OmniSummary

AI-powered content summarization agent that processes web articles (HTML), YouTube videos, and PDF documents through Slack, delivering intelligent summaries directly to your channels.

Overview

OmniSummary is a production-ready agentic system built on Strands Agents and AWS Bedrock AgentCore. It enables teams to quickly digest technical content by mentioning a Slack bot with a URL, receiving comprehensive summaries with extracted images and reference links.

Key Features

Multi-Format Support: HTML articles, YouTube videos, PDF documents
Intelligent Summarization: Claude Sonnet 4.5 with customizable opening messages
Image Processing: Automatic extraction and analysis using Claude Haiku 3 VLM
Metadata Extraction: Authors, affiliations, publication dates, keywords, reference URLs
Dual Slack Workspace: Support for both personal and business channels
Production Infrastructure: AWS Bedrock AgentCore with event deduplication and signature verification

Use Cases

Research teams tracking arxiv papers and technical blogs
Engineering teams digesting conference talks and tutorials
Product teams reviewing industry reports and whitepapers
Knowledge sharing across personal and business Slack workspaces

Architecture

┌─────────────┐
│ Slack User  │ @your-bot https://example.com/article
└──────┬──────┘
       │
       ▼
┌─────────────────────────────────────────────────┐
│ API Gateway → Lambda (Slack Event Handler)      │
│ - Signature verification                        │
│ - Event deduplication (DynamoDB)                │
│ - Immediate ACK                                 │
└──────┬──────────────────────────────────────────┘
       │ (async invoke)
       ▼
┌─────────────────────────────────────────────────┐
│ AWS Bedrock AgentCore Runtime                   │
│ - Docker container (ECS-like)                   │
│ - Strands Agent orchestration                   │
└──────┬──────────────────────────────────────────┘
       │
       ▼
┌─────────────────────────────────────────────────┐
│ Agent Tools                                     │
│ 1. parse_html / parse_pdf / parse_youtube       │
│ 2. generate_summary                             │
│ 3. send_slack_message                           │
└──────┬──────────────────────────────────────────┘
       │
       ▼
┌─────────────────────────────────────────────────┐
│ Slack Channels                                  │
│ - Summary with images                           │
│ - Reference URLs                                │
└─────────────────────────────────────────────────┘

Technology Stack

Component	Technology
Agent Framework	Strands Agents
Runtime Platform	AWS Bedrock AgentCore
Infrastructure	AWS CDK (Python)
LLM	Amazon Bedrock (Claude Sonnet 4.5)
VLM	Amazon Bedrock (Claude Haiku 3)
PDF Parsing	Unstructured / Upstage Document Parser
Messaging	Slack Bot API
Package Manager	uv

Prerequisites

Before starting, ensure you have:

AWS Account with Administrator access
AWS CLI configured with credentials
Slack Workspace with Admin permissions to create apps
Python 3.12+ installed
uv package manager (installation guide)
Docker for building AgentCore runtime images

Installation

1. Local Setup

# Clone the repository
git clone <your-repo-url>
cd omnisummary

# Install dependencies using uv
uv sync

# Create environment configuration
cp .env.template .env
# Edit .env with your credentials (see below)

# Create config file
cp config/config-template.yaml config/config.yaml
# Edit config.yaml (see Configuration section)

2. Slack App Setup

This section provides detailed instructions for creating and configuring your Slack App.

2.1 Create Slack App

Go to Slack API
Click "Create New App" → "From scratch"
Enter:
- App Name: Choose any name (e.g., "ContentSummary", "DigestBot", "OmniSummary")
- Workspace: Select your workspace
Click "Create App"

Note: Throughout this README, we use @omnisummary as an example. Replace it with your actual bot name.

2.2 Configure OAuth & Permissions

Navigate to "OAuth & Permissions" in the sidebar
Scroll to "Scopes" → "Bot Token Scopes"

Add the following scopes:

app_mentions:read    - Detect when bot is mentioned
chat:write           - Send messages to channels
files:write          - Upload images (PDF thumbnails)
channels:history     - (Optional) Read channel history

Scroll to top and click "Install to Workspace"
Authorize the app
Copy the Bot User OAuth Token (starts with xoxb-)
- Save this as SLACK_PERSONAL_TOKEN or SLACK_BUSINESS_TOKEN in .env

2.3 Enable Event Subscriptions

Important: Complete this step AFTER deploying the AWS infrastructure (Step 3).

