Atom: Your Open-Source AI Assistant for Enhanced Productivity

Atom is a powerful and flexible open-source AI assistant designed to streamline your workflow, intelligently manage your schedule, automate research, and organize your information. Take control of your productivity with an assistant that you can host yourself, ensuring privacy and customization.

Why Choose Atom?

• Empower Your Productivity: Let Atom handle the tedious tasks of scheduling, note-taking, and information gathering so you can focus on what matters most.
• Open Source & Transparent: Built with transparency in mind, Atom's open-source nature means you have full visibility into its operations. No black boxes.
• Self-Hosted for Privacy & Control: Host Atom on your own infrastructure for maximum privacy and control over your data and how the assistant operates.
• Highly Customizable: Adapt Atom to your specific needs. Its modular design and open codebase allow for extensive customization and integration possibilities.
• Intelligent Automation: Leverage AI for smart scheduling that considers your preferences, automated event templating, and efficient research capabilities.
• Seamless Integrations: Connect Atom with your favorite tools like Google Calendar, Notion, Slack, Zoom, and more to create a unified productivity hub.

Table of Contents

• Why Choose Atom?
• Example Use Cases
• Documentation
• Features
• Core Agent Capabilities & Commands
• Configuration (Environment Variables)
• Diagram
  • Meeting Assist
• Deployment Options
  • Local Docker Compose
  • AWS Cloud Deployment (Self-Hosted)
• Support the Project
• Contributing

Example Use Cases

Tired of juggling multiple apps and struggling to stay organized? Atom is here to help you reclaim your focus and boost your productivity. Here are a few ways Atom can simplify your work and personal life:

• Effortless Meeting Coordination: "Find a time next week for a 30-minute meeting with Sarah and John, prioritizing Wednesday afternoon." Atom will check everyone's availability (integrating with their calendars if permitted) and propose optimal times.
• Smart Task Management (Voice-Powered): Use natural voice commands like "Atom, create a task: follow up with marketing by Friday" or "Atom, what are my tasks for today?" Atom manages these tasks in a dedicated Notion database you configure.
• Automated Information Gathering: "Research the latest trends in AI-powered personal assistants and summarize the key findings in a Notion document." Atom's research agents can browse, collect, and synthesize information, delivering it directly to your knowledge base.
• Voice-Powered Note-Taking: While commuting, you can say: "Atom, take an audio note: Idea for marketing campaign - focus on social media engagement and influencer collaborations." Atom will transcribe the audio and save it to Notion.
• Automated Meeting Summaries & Action Items: Atom can process your meeting transcripts (e.g., from live meeting attendance or uploaded recordings) and then automatically extract key decisions and action items directly into your Notion meeting notes.
• Intelligent Information Retrieval: Ask Atom "What did we decide about Project X?" or "Search my meetings for discussions on marketing strategy." Atom can semantically search through your transcribed meeting archives (stored in Notion & LanceDB) to find relevant information quickly.
• Proactive Schedule Optimization: With Autopilot, Atom can learn your work patterns and preferences. "My mornings are for deep work. Keep them as free of meetings as possible." Atom will then intelligently schedule new events accordingly.
• Quickly Access Information: "What was the outcome of the Project Phoenix meeting last month?" Atom can search your linked Notion notes and relevant event details to provide you with the context you need.
• Stay on Top of Your Day: "What's on my agenda for today?" or "Do I have any free time this afternoon for a quick call?"
• Proactive Meeting Prep: "Atom, get me ready for my meeting with 'Project X'." Atom gathers related notes, emails, and tasks.
• Weekly Review & Preview: "Atom, what's my weekly digest?" Atom summarizes completed tasks, key meetings, and highlights upcoming critical items.
• Intelligent Follow-ups: After a meeting, ask "Atom, what follow-ups for the 'Project X' meeting?" Atom analyzes notes/transcripts for actions, decisions, and questions.

Documentation

Atom provides several resources to help you get started and make the most of its features:

• Local Deployment (Docker Compose): For detailed instructions on setting up and running Atom locally using Docker Compose, please see the Local Docker Deployment Guide.
• AWS Cloud Deployment: To deploy Atom to your own AWS account for a scalable solution, refer to the AWS Cloud Deployment Guide.
• Technical Documentation: Additional technical details, API guides, and development logs can be found in the atomic-docker/docs/ directory.
• Configuration Details: For a comprehensive list of environment variables, consult the .env.example file in the atomic-docker/project/ directory and the setup sections within the deployment guides mentioned above.

