Day 6: Build a Local ChatGPT Mini Studio 🤖

🎯 Day 6 Project – Build a Local ChatGPT: GPT Mini Studio
Today’s project was about building a complete AI assistant platform from scratch that runs locally on your machine — combining LLMs, backend architecture, authentication, and a real SaaS-style product experience.
Instead of just creating a simple chatbot, this project focuses on building a full AI product ecosystem similar to ChatGPT, including login systems, chat history, credit limits, and a modern UI.

🔗 Project Code (Open Source)
GitHub 👉 https://github.com/omkargutal/GPT-Mini-Studio

🚀 What I Built
GPT Mini Studio — a local AI assistant platform that provides a ChatGPT-like experience while running completely on local hardware.
This means:
* ⚡ Fast responses
* 🔒 Full data privacy
* 💻 No cloud dependency
* 🧠 AI inference directly on your machine
The platform uses a lightweight but powerful instruction-tuned model optimized for laptops and developer machines.

🧠 Core Features Implemented
🤖 Local AI Inference Engine
* Integrated Hugging Face Transformers
* Runs the Qwen2.5-0.5B-Instruct model locally
* Hardware detection automatically switches between:
* Apple MPS GPU
* NVIDIA CUDA
* CPU fallback
Lazy-loading ensures the model loads only when needed, preventing unnecessary RAM usage.

⚡ Backend Architecture
Built using FastAPI for high-performance asynchronous APIs.
Core backend responsibilities:
* AI prompt handling
* user authentication
* credit usage tracking
* chat session management
* database interaction
Server runs using Uvicorn for fast ASGI execution.

🗄️ Database System
Implemented using SQLAlchemy ORM with SQLite.
Database structure:
users
* email
* hashed_password
* credits
* OAuth providers
chat_sessions
* session titles
* user association
messages
* prompts
* AI responses
* timestamps
This allows permanent chat history just like ChatGPT.

🔐 OAuth Authentication
Integrated secure login using:
* Google
* GitHub
* LinkedIn
Implemented using Authlib.
Users can login with one click and their profile automatically syncs with the database.

💬 Persistent Chat History
Users can:
* create multiple conversations
* reopen past chats
* view saved prompts & responses
APIs implemented:
/api/users/{id}/sessions
/api/sessions/{id}/messages

💳 Credit-Based Usage System
Simulated a real SaaS monetization model.
Users receive:
50 credits every 8 hours
System automatically refreshes credits even if the app stays open.
This replicates usage limits used by major AI platforms.

🎨 Premium Chat UI
Built a modern interface with features like:
* collapsible sidebar
* chat history navigation
* chat rename / pin / delete
* floating ellipsis menu
* glassmorphism pricing cards
* user account dropdown

🐞 Bug Reporting System
Integrated a reporting form that:
* sends user reports to backend
* stores them in feedback.log

🧠 Tech Stack
* Python
* FastAPI
* Uvicorn
* SQLAlchemy
* SQLite
* Hugging Face Transformers
* Qwen2.5-0.5B-Instruct
* OAuth Authentication
* JavaScript / HTML / CSS
* Local AI Inference

💡 Motivation
You don’t need expensive GPUs or enterprise infrastructure to build AI products.
You need:
Consistency.
Build → Experiment → Improve → Deploy → Repeat 🔁

Follow & build with me 👇
X: https://x.com/job_update_2k25
YouTube: /https://www.youtube.com/@UCg89HKyx8h_MfRxXN_zNjWg
LinkedIn: /omkar-gutal-a25935249
GitHub: https://github.com/omkargutal

📌 Tags#AIChallenge #30DaysOfAI #BuildInPublic #AIEngineer #LLM #FastAPI #HuggingFace #GenerativeAI #Python #OpenSource #MachineLearning #AIProjects #Automation #BuildWithOmkar