📬 InboxOS — The AI Inbox Operating System

Not an email client. A decision + execution layer that reads your inbox, understands what matters, and acts on it automatically.

📋 Table of Contents

Introduction & Problem Statement
Key Features
System Architecture
The 5-Layer Pipeline (Detailed)
Database Schema
WebSocket Interface Specification
API Route Reference
Project Structure
Local Development Setup
Configuration & Environment Variables
Production Deployment Guide
Keyboard Shortcuts
Performance Optimizations
Team & Contributing
Appendix

1. Introduction & Problem Statement

The average knowledge worker receives 120+ emails per day.

Inboxes have become a cluttered dumping ground containing a chaotic mix of urgent project threads, high-priority notifications, OTP verification codes, meeting invites, invoices, automated newsletters, and spam. Current tools attempt to solve this with summaries, but a summary is simply shorter noise.

InboxOS represents a paradigm shift. It is a dedicated decision + execution layer that sits on top of your existing mail provider. If Gmail or Outlook is the hard drive, InboxOS is the operating system. It parses incoming streams, uses structured AI reasoning to identify deadlines and tasks, evaluates programmatic routing rules, and automatically acts on them—scheduling calendar events, creating dashboard tasks, and pushing urgent alerts directly to your messaging apps.

InboxOS is built from the ground up to democratize inbox automation: it is open-source, fully self-hostable, and supports local AI models (via Ollama) so that your private emails never leave your machine.

2. Key Features

⚙️ 5-Layer AI Pipeline: A structured flow that processes raw emails through ingestion, parser, intelligence, rules, action, and delivery layers.
🔌 Multi-Provider Connectors: Built-in connectors for Gmail, Outlook (Graph API), and generic IMAP servers with full OAuth 2.0 security.
🧠 AI Classification Engine: Automatic sorting of incoming mail into 8 distinct categories (academic, job, finance, meeting, OTP, newsletter, personal, support) with granular priority scoring ($0$–$100$).
📝 Smart Action Extraction: Extracts structured entities: deadlines, action items, monetary amounts, and physical/digital locations.
🎛️ Rules Engine: User-defined routing conditions with boolean logic, priority evaluation, and active time-of-day windows.
💬 Multi-Channel Delivery: Real-time push alerts to WhatsApp, Slack webhooks, Telegram, email digests, and custom WebSockets.
📊 Real-Time Dashboard: Dark glassmorphism dashboard built with Next.js 14 and Tailwind CSS, updating live via WebSocket.
🔒 Privacy-First AI: Run fully local intelligence models using Ollama (Llama 3/Mistral) or hook up to commercial endpoints (OpenAI / Gemini).
📦 Plugin SDK (v3 roadmap): Unified interfaces for building third-party connectors and execution actions.
👥 Team Inbox (v3 roadmap): Multi-user support with ticket assignments, audit logs, and response SLA tracking.

3. System Architecture

InboxOS processes data asynchronously, routing raw email streams through a structured parsing, classification, and execution pipeline.

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#1e293b', 'primaryTextColor': '#f8fafc', 'primaryBorderColor': '#3b82f6', 'lineColor': '#6366f1', 'secondaryColor': '#0f172a', 'tertiaryColor': '#1e293b'}}}%%
graph TD
    A[Email Providers <br> Gmail / Outlook / IMAP] -->|OAuth 2.0 Ingestion| B(Connector Layer)
    B -->|Raw Multipart Email| C(Parser Layer)
    C -->|Clean text/plain & Markdown| D(Intelligence Layer)
    D -->|LLM Structured Extraction JSON| E(Rules Engine DSL)
    E -->|Route & Rule Matches| F(Action Engine)
    F -->|System Task Executed| G(Output Adapters)
    G -->|WhatsApp / Slack / WebSocket / SMS| H[Delivery Channels]

    classDef default fill:#1e293b,stroke:#3b82f6,stroke-width:2px,color:#f8fafc;
    classDef storage fill:#0f172a,stroke:#10b981,stroke-width:2px,color:#f8fafc;
    classDef queue fill:#312e81,stroke:#8b5cf6,stroke-width:2px,color:#f8fafc;

    subgraph Infrastructure
        DB[(PostgreSQL 15)]:::storage
        CACHE[(Redis Cache)]:::storage
        WORKER(Celery Workers):::queue
    end

    B -.->|Save Raw Email| DB
    D -.->|Cache Tokens & Embeddings| CACHE
    F -.->|Queue Async Actions| WORKER
    G -.->|WebSocket Stats Broadcast| CACHE

4. The 5-Layer Pipeline (Detailed)

[Raw Email] -> (Ingestion) -> (Parse) -> (Intelligence) -> (Rules) -> (Action) -> (Delivery)

Layer 1 — Ingestion

Establishes connection to mail servers. Includes the Gmail Connector (using Google API v1 with Pub/Sub push notifications for instant sync), Outlook Connector (Microsoft Graph API with delta sync), and an IMAP Connector that polls standard mail servers at configurable intervals.