Navigate to "Event Subscriptions"
Toggle "Enable Events" to ON
Enter Request URL:
```
https://<your-api-gateway-url>/slack/events
```
You'll get the API Gateway URL from AWS deployment outputs.
Under "Subscribe to bot events", add:
```
app_mention
```
Click "Save Changes"

2.4 Get Channel IDs

Channel IDs are required in your configuration to specify where summaries should be posted.

Method 1: Slack Web/Desktop App

Right-click the channel name
Select "View channel details"
Scroll down and copy the Channel ID (starts with C)

Method 2: Slack API

curl -H "Authorization: Bearer xoxb-YOUR-TOKEN" \
  https://slack.com/api/conversations.list

Multiple Channels: Separate with commas (no spaces)

SLACK_PERSONAL_CHANNEL_IDS=C123ABC456,C789DEF012,C345GHI678

2.5 Get Signing Secret

Navigate to "Basic Information"
Scroll to "App Credentials"
Copy the Signing Secret
Save as SLACK_SIGNING_SECRET in .env

2.6 Invite Bot to Channels

After app installation, invite your bot to channels:

/invite @your-bot-name

The bot must be a member of all channels listed in SLACK_PERSONAL_CHANNEL_IDS and SLACK_BUSINESS_CHANNEL_IDS.

3. Environment Variables Configuration

Edit .env with your credentials:

# Slack Configuration
SLACK_PERSONAL_TOKEN=xoxb-your-personal-workspace-token
SLACK_PERSONAL_CHANNEL_IDS=C123ABC456,C789DEF012
SLACK_BUSINESS_TOKEN=xoxb-your-business-workspace-token  # Optional
SLACK_BUSINESS_CHANNEL_IDS=C345GHI678                    # Optional
SLACK_SIGNING_SECRET=your-signing-secret

# Optional: Upstage Document Parser (for PDF parsing)
UPSTAGE_API_KEY=your-upstage-api-key

# Optional: Langchain Tracing
LANGCHAIN_API_KEY=your-langchain-api-key
LANGCHAIN_TRACING_V2=false
LANGCHAIN_ENDPOINT=https://api.smith.langchain.com
LANGCHAIN_PROJECT=omnisummary

# Logging
LOG_LEVEL=INFO

Slack Configuration Notes:

Personal vs Business Tokens: You can configure separate Slack workspaces
- SLACK_PERSONAL_*: Always used for summaries
- SLACK_BUSINESS_*: Only used when explicitly requested (e.g., "send to business channels")
Channel IDs: Comma-separated, no spaces
Signing Secret: Required for request verification and security

4. Config File Configuration

Edit config/config.yaml:

resources:
  project_name: omnisummary
  stage: dev  # or prod
  profile_name: your-aws-profile  # AWS CLI profile name
  default_region_name: us-west-2
  bedrock_region_name: us-west-2  # Must support Claude models

  # Optional: Use existing VPC
  vpc_id: vpc-xxxxx
  subnet_ids:
    - subnet-xxxxx
    - subnet-yyyyy

  # Enable business Slack workspace
  enable_business_slack_channels: true  # Set to false to disable

content_parser:
  figure_analysis_model_id: anthropic.claude-3-haiku-20240307-v1:0
  metadata_extraction_model_id: anthropic.claude-sonnet-4-5-20250929-v1:0
  pdf_parser_type: unstructured  # or upstage

summarization:
  summarization_model_id: anthropic.claude-sonnet-4-5-20250929-v1:0
  n_thumbnails: 1  # Number of images to extract

agent:
  agent_model_id: anthropic.claude-sonnet-4-5-20250929-v1:0

infrastructure:
  lambda_timeout_seconds: 60
  lambda_memory_mb: 128
  agentcore_image_tag: latest

Important Configuration Notes:

bedrock_region_name: Must be a region where Claude Sonnet 4.5 is available (e.g., us-west-2, us-east-1)
enable_business_slack_channels: Must be true for business workspace features to work
pdf_parser_type: unstructured (free, local) or upstage (API-based, requires key)

5. AWS Deployment

5.1 CDK Bootstrap (First-time only)

# Export AWS profile if using named profile
export AWS_PROFILE=your-aws-profile

# Bootstrap CDK (creates S3 bucket and IAM roles)
uv run cdk bootstrap

5.2 Deploy Foundation Stack

# Deploys: VPC, ECR, DynamoDB, IAM Roles, CodeBuild
uv run cdk deploy OmniSummaryFoundationStack

# Review changes and confirm with 'y'

This stack creates:

