🌴 HH GOA VIBE RAG

Multilingual Voice + Retrieval-Augmented Generation Assistant

Speak naturally. Ask in your language. Get knowledge-grounded answers.

HH GOA VIBE RAG is a multilingual Voice + RAG (Retrieval-Augmented Generation) assistant built to make AI interaction more natural, conversational, and accessible across Indian languages.

The project combines voice interaction, multilingual semantic search, FAISS vector retrieval, local AI generation, and a Goa-inspired glassmorphism interface into a single AI experience.

---

🌊 Project Vision

AI should not require people to change the way they communicate.

Instead of forcing users to type perfect English queries, HH GOA VIBE RAG aims to let users:

Speak → Understand → Retrieve → Generate → Respond

The current prototype focuses on English and Gujarati, with the architecture designed to expand toward all major Indic languages.

---

✨ Key Features

🎙️ Voice-First AI

Designed around natural human interaction.

Users can interact with the system using voice or text and receive contextual answers from the RAG pipeline.

---

🌐 Multilingual Understanding

Current language support:

• 🇬🇧 English
• 🇮🇳 Gujarati

The project is designed to expand into:

• Hindi
• Marathi
• Bengali
• Tamil
• Telugu
• Kannada
• Malayalam
• Punjabi
• Odia
• Assamese
• Urdu
• Nepali
• Konkani
• Sanskrit
• Other Indic languages

---

🔎 Semantic Retrieval

Instead of depending only on keyword matching, the system converts questions and knowledge chunks into multilingual semantic embeddings.

Current embedding model:

sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

Embedding dimension:

384

---

⚡ FAISS Vector Search

Knowledge embeddings are indexed using:

FAISS IndexFlatIP

Since the embeddings are normalized, inner-product search provides cosine-similarity-based retrieval.

This enables lightweight and fast local semantic search.

---

🧠 Local AI Generation

The current generation model is:

Gemma 3:4b

The model runs locally through Ollama, allowing the core system to operate without relying on paid external LLM APIs.

---

🧩 Smart Retrieval Pipeline

HH GOA VIBE RAG performs multiple retrieval stages:

User Question
      ↓
Query Embedding
      ↓
FAISS Vector Search
      ↓
Candidate Retrieval
      ↓
Metadata-Aware Ranking
      ↓
Duplicate Removal
      ↓
Top-K Contexts
      ↓
Gemma 3:4b
      ↓
Grounded Answer




























<div align="center">
<img width="1200" height="475" alt="GHBanner" src="https://ai.google.dev/static/site-assets/images/share-ais-513315318.png" />
</div>

# Run and deploy your AI Studio app

This contains everything you need to run your app locally.

View your app in AI Studio: https://ai.studio/apps/f833120b-57cd-46e1-a74d-76547dc8a8f2

## Run Locally

**Prerequisites:**  Node.js


1. Install dependencies:
   `npm install`
2. Set the `GEMINI_API_KEY` in [.env.local](.env.local) to your Gemini API key
3. Run the app:
   `npm run dev`