Layer 2 — Parser

Converts dirty multipart HTML and plain text into clean Markdown. Strips signatures using regex and heuristic layouts, removes quoted reply chains, extracts attachment metadata, and sanitizes dangerous scripts.

Layer 3 — Intelligence

Interacts with the configured LLM client through a provider-agnostic interface (LLMClient). Executes system prompting using few-shot classification templates, utilizing OpenAI's JSON Mode or Gemini's Structured Output to return reliable JSON structures.

Layer 4 — Rules

An evaluation engine that executes user-defined constraints. It processes logical conditions (sender_is, domain_is, category_is, priority_above, has_attachment, sent_before) and applies priority overrides and active scheduling windows.

Layer 5 — Delivery

Distributes outputs to external target endpoints. Integrates with the Twilio WhatsApp API, Slack Webhook client, Telegram Bot API, and Next.js WebSockets to keep developers and users notified.

Example Data Flow

Layer	Input / Data Format	Output / Extracted Data
Layer 1: Ingestion	Raw SMTP / IMAP Socket Connection	`{"id": "msg_9918", "raw_headers": "...", "body_html": "..."}`
Layer 2: Parser	Raw HTML / Multipart Text	`{"clean_text": "Project update due by Friday 5 PM. Total invoice is $1,200."}`
Layer 3: Intelligence	Clean text/plain	`{"category": "finance", "priority": 95, "deadline": "2026-07-03T17:00:00Z"}`
Layer 4: Rules	JSON metadata & categories	`MATCH RULE "Finance Priority Alert" -> Action: WhatsApp Alert & Task Creator`
Layer 5: Delivery	Scheduled Task & Notification payload	WhatsApp sent: "Finance Alert: Invoice of $1200 due on Friday at 5 PM."

5. Database Schema

The persistent layer is backed by PostgreSQL 15, optimized for JSONB indexing to handle evolving metadata fields.

-- Core users table
CREATE TABLE users (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    email VARCHAR(255) UNIQUE NOT NULL,
    name VARCHAR(255) NOT NULL,
    preferences JSONB NOT NULL DEFAULT '{}',
    created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);

-- Email connections mapping
CREATE TABLE email_accounts (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id) ON DELETE CASCADE,
    provider VARCHAR(50) NOT NULL, -- 'gmail', 'outlook', 'imap'
    encrypted_credentials BYTEA NOT NULL,
    sync_state VARCHAR(50) NOT NULL,
    last_sync_at TIMESTAMP WITH TIME ZONE,
    created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);

-- Persistent email storage
CREATE TABLE emails (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    account_id UUID REFERENCES email_accounts(id) ON DELETE CASCADE,
    provider_message_id VARCHAR(255) NOT NULL,
    thread_id VARCHAR(255),
    sender VARCHAR(255) NOT NULL,
    subject VARCHAR(500) NOT NULL,
    body_text TEXT NOT NULL,
    body_html TEXT,
    received_at TIMESTAMP WITH TIME ZONE NOT NULL,
    raw_metadata JSONB NOT NULL DEFAULT '{}'
);

-- Structured AI analysis results
CREATE TABLE email_analysis (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    email_id UUID REFERENCES emails(id) ON DELETE CASCADE,
    category VARCHAR(50) NOT NULL,
    priority_score INT NOT NULL,
    urgency_score INT NOT NULL,
    actionability_score INT NOT NULL,
    confidence_score INT NOT NULL,
    summary TEXT NOT NULL,
    extracted_data JSONB NOT NULL DEFAULT '{}',
    analyzed_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);

-- Programmatic Action Rules
CREATE TABLE rules (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id) ON DELETE CASCADE,
    name VARCHAR(255) NOT NULL,
    conditions JSONB NOT NULL DEFAULT '[]',
    actions JSONB NOT NULL DEFAULT '[]',
    priority_order INT NOT NULL,
    is_active BOOLEAN NOT NULL DEFAULT TRUE,
    created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);

