---
title: Student Support Chatbot
emoji: 🎓
colorFrom: purple
colorTo: indigo
sdk: docker
pinned: false
---

# 🎓 Student Support Chatbot

A fully deployed AI-powered RAG (Retrieval-Augmented Generation) chatbot that answers student queries about admissions, fees, courses, campus life, and more. Built with LangChain, Google Gemini, Pinecone, and Flask.

---

## 🏗️ Architecture

```
Student Question
      ↓
HuggingFace Embeddings (all-MiniLM-L6-v2)
      ↓
Pinecone Vector DB (semantic search → top 3 results)
      ↓
Google Gemini LLM (generates final answer)
      ↓
Flask Web App (chat interface)
```

**Dataset**: [`bot-remains/student-assistance-chatbot`](https://huggingface.co/datasets/bot-remains/student-assistance-chatbot) — 217 student Q&A pairs covering admissions, eligibility, fees, academics, and campus life.

---

## 🛠️ Tech Stack

| Layer           | Technology                           |
| --------------- | ------------------------------------ |
| LLM             | Google Gemini 2.5 Flash              |
| Embeddings      | HuggingFace `all-MiniLM-L6-v2`       |
| Vector Database | Pinecone                             |
| Framework       | LangChain                            |
| Web App         | Flask                                |
| Deployment      | AWS App Runner + Amazon ECR / Docker |

---

## 🚀 How to Run Locally

### Step 1: Clone the repository

```bash
git clone https://github.com/aak007/Build-a-Complete-Medical-Chatbot-with-LLMs-LangChain-Pinecone-Flask-AWS.git
cd Build-a-Complete-Medical-Chatbot-with-LLMs-LangChain-Pinecone-Flask-AWS
```

### Step 2: Create and activate conda environment

```bash
conda create -n medibot python=3.10 -y
conda activate medibot
```

### Step 3: Install dependencies

```bash
pip install -r requirements.txt
```

### Step 4: Set up environment variables

Create a `.env` file in the root directory:

```ini
PINECONE_API_KEY=your_pinecone_api_key
GOOGLE_API_KEY=your_gemini_api_key

# Optional: choose your index name
PINECONE_INDEX_NAME=student-chatbot

# Dataset source options: hf | local | pdf
DATA_SOURCE=hf

# If DATA_SOURCE=hf
HF_DATASET_NAME=bot-remains/student-assistance-chatbot

# If DATA_SOURCE=local (csv/json/jsonl/txt/md)
LOCAL_DATASET_PATH=data/my_dataset.csv
TEXT_COLUMNS=question,answer

# If DATA_SOURCE=pdf (single file or folder)
# LOCAL_DATASET_PATH=data/student_buddy_qa.pdf
```

### Step 5: Index the dataset into Pinecone _(Run only once)_

```bash
python store_index.py
```

This downloads the student Q&A dataset from HuggingFace and stores it as vector embeddings in your Pinecone `student-chatbot` index.

### Use Your Own Dataset

You can replace the default dataset with your own data and re-index it.

1. Put your file inside the project, for example `data/my_dataset.csv`.
2. Update `.env`:

```ini
DATA_SOURCE=local
LOCAL_DATASET_PATH=data/my_dataset.csv
TEXT_COLUMNS=question,answer
PINECONE_INDEX_NAME=my-student-bot-index
```

3. Rebuild vectors:

```bash
python store_index.py
```

4. Start app:

```bash
python app.py
```

Supported local file formats:

- `csv` (recommended)
- `json` (array of objects)
- `jsonl` (one JSON object per line)
- `txt` / `md` (split by blank lines)

For PDF knowledge bases:

```ini
DATA_SOURCE=pdf
LOCAL_DATASET_PATH=data/student_buddy_qa.pdf
PINECONE_INDEX_NAME=student-buddy-index
```

Then run:

```bash
python store_index.py
python app.py
```

Example CSV schema:

```csv
question,answer,category
What is eligibility for B.Tech?,Candidates must pass 10+2 with PCM and required cutoff,admissions
What are hostel fees?,Hostel fees vary by room type and campus policy,fees
```

### Step 6: Start the app

```bash
python app.py
```

### Step 7: Open in browser

```
http://localhost:7860
```

---

## ☁️ Deployment

### AWS App Runner + ECR (Recommended)

```bash
# 1. Authenticate Docker to AWS ECR
aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin <account-id>.dkr.ecr.us-east-1.amazonaws.com

# 2. Build Docker image
docker build --no-cache -t medical-bot .

# 3. Tag image
docker tag medical-bot:latest <account-id>.dkr.ecr.us-east-1.amazonaws.com/medical-bot:latest

# 4. Push to ECR
docker push <account-id>.dkr.ecr.us-east-1.amazonaws.com/medical-bot:latest
```

Then deploy via AWS App Runner, set **port 7860**, and add your `PINECONE_API_KEY` and `GOOGLE_API_KEY` as environment variables.

---

## 📁 Project Structure

```
├── app.py                  # Flask web server
├── store_index.py          # Downloads dataset & indexes into Pinecone
├── Dockerfile              # Docker config for deployment
├── requirements.txt        # Python dependencies
├── setup.py
├── src/
│   ├── helper.py           # load_hf_dataset(), text_split(), embeddings
│   └── prompt.py           # System prompt for student support
├── static/
│   └── style.css           # Premium student-themed UI
└── templates/
    └── chat.html           # Chat interface
```

---

## 🔑 Required API Keys

| Key                | Where to Get                                                  |
| ------------------ | ------------------------------------------------------------- |
| `PINECONE_API_KEY` | [app.pinecone.io](https://app.pinecone.io/)                   |
| `GOOGLE_API_KEY`   | [aistudio.google.com](https://aistudio.google.com/app/apikey) |