ECR repository for AgentCore Docker images
DynamoDB table for event deduplication
CodeBuild project for automatic Docker builds
IAM roles for Lambda and AgentCore

5.3 Build and Push Docker Image

After Foundation Stack deployment, the CodeBuild project automatically triggers to build and push the Docker image to ECR. You can monitor the build:

# Get build project name from CloudFormation outputs
aws codebuild list-builds-for-project \
  --project-name omnisummary-dev-summary-bot-builder

# Check build status
aws codebuild batch-get-builds --ids <build-id>

Manual build (if needed):

# Get ECR repository URI from CloudFormation outputs
aws ecr describe-repositories --repository-names omnisummary-dev-summary-bot

# Login to ECR
aws ecr get-login-password --region us-west-2 | \
  docker login --username AWS --password-stdin <account-id>.dkr.ecr.us-west-2.amazonaws.com

# Build and push
docker build -t omnisummary-dev-summary-bot:latest .
docker tag omnisummary-dev-summary-bot:latest <ecr-uri>:latest
docker push <ecr-uri>:latest

5.4 Deploy Application Stack

# Deploys: AgentCore Runtime, Lambda, API Gateway, SSM Parameters
uv run cdk deploy OmniSummaryApplicationStack

# Save the API Gateway URL from outputs

Important: Copy the API Gateway URL from deployment outputs. You'll need it for Slack Event Subscriptions (Step 2.3).

Example output:

OmniSummaryApplicationStack.ApiEndpoint = https://abcdef1234.execute-api.us-west-2.amazonaws.com/dev

5.5 Update Slack Event Subscription URL

Now go back to Step 2.3 and configure the Event Subscription URL using the API Gateway endpoint:

https://abcdef1234.execute-api.us-west-2.amazonaws.com/dev/slack/events

5.6 Verify Deployment

# Check AgentCore Runtime status
aws bedrock-agentcore list-agent-runtimes

# Check Lambda function
aws lambda list-functions --query "Functions[?contains(FunctionName, 'slack-event-handler')]"

# Test API Gateway
curl -X POST https://<your-api-gateway-url>/slack/events
# Should return 401 (Unauthorized) without valid Slack signature

Usage

Basic Commands

Once deployed and configured, interact with your bot in Slack:

Simple Summarization

@your-bot-name https://example.com/blog-post

The bot will:

Parse the HTML content
Extract metadata (title, authors, date, keywords)
Generate a summary
Post to personal Slack channels

YouTube Videos

@your-bot-name https://youtube.com/watch?v=abcd1234

Extracts transcript and summarizes video content.

PDF Documents

@your-bot-name https://arxiv.org/pdf/2301.12345.pdf

Parses PDF, extracts figures, and generates summary with image thumbnails.

Advanced Features

Send to Business Channels

Add keywords to send summaries to both personal AND business workspaces:

@your-bot-name https://example.com/article 회사 채널에도 보내줘
@your-bot-name https://example.com/article send to business channels
@your-bot-name https://example.com/article 전체 채널에

Keywords detected:

Korean: 회사 채널, 비즈니스 채널, 업무 채널, 전체 채널, 다 보내
English: business channel, work channel, send to all

Requirement: enable_business_slack_channels: true in config.yaml

Custom Opening Message

Start the summary with a specific phrase:

@your-bot-name https://example.com/article AWS 신기능이네요로 시작해줘

The summary will begin with: "AWS 신기능이네요"

Keywords: 로 시작해줘, 로 시작, 부터 시작

Combined Features

@your-bot-name https://example.com/article
흥미로운 아키텍처네요로 시작. 회사 채널에도 공유해줘

This will:

Generate summary starting with "흥미로운 아키텍처네요"
Post to both personal and business channels

Output Format

Summaries include:

🗞️ 2024-01-15에 John Doe, Jane Smith에서 발행한 Article Title의 요약입니다.

[Opening commentary in Gruber-style voice]

📌 왜 여기에 주목해야 하나요?
[Significance and context]

🔄 핵심 아이디어 및 접근 방식
[Key concepts and methodologies]

🛠️ 기술적 심층 분석
[Technical details and implementation]

📊 성과 및 비즈니스 임팩트
[Results and metrics]

📎 참고 링크
- https://github.com/example/repo
- https://arxiv.org/abs/2301.12345

For PDFs, image thumbnails are uploaded separately after the text summary.