-- Tasks generated from emails
CREATE TABLE tasks (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id) ON DELETE CASCADE,
    title VARCHAR(255) NOT NULL,
    description TEXT,
    deadline TIMESTAMP WITH TIME ZONE,
    status VARCHAR(50) NOT NULL DEFAULT 'pending', -- 'pending', 'in_progress', 'completed'
    source_email_id UUID REFERENCES emails(id) ON DELETE SET NULL,
    created_at TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
);

-- Multi-channel system notifications
CREATE TABLE notifications (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id) ON DELETE CASCADE,
    channel VARCHAR(50) NOT NULL, -- 'whatsapp', 'slack', 'telegram', 'websocket'
    content TEXT NOT NULL,
    status VARCHAR(50) NOT NULL DEFAULT 'queued', -- 'queued', 'sent', 'failed'
    sent_at TIMESTAMP WITH TIME ZONE,
    read_at TIMESTAMP WITH TIME ZONE
);

-- Core Database Indexes for Performance
CREATE INDEX idx_emails_received_at ON emails (received_at DESC);
CREATE INDEX idx_email_analysis_scores ON email_analysis (priority_score, category);
CREATE INDEX idx_emails_raw_metadata_gin ON emails USING gin (raw_metadata);
CREATE INDEX idx_email_analysis_data_gin ON email_analysis USING gin (extracted_data);
CREATE INDEX idx_tasks_deadline ON tasks (deadline NULLS LAST);

6. WebSocket Interface Specification

Real-time interface connection point: /ws/{user_id} (requires JWT in headers or query parameters).

Server → Client Events

`email.received`

Pushed when an incoming email is fetched, parsed, and analyzed by the AI pipeline.

{
  "event": "email.received",
  "timestamp": "2026-06-29T20:15:00Z",
  "data": {
    "email_id": "8a09b2e1-456c-789d-01ef-23456789abcd",
    "sender": "billing@cloudprovider.com",
    "subject": "Overdue Invoice: #INV-29091",
    "analysis": {
      "category": "finance",
      "priority_score": 92,
      "urgency_score": 85,
      "summary": "Urgent request to pay overdue monthly cloud hosting bill.",
      "extracted_data": {
        "due_date": "2026-07-01T23:59:59Z",
        "amount_due": 1240.50,
        "currency": "USD"
      }
    }
  }
}

`task.created`

Pushed when the action engine automatically extracts a task from an analyzed email.

{
  "event": "task.created",
  "timestamp": "2026-06-29T20:15:02Z",
  "data": {
    "task_id": "9b10c3f2-567d-890e-12fa-34567890bcde",
    "title": "Pay Overdue Invoice: #INV-29091",
    "deadline": "2026-07-01T23:59:59Z",
    "source_email_id": "8a09b2e1-456c-789d-01ef-23456789abcd"
  }
}

`dashboard.stats`

Broadcast to update real-time graphs and figures in the frontend client.

{
  "event": "dashboard.stats",
  "timestamp": "2026-06-29T20:15:05Z",
  "data": {
    "unread_count": 14,
    "processed_today": 182,
    "categories_distribution": {
      "finance": 12,
      "meeting": 8,
      "newsletter": 145,
      "personal": 17
    }
  }
}

Client → Server Events

`subscribe_user`

Subscribes the connection channel to user-specific events.

{
  "action": "subscribe_user",
  "user_id": "1a23b4c5-678d-90ee-fffa-bcde12345678"
}

`mark_read`

Instructs the server to mark a specific email and its tasks as processed.

{
  "action": "mark_read",
  "email_id": "8a09b2e1-456c-789d-01ef-23456789abcd"
}

7. API Route Reference

All API routes require authentication using a Bearer JWT Token unless explicitly labeled Public.