Features

Atom comes packed with a variety of features designed to enhance your productivity. These include:

• Smart Scheduling & Time Management: AI-powered tools to manage your calendar effectively.
• Integrated Note-Taking & Research: Seamlessly create notes and conduct research with AI assistance.
• Live Meeting Attendance and Transcription: Atom can join Zoom, Google Meet, and Microsoft Teams meetings, transcribe them in real-time, and save the notes to Notion. Includes status tracking and audio device selection (in development).
• Comprehensive voice-driven task management using a Notion database backend.
• Semantic search capabilities across your meeting transcript archive.
• Automated extraction of action items and decisions from meeting transcripts.
• Smart Meeting Preparation: Gathers relevant documents, emails, and tasks before your meetings.
• Automated Weekly Digest: Summarizes your past week's accomplishments and previews upcoming critical items.
• Intelligent Follow-up Suggester: Analyzes meeting notes or project documents to suggest action items, decisions, and questions.

For a detailed list and explanation of all features, please see our Features Document.

Core Agent Capabilities & Commands

The Atom Agent understands a variety of commands to interact with your integrated services. Commands are typically issued in a chat interface with the agent. The agent's natural language understanding (NLU) has been significantly enhanced, particularly for creating and querying calendar events and for managing tasks using voice commands.

Atom provides a rich set of commands to manage your productivity. You can interact with Atom to:

• Manage your calendar: List, create, and modify events across Google Calendar and Microsoft Teams.
• Handle CRM tasks: Create and retrieve contacts in HubSpot.
• Manage tasks in Notion: Create, query, and update tasks in your designated Notion task database using natural language (e.g., "Atom, add 'buy milk' to my shopping list due tomorrow," "Atom, what are my work tasks for this week?").
• Search meeting transcripts: Perform semantic searches across your meeting notes (e.g., "Atom, find meetings where we discussed the Q3 budget").
• Communicate via Slack: Send messages and list channels.
• Manage video meetings: List and get details for Zoom and Google Meet events.
• Process payments: Interact with Stripe to list payments.
• Handle accounting: Work with QuickBooks Online to manage invoices.
• Take notes and conduct research: Create notes in Notion, initiate multi-agent research projects, and process research queues.
• Prepare for meetings: "Atom, prep me for my meeting on [topic/date]."
• Get weekly summaries: "Atom, what's my weekly digest?"
• Suggest follow-ups: "Atom, what are the follow-ups for [meeting/project context]?"
• And much more, including general commands like help.

For a more detailed list of commands and their specific syntax, please refer to the agent's help command or explore the agent's capabilities through interaction.

Configuration (Environment Variables)

The Atom Agent uses environment variables for its configuration and to connect to various third-party services.

Atom requires various environment variables to be set for full functionality. These include API keys for services like OpenAI, Notion, Deepgram, and credentials for various integrations. Below are some key variables, including those for newer features. Always consult the .env.example file in atomic-docker/project/ for a comprehensive list and the deployment guides for detailed setup.

Core Service & API Keys:

• OPENAI_API_KEY: For AI capabilities.
• NOTION_API_KEY: For Notion integration (used by various skills, including task management).
• DEEPGRAM_API_KEY: For transcription services.
• ATOM_NOTION_TASKS_DATABASE_ID: The ID of your Notion database dedicated to task management. Refer to the "Notion Task Management" section in FEATURES.md for required database properties.
• LANCEDB_URI: URI for your LanceDB instance (e.g., ./lance_db for local, or a remote URI). Required for semantic search of meeting transcripts.
• PYTHON_API_SERVICE_BASE_URL: (Optional) URL for the Python backend service if not http://localhost:8080.

Slack Integration (Agent):

• ATOM_SLACK_BOT_TOKEN: Slack Bot User OAuth Token for Slack integration (sending messages, searching, etc.). Requires appropriate scopes.
  • Note: OPENAI_API_KEY is also used by the agent for AI-powered Slack query understanding and information extraction.

Microsoft Teams Integration (Agent & UI - User OAuth):

• MSTEAMS_CLIENT_ID: Azure AD App Registration Client ID (for frontend/Next.js backend's OAuth flow).
• MSTEAMS_CLIENT_SECRET: Azure AD App Registration Client Secret (for Next.js backend's OAuth flow).
• MSTEAMS_REDIRECT_URI: Your app's redirect URI, e.g., http://localhost:3000/api/atom/auth/msteams/callback. Must be registered in Azure AD.
• MSTEAMS_AUTHORITY: (Optional) Microsoft Identity Platform authority, e.g., https://login.microsoftonline.com/common or specific tenant.
  • Note: This OAuth flow uses delegated permissions (e.g., Chat.Read, ChannelMessage.Read.All) to access user's Teams data. OPENAI_API_KEY is used by the agent for AI-powered Teams message query understanding and information extraction.