Project Structure

omnisummary/
├── agent/                      # Strands Agent definition
│   ├── agent.py               # summarization_agent
│   ├── agent_tools.py         # Tool implementations
│   └── tool_state.py          # Stateful tool execution
├── agent_runtime/             # Bedrock AgentCore entrypoint
│   └── app.py                 # BedrockAgentCoreApp
├── config/                    # Configuration files
│   ├── config-template.yaml
│   └── config.yaml            # User config (gitignored)
├── infrastructure/            # AWS CDK stacks
│   ├── foundation_stack.py    # VPC, ECR, DynamoDB, IAM
│   └── application_stack.py   # AgentCore, Lambda, API Gateway
├── lambda_handlers/           # Lambda functions
│   └── slack_event_handler.py # Slack event processing
├── shared/                    # Shared utilities
│   ├── config.py              # Config models
│   ├── constants.py           # Enums and constants
│   ├── formatters.py          # Slack message formatting
│   ├── models.py              # Data models (Pydantic)
│   └── prompts/               # LLM prompts
│       └── prompts.py
├── tools/                     # Content processing tools
│   ├── content_parser/        # HTML, PDF, YouTube parsers
│   ├── output_handler/        # Slack message sender
│   └── summarizer/            # LLM summarization
├── tests/                     # Unit and integration tests
├── .env.template              # Environment variable template
├── config.yaml                # Main configuration
├── pyproject.toml             # Python project definition
├── Dockerfile                 # AgentCore runtime image
└── README.md                  # This file

Development

Running Tests

# Run all tests
uv run pytest

# Run with coverage
uv run pytest --cov=omnisummary --cov-report=html

# Run specific test file
uv run pytest tests/test_agents.py

Code Quality

# Format code
uv run black .

# Lint
uv run ruff check .

# Type checking
uv run mypy .

Local Testing (Without AWS)

You can test agent logic locally:

from agent import summarization_agent

# Test the agent
response = summarization_agent("https://example.com/article")
print(response)

Note: This requires configuring local .env with Slack tokens and setting up AWS credentials for Bedrock access.

Adding New Tools

Define tool function in agent/agent_tools.py:

@tool
async def my_new_tool(param: str) -> str:
    """Tool description for LLM."""
    # Implementation
    return result

Add tool to agent in agent/agent.py:

tools = [
    generate_summary,
    parse_html,
    parse_pdf,
    parse_youtube,
    send_slack_message,
    my_new_tool,  # Add here
]

Update system prompt to instruct agent when to use the tool

Rebuild and redeploy:

# Trigger CodeBuild
aws codebuild start-build --project-name omnisummary-dev-summary-bot-builder

# After build completes, update AgentCore runtime
uv run cdk deploy OmniSummaryApplicationStack

Troubleshooting

Bot doesn't respond to mentions

Check 1: Bot is invited to channel

/invite @your-bot-name

Check 2: Event Subscription URL is verified

Go to Slack App → Event Subscriptions
URL should show green checkmark

Check 3: Check Lambda logs

aws logs tail /aws/lambda/omnisummary-dev-slack-event-handler-lambda --follow

Check 4: Verify app_mention scope

Slack App → OAuth & Permissions → Scopes
Ensure app_mentions:read is added

Summaries not appearing

Check 1: AgentCore Runtime logs

# Get runtime name
aws bedrock-agentcore list-agent-runtimes

# Get log group
aws logs describe-log-groups --query "logGroups[?contains(logGroupName, 'agentcore')]"

# Tail logs
aws logs tail <log-group-name> --follow

Check 2: Slack tokens are correct

# Verify token works
curl -H "Authorization: Bearer xoxb-YOUR-TOKEN" \
  https://slack.com/api/auth.test

Check 3: Channel IDs are correct

# List channels the bot is member of
curl -H "Authorization: Bearer xoxb-YOUR-TOKEN" \
  https://slack.com/api/conversations.list

Business channels not receiving messages

Check 1: Config setting

# config/config.yaml
resources:
  enable_business_slack_channels: true  # Must be true

Check 2: Mention business keywords

@bot https://example.com/article 회사 채널에도 보내줘

Check 3: Business token and channel IDs are set

# .env
SLACK_BUSINESS_TOKEN=xoxb-...
SLACK_BUSINESS_CHANNEL_IDS=C123ABC456

Docker build fails

Check 1: ECR authentication

aws ecr get-login-password --region us-west-2 | \
  docker login --username AWS --password-stdin <account-id>.dkr.ecr.us-west-2.amazonaws.com