Method	Route	Description	Auth
POST	`/auth/connect/gmail`	Initiates Gmail OAuth credentials connection flow.	JWT
GET	`/auth/callback/gmail`	Callback handler for Google credentials token exchange.	Public
GET	`/emails`	Query processed emails. Supports pagination, category, and score filtering.	JWT
GET	`/emails/{id}`	Fetches detailed parsed message body, AI analysis, and action items.	JWT
GET	`/emails/{id}/analysis`	Fetches the raw AI classification output data.	JWT
POST	`/tasks`	Create manual tasks or trigger an extraction from a specific email.	JWT
PATCH	`/tasks/{id}`	Update task status (`pending`, `in_progress`, `completed`).	JWT
GET	`/rules`	Retrieves user-defined evaluation rules.	JWT
POST	`/rules`	Declares a new rules parsing route definition.	JWT
DELETE	`/rules/{id}`	Removes a rules parsing route definition.	JWT
GET	`/dashboard/stats`	Fetches core metrics, category count, and queue states.	JWT
GET	`/dashboard/heatmap`	Generates activity maps for dashboard analytics.	JWT
GET	`/settings`	Update/fetch active AI clients, priorities, and secrets.	JWT
POST	`/webhooks/gmail`	Push endpoint for Google Cloud Pub/Sub instant notification.	Signed
GET	`/health`	Simple diagnostic route verifying database and Redis connections.	Public
WS	`/ws/{user_id}`	Live update connection link.	JWT

8. Project Structure

This monorepo isolates individual layers and clients under clean folders. All config files are structured under config/ to keep the root directory pristine.

InboxOS/
├── apps/
│   ├── api/                    # FastAPI python server backend
│   └── web/                    # Next.js 14 frontend client (React + Vite)
├── config/                     # Configuration and Environment mappings
│   ├── docker/
│   │   └── docker-compose.yml  # Relocated Docker Compose runtime configuration
│   ├── env/
│   │   └── .env.example        # Environment variables template file
│   └── git/
│       └── gitignore           # Symbolic link pointing to root .gitignore
├── docs/                       # Developer manuals & ADR logs
│   └── implementation-plan.md  # Core implementation design document
├── packages/                   # Decoupled libraries
│   ├── core/                   # Pydantic schemas, exception logic
│   ├── intelligence/           # AI classifier engines & templates
│   ├── outputs/                # Adapters for Slack, Telegram, WhatsApp
│   └── rules-engine/           # Evaluator DSL parser logic
├── .gitignore                  # Git tracking rules
└── README.md                   # Repository Hero and Reference manual (this file)

9. Local Development Setup

Prerequisites

Python $3.11+$
Node.js $18+$
PostgreSQL $15+$
Redis $7+$

Quick Start (Docker Compose)

The entire pipeline can be deployed locally using the Docker Compose configuration located in the config/ directory.

# 1. Clone the repository
git clone https://github.com/inboxos/inboxos.git
cd inboxos

# 2. Copy and configure variables
cp config/env/.env.example config/env/.env
# Edit config/env/.env with your secrets

# 3. Spin up all services (PostgreSQL, Redis, API Backend, Celery, Frontend Dashboard)
docker compose -f config/docker/docker-compose.yml up -d

# 4. Execute database migrations
docker compose -f config/docker/docker-compose.yml exec api alembic upgrade head

# 5. Open the Dashboard UI
# Access the dashboard at http://localhost:5173

Manual Setup (Development)

For active local debugging of individual services without Docker containerization.

1. Backend API Server

cd apps/api
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp ../../config/env/.env.example ../../config/env/.env
uvicorn main:app --reload --port 8000

2. Celery Worker Queue (new terminal)

cd apps/api
source venv/bin/activate
celery -A tasks worker --loglevel=info

3. Frontend Dashboard UI (new terminal)

cd apps/web
npm install
npm run dev

10. Configuration & Environment Variables

Create your configuration files inside config/env/.

Backend Configuration (`config/env/.env`)

Key	Type	Default	Description
`DATABASE_URL`	String	`postgresql+asyncpg://inboxos:inboxos_dev@postgres:5432/inboxos`	SQLAlchemy database URI.
`REDIS_URL`	String	`redis://localhost:6379/0`	Redis caching & async broker address.
`OPENAI_API_KEY`	String	—	OpenAI key (required if `AI_PROVIDER` is `openai`).
`GEMINI_API_KEY`	String	—	Google AI studio key (required if provider is `gemini`).
`OLLAMA_BASE_URL`	String	`http://localhost:11434`	Endpoint for local running model.
`AI_PROVIDER`	String	`mock`	Selected AI engine (`openai`, `gemini`, `ollama`, `mock`).
`JWT_SECRET`	String	`dev-secret-change-in-production-32chars!!`	Cryptographic secret for signing API tokens.
`GMAIL_CLIENT_ID`	String	—	Google Developer Console Client Identity credentials.
`GMAIL_CLIENT_SECRET`	String	—	Google Developer Console Client Secret credentials.
`TWILIO_ACCOUNT_SID`	String	—	Twilio SID credentials (for WhatsApp dispatch).
`TWILIO_AUTH_TOKEN`	String	—	Twilio token verification credentials.

