Event Subscriptions
• Subscribe to:
  • app_mention
• Add OAuth scopes:
  • chat:write
  • app_mentions:read
• Install app to your workspace
• Get SLACK_BOT_TOKEN and SLACK_SIGNING_SECRET

3. Deploy Lambda Function

• Create a Lambda with Python 3.10+
• Add environment variables:
  • SLACK_BOT_TOKEN
  • SLACK_SIGNING_SECRET
• Set IAM policy to allow bedrock:InvokeModel
• Connect via API Gateway (proxy integration, payload v2)

4. Grant Bedrock Access

• Use modelId: anthropic.claude-3-haiku-20240307-v1:0
• Opt-in via AWS Console → Bedrock → Model access

Directory Structure

.
├── lambda_function.py       # Main Lambda handler
├── requirements.txt         # (Optional: for packaging)
├── test_event.json          # Sample Slack event payload
├── README.md                # This file

Sample Message & Bot Response

Input:

@chatbot the application is not responding

Response:

Incident classified as: *Based on the incident message provided, "the application is not responding, this would be classified as an Application Bug.
The message indicates that a specific application is not functioning properly, which suggests an issue with the application itself rather than a network problem or a security alert.*

Demo

Here's a sample interaction with the bot in Slack:

License

MIT License