README.md

---
title: Chat API - Multi Mode AI
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: docker
app_file: app.py
pinned: false
---


# 🤖 Multimodal AI Chatbot API (FastAPI + Hugging Face + Qwen2.5)

This project is a **FastAPI-based AI chatbot backend** designed for deployment on **Google Colab + Hugging Face Spaces**. It uses a lightweight LLM (`Qwen2.5-1.5B-Instruct`) and provides multiple domain-specific AI endpoints.


## 🚀 Features

* ⚡ FastAPI backend for high-performance inference
* 🧠 Powered by `Qwen2.5-1.5B-Instruct` (lightweight LLM)
* 🌐 Multiple specialized AI modes:

  * English assistant
  * Bangla assistant
  * Math solver
  * Physics tutor
* 🔗 REST API ready for frontend integration
* 📦 Hugging Face / Colab deployment friendly

---

## 📁 Project Structure

```
main.py        # FastAPI backend
requirements.txt
README.md
```

---

## 🧠 AI Modes / Endpoints

| Mode    | Endpoint   | Description                       |
| ------- | ---------- | --------------------------------- |
| General | `/chat`    | General AI assistant              |
| English | `/english` | Grammar, writing, correction      |
| Bangla  | `/bangla`  | Bangla language assistant         |
| Math    | `/math`    | Step-by-step problem solving      |
| Physics | `/physics` | Physics explanations & numericals |

---

## ⚙️ Installation (Local / Colab)

### 1️⃣ Install dependencies

```bash
pip install fastapi uvicorn transformers torch accelerate sentencepiece
```

---

### 2️⃣ Run server

```bash
uvicorn main:app --host 0.0.0.0 --port 8000
```

For Colab:

```bash
nohup uvicorn main:app --host 0.0.0.0 --port 8000 &
```

---

## 🌐 API Usage

### 🔹 General Chat

```bash
POST /chat
```

```json
{
  "message": "Hello, what is AI?",
  "mode": "none"
}
```

---

### 🔹 English Mode

```bash
POST /english
```

```json
{
  "message": "Correct this sentence: I has a book"
}
```

---

### 🔹 Bangla Mode

```bash
POST /bangla
```

```json
{
  "message": "বাংলা ব্যাকরণ কি?"
}
```

---

### 🔹 Math Mode

```bash
POST /math
```

```json
{
  "message": "Solve: 2x + 5 = 15"
}
```

---

### 🔹 Physics Mode

```bash
POST /physics
```

```json
{
  "message": "Explain Newton's second law"
}
```

---

## 🧾 Response Format

```json
{
  "mode": "math",
  "response": "Step-by-step solution..."
}
```

---

## 📌 Disclaimer (Included in every response)

Each response automatically includes:

> The chat interface is built by MD Saib Hossain.
> For AI systems, RAG pipelines, or Agentic AI development contact:
> [https://www.linkedin.com/in/saib-hossain-182834229/](https://www.linkedin.com/in/saib-hossain-182834229/)

---

## 🚀 Deployment (Hugging Face Spaces)

1. Create a new Space
2. Choose **Docker or Gradio SDK**
3. Upload `main.py`
4. Set runtime to Python 3.10+
5. Add dependencies in `requirements.txt`

---

## 🔥 Tech Stack

* FastAPI
* Hugging Face Transformers
* PyTorch
* Qwen2.5 LLM
* Google Colab (for development)

---

## 📈 Future Improvements

* 🔥 Streaming responses (token-by-token)
* 🔥 LangGraph-based routing
* 🔥 RAG integration (PDF + web data)
* 🔥 Memory-based chat system
* 🔥 Frontend React UI integration

---

## 👨‍💻 Author

**MD Saib Hossain**
AI/ML Engineer | Agentic AI & RAG Systems Builder

🔗 LinkedIn: [https://www.linkedin.com/in/saib-hossain-182834229/](https://www.linkedin.com/in/saib-hossain-182834229/)

---

## ⭐ If you like this project

You can extend it into:

* AI tutor system
* Multilingual chatbot
* Exam preparation assistant
* Medical/education RAG system

---

🚀 Built for scalable AI experimentation on lightweight infrastructure.