Frontend Configuration (`apps/web/.env`)

Key	Type	Default	Description
`NEXT_PUBLIC_API_URL`	String	`http://localhost:8000`	Address location of FastAPI backend server.
`NEXT_PUBLIC_WS_URL`	String	`ws://localhost:8000`	Address location for real-time WebSocket.

11. Production Deployment Guide

Backend Deploy (DigitalOcean App Platform / Render)

Provision a managed PostgreSQL 15 and Redis instance.
Build the Dockerfile context from apps/api/ folder.
Bind the environment variables to point to production endpoints (ensure APP_ENV=production is set).
Run alembic upgrade head inside a release command wrapper hook before launching.

Frontend Deploy (Vercel)

Connect the repository and configure target build directory to apps/web.
Configure build framework presets to Vite / Next.js.
Supply NEXT_PUBLIC_API_URL and NEXT_PUBLIC_WS_URL env parameters pointing to the live API backend server.

Self-Hosted (Single VPS with Nginx)

For deploying everything on a single virtual private server.

Run docker compose -f config/docker/docker-compose.yml up -d on the VPS.
Bind local Nginx to reverse proxy port 8000 (API backend) and port 5173 (Frontend).

Secure the connection with Let's Encrypt SSL certificates:

sudo apt-get install certbot python3-certbot-nginx
sudo certbot --nginx -d yourdomain.com

12. Keyboard Shortcuts

Increase efficiency inside the dark glassmorphism dashboard UI.

Shortcut Key	Action
`/`	Focus Global Search Bar
`g`	Navigate to Inbox view
`a`	Navigate to Alerts and Tasks dashboard
`r`	Force reload sync channels
`n`	Create New Routing Rule
`?`	Toggle Keyboard Shortcuts Overlay
`Esc`	Dismiss active panels, modals, and filters

13. Performance Optimizations

📡 Parser Caching: Parsed multipart email contexts and structures are serialized and stored in Redis with a 24-hour expiration window.
🤖 AI Batch Classification: Reduces LLM latency by grouping low-importance emails (such as newsletters and support tickets) and classifying them in batches of 5. Uses a lightweight routing model (like gpt-4o-mini) for initial sorting before processing with complex extraction pipelines.
🗂️ Database Composite Indexing: High-speed queries are achieved using composite indexes built on (user_id, received_at DESC) and (category, priority_score DESC).
🖥️ Instanced Rendering: The frontend dashboard list utilizes Virtual Scrolling (react-window) to smoothly render long inbox logs containing 10,000+ entries.
⚡ Connection Pooling: Backend connections are managed using AsyncPG with SQLALchemy async engines, maintaining connections without initialization lag.
🛡️ Redis-Backed Throttling: Protects pipeline processes using rate limiters configured to allow a burst of 50 emails/hour per connected inbox user.

14. Team & Contributing

InboxOS is built by a distributed core team of open-source developers.

Name / Role	Scope	Experience	Core Focus
Siddharth Mehta (Lead Architect)	Core Engine	Advanced	Orchestration layer, async database flow
Elena Rostova (AI Lead)	Intelligence	Advanced	Prompt templates, structured outputs
Marcus Vance (DevOps Architect)	Infrastructure	Advanced	Celery, Redis scaling, Docker environments
Hana Tanaka (Frontend Lead)	Web Client	Advanced	Next.js glassmorphism client, WebSockets
Lucas Dupont (Connector Lead)	Integrations	Advanced	Gmail OAuth and Microsoft Graph sync
8 Community Contributors	Extensions	Beginners	Webhook integrations, CLI scripts, translation templates

Contributing to InboxOS

We welcome contributions to the pipeline connectors and adapters!

Check the Good First Issues label—we currently maintain 102+ onboarding tasks!
Read the standard CONTRIBUTING.md guidelines for branch styles.
InboxOS proudly participates in Hacktoberfest. Submit your PRs according to guidelines!

