Value Added Program · Day 2
AI Chatbot
A full-featured multimodal chatbot powered by Mistral Pixtral — with multilingual support, regional TTS, file understanding, and session memory.
Node.js
Express
JavaScript
HTML5
CSS3
SQLite
Mistral Pixtral
Edge TTS
npm
git
Features
AI Chat
Powered by Mistral Pixtral with full session memory. Understands context across the
entire conversation.
File Understanding
Upload JPG, PNG, or PDFs up to 10MB. Pixtral reads and reasons about images and
scanned documents natively.
Multilingual
Respond in 11 languages including Hindi, Marathi, Tamil, Telugu, Bengali, and more —
switchable mid-conversation.
Text-to-Speech
Edge TTS voices for English and regional Indian languages. Zero extra API calls —
runs via a local endpoint.
Session Memory
SQLite stores full chat history per browser tab. Auto-created on first run — no
setup needed.
Response Styles
Concise, Detailed, Formal, or Simple (ELI5). Injected into the system prompt — ~10
token overhead only.
Setup
01
Clone & Install
Clone the repo and install all dependencies.
git clone <your-repo-url> cd chatbot npm install
02
Configure Environment
Copy the example env file and add your Mistral API key. Get one at
console.mistral.ai.
cp .env.example .env
MISTRAL_API_KEY=your_key_here PORT=3000
03
Start the Server
The SQLite database and uploads folder are created automatically on first run.
node server.js
04
Open in Browser
Navigate to
http://localhost:3000 — the chatbot is ready.Project Structure
chatbot/
├── public/ ← Frontend (served statically)
│ ├── index.html
│ ├── style.css
│ └── js/
│ └── app.js
├── src/
│ ├── routes/
│ │ ├── chat.js ← Pixtral API + history
│ │ ├── tts.js ← Edge TTS audio
│ │ ├── upload.js ← File upload (10MB)
│ │ └── session.js ← Session cleanup
│ ├── db.js ← SQLite setup
│ └── config.js
├── docs/ ← Implementation guides
├── uploads/ ← Gitignored, auto-cleaned
├── chatbot.db ← Auto-created, gitignored
├── .env
├── package.json
└── server.js
├── public/ ← Frontend (served statically)
│ ├── index.html
│ ├── style.css
│ └── js/
│ └── app.js
├── src/
│ ├── routes/
│ │ ├── chat.js ← Pixtral API + history
│ │ ├── tts.js ← Edge TTS audio
│ │ ├── upload.js ← File upload (10MB)
│ │ └── session.js ← Session cleanup
│ ├── db.js ← SQLite setup
│ └── config.js
├── docs/ ← Implementation guides
├── uploads/ ← Gitignored, auto-cleaned
├── chatbot.db ← Auto-created, gitignored
├── .env
├── package.json
└── server.js
API Endpoints
| Method | Endpoint | Description |
|---|---|---|
| POST | /api/chat | Send message, receive AI reply with session history |
| POST | /api/upload | Upload file (JPG / PNG / PDF, max 10MB) |
| POST | /api/tts | Convert text to speech audio (Edge TTS) |
| DELETE | /api/session/:id | Clear session data and delete uploaded files |
Supported Languages
🇺🇸
English (US)
en-US-AriaNeural
🇬🇧
English (UK)
en-GB-SoniaNeural
🇮🇳
Hindi
hi-IN-SwaraNeural
🇮🇳
Marathi
mr-IN-AarohiNeural
🇮🇳
Tamil
ta-IN-PallaviNeural
🇮🇳
Telugu
te-IN-MohanNeural
🇮🇳
Bengali
bn-IN-TanishaaNeural
🇮🇳
Kannada
kn-IN-SapnaNeural
🇮🇳
Gujarati
gu-IN-DhwaniNeural
🇮🇳
Punjabi
pa-IN-OjasNeural
Dependencies
| Package | Purpose |
|---|---|
| express | Web server and routing |
| multer | File upload handling with 10MB limit enforcement |
| better-sqlite3 | Session memory and chat history (synchronous SQLite) |
| pdf-parse | Text extraction from text-based PDFs |
| edge-tts | Text-to-speech via Microsoft Edge neural voices |
| uuid | Session ID and upload filename generation |
| dotenv | Environment variable loading |