Check 2: CodeBuild logs

aws codebuild list-builds-for-project --project-name omnisummary-dev-summary-bot-builder
aws codebuild batch-get-builds --ids <build-id>

DynamoDB event deduplication issues

If you're seeing duplicate events:

# Check DynamoDB table
aws dynamodb scan --table-name omnisummary-dev-event-deduplication

# Verify TTL is enabled
aws dynamodb describe-time-to-live --table-name omnisummary-dev-event-deduplication

Cost Estimation

Approximate monthly costs for moderate usage (100 summaries/day):

Service	Usage	Cost
Bedrock Claude Sonnet 4.5	~50K tokens/summary × 100	~$15
Bedrock Claude Haiku 3	~10K tokens/image × 100	~$1
Lambda	100 invocations/day	~$0.01
API Gateway	100 requests/day	~$0.01
DynamoDB	On-demand, minimal	~$0.25
AgentCore Runtime	Runtime hours	Variable*
ECR	10 images stored	~$0.10
Total		~$16-20/month

*AgentCore Runtime costs depend on task duration and concurrency. Idle time may incur minimal charges.

Cost Optimization Tips:

Use dev stage for testing (enables aggressive cleanup)
Set n_thumbnails: 1 to minimize VLM API calls
Configure lifecycle rules on ECR (already included)
Use Bedrock on-demand pricing for variable workloads

Security Best Practices

Slack Request Verification: Always enabled via signing secret
Secrets Management: All credentials stored in SSM Parameter Store (encrypted)
Event Deduplication: Prevents replay attacks and duplicate processing
IAM Least Privilege: Separate roles for Lambda and AgentCore with minimal permissions
VPC Isolation: Optional VPC deployment for network isolation
Environment Separation: Use dev/prod stages with separate AWS accounts

Contributing

We welcome contributions! Please follow these guidelines:

Fork the repository
Create a feature branch (git checkout -b feature/amazing-feature)
Follow code standards in CLAUDE.md
Write tests for new functionality
Run code quality checks (black, ruff, mypy)
Commit with descriptive messages
Push and create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Built on Strands Agents by Anthropic
Uses AWS Bedrock AgentCore for production deployment
Inspired by Daring Fireball's analytical writing style
PDF parsing powered by Unstructured and Upstage

Support

For issues and questions:

Open an issue on GitHub
Check existing issues for solutions
Review CloudWatch Logs for debugging

Note: Replace @your-bot-name with your actual Slack bot display name throughout usage examples.

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
agent		agent
agent_runtime		agent_runtime
assets		assets
config		config
infrastructure		infrastructure
lambda_handlers		lambda_handlers
scripts		scripts
shared		shared
tests		tests
tools		tools
.dockerignore		.dockerignore
.env.template		.env.template
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
Dockerfile		Dockerfile
LICENSE		LICENSE
README.md		README.md
cdk.json		cdk.json
pyproject.toml		pyproject.toml
uv.lock		uv.lock

License

bits-bytes-nn/omnisummary

Folders and files

Latest commit

History

Repository files navigation

OmniSummary

Overview

Key Features

Use Cases

Architecture

Technology Stack

Prerequisites

Installation

1. Local Setup

2. Slack App Setup

2.1 Create Slack App

2.2 Configure OAuth & Permissions

2.3 Enable Event Subscriptions

2.4 Get Channel IDs

2.5 Get Signing Secret

2.6 Invite Bot to Channels

3. Environment Variables Configuration

4. Config File Configuration

5. AWS Deployment

5.1 CDK Bootstrap (First-time only)

5.2 Deploy Foundation Stack

5.3 Build and Push Docker Image

5.4 Deploy Application Stack

5.5 Update Slack Event Subscription URL

5.6 Verify Deployment

Usage

Basic Commands

Simple Summarization

YouTube Videos

PDF Documents

Advanced Features

Send to Business Channels

Custom Opening Message

Combined Features

Output Format

Project Structure

Development

Running Tests

Code Quality

Local Testing (Without AWS)

Adding New Tools

Troubleshooting

Bot doesn't respond to mentions

Summaries not appearing

Business channels not receiving messages

Docker build fails

DynamoDB event deduplication issues

Cost Estimation

Security Best Practices

Contributing

License

Acknowledgments

Support

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Contributors 2

Uh oh!

Languages

Packages