15. Appendix

Example Flow 1: Academic Deadline Email

Layer 1 (Ingestion): Gmail connector fetches notification: "A reminder that your DBMS Mini-Project is due on July 3, 2026, at 11:59 PM."
Layer 2 (Parser): Converts HTML elements, removes headers and signature signatures.
Layer 3 (Intelligence): Sorts to category: academic. Assigns priority_score: 95. Extracts structured actions: {"title": "Submit DBMS Mini-Project", "deadline": "2026-07-03T23:59:00Z"}.
Layer 4 (Rules): Evaluates matches: Priority is $>90$ and category is academic -> Trigger Action: WhatsApp and Dashboard task creation.
Layer 5 (Delivery): Task added to PostgreSQL database. Send message: "Urgent: Submit DBMS Mini-Project due on July 3, 2026."

Example Flow 2: Job Recruiter Email

Layer 1 (Ingestion): Outlook graph webhook fetches notification: "Hi applicant, we loved your profile! Schedule a technical call by Friday using this link."
Layer 2 (Parser): Converts email formatting to clean text.
Layer 3 (Intelligence): Sorts to category: job. Assigns priority_score: 98. Extracts task details: {"title": "Schedule Technical Interview Call", "deadline": "Friday 5 PM"}.
Layer 4 (Rules): Evaluates matches: Matches rule "Job Application Tracker" -> Priority boosted, route task.
Layer 5 (Delivery): Schedules calendar reminder, pushes task card to user alerts dashboard.

Example Flow 3: Digest Newsletter Email

Layer 1 (Ingestion): IMAP polling retrieves weekly bulletin newsletter.
Layer 2 (Parser): Strips tracking pixels and promotional headers.
Layer 3 (Intelligence): Sorts to category: newsletter. Assigns priority_score: 15 (low importance).
Layer 4 (Rules): Evaluates matches: Category is newsletter -> Match rule: "Newsletter Digest Rollup". Route to newsletter aggregate, bypass direct phone alert.
Layer 5 (Delivery): Inserts message summary into database weekly briefing logs.

Glossary

Connector: Component that authenticates and fetches raw data from a specific email provider.
Adapter: Output router that delivers system messages to targeted messaging APIs (e.g., Slack Webhooks).
Pipeline: The continuous asynchronous process of ingesting, parsing, classifying, evaluating, and delivering.
LLMClient: Provider-agnostic engine wrapper handling structured communications with AI models.
Rules Engine: DSL-based evaluator verifying user constraints on processed email metadata.
Digest: Grouped summaries of low-importance emails compiled into a single daily or weekly report.
RAG: Retrieval-Augmented Generation. Used to provide context on previous mail history to the AI agent.
Conjunction: Logical matching criteria (AND/OR operations) in the Rules Engine.
Orchestration: Core server code running background pipelines and scheduling Celery queues.

Developer Resources

..

Name		Name	Last commit message	Last commit date
Latest commit History 71 Commits
apps/api		apps/api
backend		backend
config		config
docs		docs
frontend		frontend
packages		packages
.gitignore		.gitignore
README.md		README.md
package-lock.json		package-lock.json

Uh oh!

Folders and files

Latest commit

History

Repository files navigation

📬 InboxOS — The AI Inbox Operating System

📋 Table of Contents

1. Introduction & Problem Statement

2. Key Features

3. System Architecture

4. The 5-Layer Pipeline (Detailed)

Layer 1 — Ingestion

Layer 2 — Parser

Layer 3 — Intelligence

Layer 4 — Rules

Layer 5 — Delivery

Example Data Flow

5. Database Schema

6. WebSocket Interface Specification

Server → Client Events

email.received

task.created

dashboard.stats

Client → Server Events

subscribe_user

mark_read

7. API Route Reference

8. Project Structure

9. Local Development Setup

Prerequisites

Quick Start (Docker Compose)

Manual Setup (Development)

1. Backend API Server

2. Celery Worker Queue (new terminal)

3. Frontend Dashboard UI (new terminal)

10. Configuration & Environment Variables

Backend Configuration (config/env/.env)

Frontend Configuration (apps/web/.env)

11. Production Deployment Guide

Backend Deploy (DigitalOcean App Platform / Render)

Frontend Deploy (Vercel)

Self-Hosted (Single VPS with Nginx)

12. Keyboard Shortcuts

13. Performance Optimizations

14. Team & Contributing

Contributing to InboxOS

15. Appendix

Example Flow 1: Academic Deadline Email

Example Flow 2: Job Recruiter Email

Example Flow 3: Digest Newsletter Email

Glossary

Developer Resources

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`email.received`

`task.created`

`dashboard.stats`

`subscribe_user`

`mark_read`

Backend Configuration (`config/env/.env`)

Frontend Configuration (`apps/web/.env`)

Packages