Google Account Integration (Calendar & Gmail - Agent & UI):

• ATOM_GOOGLE_CALENDAR_CLIENT_ID: Google Cloud OAuth Client ID (used by atom-agent for services like Calendar).
• ATOM_GOOGLE_CALENDAR_CLIENT_SECRET: Google Cloud OAuth Client Secret (used by atom-agent).
• NEXT_PUBLIC_GOOGLE_CLIENT_ID_ATOMIC_WEB: Google Cloud OAuth Client ID used by the frontend/Next.js backend for the OAuth flow.
• GOOGLE_CLIENT_SECRET_ATOMIC_WEB: Google Cloud OAuth Client Secret used by the Next.js backend for the OAuth flow.
• GOOGLE_OAUTH_ATOMIC_WEB_REDIRECT_URL: Your app's redirect URI, e.g., http://localhost:3000/api/atom/auth/calendar/callback. This must be registered in your Google Cloud Console.
  • Note: This OAuth flow now requests scopes for both Google Calendar and Gmail (read-only). OPENAI_API_KEY is also used by the agent for AI-powered Gmail query understanding and information extraction.

Wake Word Detection (Frontend - Experimental):

• NEXT_PUBLIC_AUDIO_PROCESSOR_URL: (Optional) URL of your external WebSocket-based STT service for wake word detection.
• NEXT_PUBLIC_MOCK_WAKE_WORD_DETECTION: (Optional) Set to true to simulate wake word detection for UI testing without a live STT service.

Live Meeting Attendance (Frontend & Worker - Experimental):

• NEXT_PUBLIC_LIVE_MEETING_WORKER_URL: URL for the live_meeting_worker service (e.g., http://localhost:8081), used by the frontend to list audio devices.
• For the live_meeting_worker and attend_live_meeting handler:
  • ZOOM_SDK_KEY & ZOOM_SDK_SECRET: Required if using the NewZoomSdkAgent. These are passed from the client during the API call but might also be relevant for the handler or worker environment depending on exact flow. The setup guide indicates these are passed in handler_input.

Database & Hasura:

• HASURA_GRAPHQL_URL: URL for your Hasura GraphQL endpoint.
• HASURA_ADMIN_SECRET: Admin secret for Hasura.
• POSTGRES_USER, POSTGRES_PASSWORD, POSTGRES_DB: Credentials for the PostgreSQL database.

For a comprehensive list of all environment variables and their setup, please refer to the .env.example file in the atomic-docker/project/ directory and the detailed setup instructions in the deployment guides for Local Docker Compose and AWS Cloud Deployment.

Diagram

Meeting Assist

    sequenceDiagram
    actor Alice
    participant A as Atom
    actor Bob
    actor John
    participant H as handshake.Atomlife.app
    participant P as AI Scheduler

    participant G as Google Calendar
    Alice->>A: Create a new meeting assist with John & Bob as attendees
    A->>John: Sends handshake link for a possible meeting
    A->>Bob: Sends another handshake link to another attendee 
    John->>H: Selects time preferences (not availability like a booking link)
    Bob->>H: Also selects time preferences
    H->>P: Submits & starts AI planner after minimum threshold met
    P->>G: finds an optimal slot & creates the new event

Loading

Deployment Options

This project offers multiple ways to deploy and run the Atom application stack.

Local Docker Compose

For local development, testing, and self-hosting on a single machine, the project can be run using Docker Compose. This method utilizes the services defined in the atomic-docker/ directory.

• Setup and Instructions: See the detailed guide in atomic-docker/README.md.

AWS Cloud Deployment (Self-Hosted)

For a scalable and robust cloud environment, you can deploy the entire application stack to your own AWS account. This deployment is managed by the AWS Cloud Development Kit (CDK) and provisions all necessary infrastructure, including managed services for databases, messaging, and search where appropriate.

• Features: Deploys core application services, Optaplanner, a new python-agent service (for notes and research), and utilizes AWS S3, Amazon EFS (for LanceDB vector stores), and Amazon MSK Serverless. Amazon OpenSearch Service is no longer used. Note: The Python backend service (often named python_api_service or python-agent in deployment configurations) has been updated with new responsibilities. It now handles task management operations with Notion and semantic search queries using LanceDB, in addition to its previous roles. Ensure it is deployed with access to necessary configurations (Notion tokens, LanceDB URI) and that the notion-client>=2.0.0 dependency is included in its environment.

• Detailed Guide: For prerequisites, setup instructions, deployment steps, and management, please refer to the comprehensive AWS Deployment Guide.

Support the Project

• I'm spending 100% of my work time on this project
• Star this repository, so I can start an Open Collective to support this project
• In process of setting up Github Sponsors
• Follow me on Twitter: https://twitter.com/rish1_2
• Used Atom? write a review or let me know!

Contributing

1. Fork this repository and clone the fork to your machine
2. Create a branch (git checkout -b my-new-feature)
3. Implement a new feature or fix a bug and add some tests or proof of fix
4. Commit your changes (git commit -am 'Added a new feature')
5. Push the branch to your fork on GitHub (git push origin my-new-feature)
6. Create new Pull Request from your fork