CalSync AI

AI-Powered Smart Scheduling & Productivity Assistant (Local LLM + Google Calendar)

CalSync AI is a local-first AI productivity system that generates structured learning plans, converts them into optimized schedules, and automatically syncs them with Google Calendar.

It uses a local LLM via Ollama combined with a tool-based architecture (MCP) to create a powerful, extensible personal AI assistant.

Features

AI-Generated Learning Plans
Smart Schedule Generation (No Conflicts)
Google Calendar Integration (OAuth2)
MCP Tooling (AI → Real Actions)
Fully Dockerized Setup
Local LLM (No external AI APIs required)

Screenshots

Planner Dashboard

Define your goal, configure study preferences, and manage your entire workflow in one place.

Google Calendar Sync

Approved schedules are instantly converted into real Google Calendar events.

Calendar & History View

View upcoming events, track applied schedules, and manage saved plans and sessions.

Architecture

Frontend (HTML + JS)
        ↓
Brain (FastAPI - AI Orchestration)
        ↓ MCP Calls
Backend (Go API)
        ↓
Google Calendar API

System Flow

User enters goal → Frontend
Frontend calls Backend APIs
Backend sends request to Brain (AI)
Brain uses Ollama (LLM) to generate plan/schedule
Backend validates + syncs with Google Calendar

Services Overview

Service	Purpose
frontend	UI for user interaction
brain	AI logic (plan + schedule generation)
backend	OAuth, DB, calendar sync, MCP
ollama	Local LLM inference
postgres	Persistent storage

Tech Stack

Frontend: HTML, Vanilla JS
Backend: Go (net/http)
AI Layer: FastAPI (Python)
LLM Runtime: Ollama (llama3.2:3b)
Integration: Google Calendar API
Infra: Docker + Docker Compose

Project Structure

CalSync-AI/
├── backend/        # Go API (OAuth + Calendar + MCP)
├── brain/          # FastAPI AI engine
├── frontend/       # UI (HTML + JS)
├── docker-compose.yml
└── README.md

Getting Started

Prerequisites

Docker
Docker Compose
Google Account
Ollama-compatible environment

Google OAuth Setup

Go to Google Cloud Console
Create OAuth 2.0 credentials
Add this redirect URI: http://localhost:8080/auth/google/callback
Copy your credentials into backend/.env

1. Clone the Repository

git clone https://github.com/your-username/CalSync-AI.git
cd CalSync-AI

2. Start the Application

docker compose up --build

3. Pull Ollama Model

docker exec -it calsync-ollama ollama pull llama3.2:3b

4. Access the App

Service	URL
Frontend	http://localhost:8000
Backend	http://localhost:8080
Brain	http://localhost:5005
Ollama	http://localhost:11434

Environment Variables

Create a .env file inside backend/:

GOOGLE_CLIENT_ID=your_client_id
GOOGLE_CLIENT_SECRET=your_client_secret
GOOGLE_REDIRECT_URL=http://localhost:8080/auth/google/callback

SESSION_SECRET=your_secret

DATABASE_URL=postgres://calsync:calsyncpassword@postgres:5432/calsync?sslmode=disable

FRONTEND_ORIGIN=http://localhost:8000
FRONTEND_REDIRECT_URL=http://localhost:8000/main.html

BRAIN_BASE_URL=http://calsync-brain:5005
POST /api/v1/ai/generate-learning-plan

AI Workflow

1. Generate Learning Plan

POST /api/v1/ai/generate-learning-plan

Input:

Learning goal
Estimated total hours

Output:

Structured JSON learning plan

2. Generate Schedule

Field	Value
Endpoint	`POST /api/v1/ai/generate-schedule`
Constraints	No past dates, no conflicts, one session/day, respects time window

3. Apply Schedule

Endpoint: POST /api/v1/ai/apply-schedule

Creates real Google Calendar events from the generated schedule.

Google Calendar Integration

Current Status: Moving toward app-owned multi-user OAuth

Planned OAuth Flow

User clicks Sign in with Google
Backend redirects to Google OAuth
Google returns callback to backend
Backend creates user session
Calendar actions run on behalf of the user

Current Limitations

Local/dev-oriented setup
Production deployment in progress
Depends on local Ollama availability
Multi-user support still improving

Debugging

View Logs

docker logs -f calsync-backend
docker logs -f calsync-brain
docker logs -f calsync-frontend
docker logs -f calsync-ollama

Test Backend

curl http://localhost:8080/api/v1/calendar/events
docker logs -f calsync-ollama

Test Brain

curl http://localhost:5005/health

Roadmap

Authentication & Scaling

Full multi-user Google OAuth
Persistent database-backed user/session storage
Secure production session handling

AI Improvements

Better JSON parsing reliability
Multi-week scheduling
Adaptive scheduling
Improved fallback logic

Product Expansion

Gmail integration
Notes & documents
Task management
Full productivity assistant

Deployment

HTTPS (SSL)
Reverse proxy
Production configs
Background jobs

Model Strategy

Current Model:

llama3.2:3b (Ollama)

Future:

Stronger local models
Tool-using agents
Retrieval over user data
Fine-tuning / adapters

Author

Naitik Shah
Master of Engineering in Computer Science
Oregon State University
🔗 LinkedIn
🔗 GitHub

License

MIT License

Acknowledgements

Ollama
Google Calendar API
FastAPI
Docker

Vision

CalSync AI is evolving into a full personal AI system for:

Planning
Scheduling
Calendar automation
Workflow assistance
Intelligent task execution

Name		Name	Last commit message	Last commit date
Latest commit History 127 Commits
backend		backend
brain		brain
frontend		frontend
.gitignore		.gitignore
README.md		README.md
docker-compose.yml		docker-compose.yml

Folders and files

Latest commit

History

Repository files navigation