Skip to content

Repository files navigation

Sentient

Your Second Brain. Powered by Logic.

An autonomous "Chief of Staff" agent that manages your inbox, calendar, and memory.


✦ The Vision

Sentient is not just a chatbot; it is an agentic workflow engine designed to reclaim your time. Unlike standard assistants that wait for commands, Sentient actively manages your digital life.

Demo Video: https://drive.google.com/file/d/1O41VfIk7UnfQQ8_OedxftvDhYJ9MUOM3/view?usp=sharing

It connects a Reactive Frontend (Next.js/Tailwind) with a Cognitive Backend (LangGraph/Gemini), allowing it to:

  • Read & Reason: Analyze emails and calendar for action items, not just keywords.
  • Act Autonomously: Schedule, reschedule, and delete meetings based on deep context.
  • Remember Forever: Utilizes vector-based/PostgreSQL long-term memory to recall user preferences.

Key Capabilities

1. Inbox Intelligence

Sentient connects to Gmail via OAuth to perform semantic analysis on your inbox.

  • Summarization: Condenses long thread chains into actionable bullet points.
  • Contextual Drafting: Drafts replies using the correct "From" aliases and historical context.

2. Time Sovereignty

Full bi-directional control of Google Calendar.

  • Conflict Resolution: Checks for overlaps before proposing times.
  • Smart Scheduling: "Book a meeting with Joel" automatically finds the best slot based on your preferences.
  • Management: Can delete and reschedule events autonomously when plans change.

3. Infinite Context (Active Memory)

Powered by PostgreSQL, Sentient maintains a persistent user profile.

  • Fact Extraction: Automatically extracts constraints from emails and saves them.
  • Preference Recall: Consults memory before every action.

Tech Stack & Architecture

Frontend (The Face)

  • Framework: Next.js 14 (React)
  • Styling: Tailwind CSS v4 + Framer Motion (Animations)
  • UI/UX: Glassmorphism, Fluid Mesh Gradients, Lucide Icons

Backend (The Brain)

  • Core: Python FastAPI
  • Agent Framework: LangGraph + LangChain
  • LLM: Google Gemini 3.0 Flash (optimized for speed/cost)
  • Database: PostgreSQL (hosted on Render)

Infrastructure (The Cloud)

  • Containerization: Docker (Multi-stage builds)
  • Deployment: Render (Auto-Deploy from GitHub)
  • CI/CD: Automated pipelines trigger on every push to main.
Why Render instead of AWS?

AWS didn't like my debit cards, cancelled 4 of them! so couldn't set up billing.


Getting Started

Prerequisites

  • Docker & Docker Compose
  • Google Cloud Console Project (with Gmail/Calendar APIs enabled)
  • Google Gemini API Key

1. Installation

# Clone the repository
git clone https://github.com/AryanAT13/personalAI.git
cd personalAI

# Setup Backend Environment
cd backend
cp .env.example .env
# Fill in your GOOGLE_CLIENT_ID, SECRET, and GEMINI_API_KEY

2. Run with Docker (Recommended)

Sentient is fully containerized. You can spin up the entire stack (Frontend + Backend) with one command:

docker-compose up --build
The app will be available at:

Frontend: http://localhost:3000
Backend: http://localhost:8000

3. Deployment Strategy

This project utilizes a Microservices Architecture deployed on Render.

  • Frontend Service: Deployed as a Static Site (Next.js export).
  • Backend Service: Deployed as a Web Service (Python/FastAPI) running inside a Docker Container.
  • Database: Managed PostgreSQL instance on Render.

4. CI/CD Pipeline

Deployment is fully automated. Pushing to the main branch triggers Render's build pipeline, which:

  • Builds the Docker image from backend/Dockerfile.
  • Runs health checks.
  • Performs a zero-downtime rollout.

5. Security & Privacy

  • OAuth 2.0: Uses Google's official OAuth flow. User tokens are stored securely and never exposed to the client.
  • Data Isolation: User memory is isolated in the database.
  • No Training: User data is processed ephemerally by the LLM and is not used for model training.

About

Internship

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages