title	Semantic RAG Engine
colorFrom	blue
colorTo	purple
sdk	docker
pinned	false

Semantic RAG Engine

Overview

This project is an advanced Dual-Engine Retrieval-Augmented Generation (RAG) System. It goes beyond traditional keyword search by combining dense vector embeddings with ultra-fast LLM inference to provide three simultaneous layers of information:

1. Semantic Results: Raw, highly relevant source documents fetched via vector similarity.
2. Contextual RAG Answer: A highly synthesized, strict answer grounded exclusively in the private source data.
3. General AI Knowledge: A broad, worldly perspective drawn from the LLM's baseline training, providing context even when private data is sparse.

Designed with a premium SaaS-style user interface, the system showcases robust backend engineering, cloud API integration, and modern frontend design.

---

Technical Architecture

1. Vector Database & Embeddings

• ChromaDB: Utilized as the high-speed, local vector store for lightning-fast similarity searches.
• SentenceTransformers (all-MiniLM-L6-v2): Generates 384-dimensional dense semantic embeddings, balancing deep contextual understanding with low latency.

2. Dual-Pipeline LLM Integration

• Groq Cloud API: Powers the generation pipelines using the llama-3.1-8b-instant model.
• Concurrent Processing: The backend architecture triggers the RAG synthesis and the General Knowledge synthesis concurrently to drastically reduce user wait times.

3. Frontend & UX

• Built with pure Vanilla HTML/CSS/JS to demonstrate core frontend proficiency without relying on heavy frameworks.
• Features a custom CSS Grid layout, a retractable pull-tab sidebar, and a clean, book-style reading UI for optimal readability.

---

UI Showcase

The 3-Tier Book-Style Results Layout (Semantic Source + RAG + General AI)

Retractable Sidebar with Independent Accordion Project Metadata

---

Architecture Flow

graph TD
    %% Styling
    classDef user fill:#2563eb,stroke:#fff,stroke-width:2px,color:#fff;
    classDef frontend fill:#f1f5f9,stroke:#94a3b8,stroke-width:2px;
    classDef backend fill:#e2e8f0,stroke:#64748b,stroke-width:2px;
    classDef database fill:#10b981,stroke:#fff,stroke-width:2px,color:#fff;
    classDef cloud fill:#f59e0b,stroke:#fff,stroke-width:2px,color:#fff;

    %% Nodes
    User([User Query]):::user
    UI[Vanilla JS / CSS Frontend]:::frontend
    FastAPI[FastAPI Backend Server]:::backend
    Embed[SentenceTransformers<br>all-MiniLM-L6-v2]:::backend
    Chroma[(ChromaDB<br>Vector Store)]:::database
    GroqRAG((Groq API<br>Strict Context RAG)):::cloud
    GroqGen((Groq API<br>General Knowledge)):::cloud

    %% Flow
    User -->|Enters Query| UI
    UI -->|POST /api/search| FastAPI
    FastAPI -->|Vectorize Query| Embed
    Embed -->|Search K-Nearest| Chroma
    Chroma -->|Returns Source Context| FastAPI
    
    FastAPI -->|Simultaneous Request A| GroqRAG
    FastAPI -->|Simultaneous Request B| GroqGen
    
    GroqRAG -.->|Synthesized Answer| FastAPI
    GroqGen -.->|Broad Context| FastAPI
    
    FastAPI -->|JSON Response| UI
    UI -->|Renders 3-Tier Layout| User

Loading

---

Getting Started

1. Prerequisites

• Python 3.8+
• Docker & Docker Compose (for ChromaDB)
• A Groq API Key

2. Setup Environment

# Install dependencies
pip install -r requirements.txt

# Create your environment file
echo 'GROQ_API_KEY="your_api_key_here"' > .env

# Start ChromaDB container
docker compose up -d

3. Build the Vector Index

The data source is located in data/docs.txt. To ingest this into ChromaDB:

python embed.py          # Step 1: Generate vector embeddings
python insert_vectors.py # Step 2: Store in ChromaDB

4. Run the Application

Launch the FastAPI server:

python server.py

Open your browser and navigate to http://localhost:8500.

---

Project Structure

• server.py: The FastAPI backend handling the search logic and Groq API calls.
• embed.py: Logic for converting text to semantic vectors.
• insert_vectors.py: Script to ingest data into ChromaDB.
• public/: Contains the frontend assets (index.html, styles.css, script.js).
• docker-compose.yml: ChromaDB container configuration.

---

Author

Suprit Lenkennavar AI and Data Science Engineer