Update README.md

README.md

@@ -1,10 +1,21 @@
+---
+title: ContextHub
+emoji: 🤖
+colorFrom: blue
+colorTo: green
+sdk: streamlit
+sdk_version: "1.33.0"
+python_version: "3.10"
+app_file: app.py
+pinned: false
+---
+
 # 🔍 ContextHub — Multimodal RAG Intelligence Platform
 
 <div align="center">
 
 ![Python](https://img.shields.io/badge/Python-3.10+-blue?style=for-the-badge&logo=python)
-![FastAPI](https://img.shields.io/badge/FastAPI-0.110-009688?style=for-the-badge&logo=fastapi)
-![Streamlit](https://img.shields.io/badge/Streamlit-1.32-FF4B4B?style=for-the-badge&logo=streamlit)
+![Streamlit](https://img.shields.io/badge/Streamlit-1.33-FF4B4B?style=for-the-badge&logo=streamlit)
 ![Pinecone](https://img.shields.io/badge/Pinecone-Vector_DB-black?style=for-the-badge)
 ![Groq](https://img.shields.io/badge/Groq-LLaMA_3.3-orange?style=for-the-badge)
 
@@ -16,8 +27,9 @@
 
 ## 📌 What is ContextHub?
 
-ContextHub is a **Retrieval-Augmented Generation (RAG)** platform that lets you upload any source of information and query it using natural language.  
-Unlike general-purpose AI, ContextHub answers **only from your uploaded content** — no hallucinations, no guesswork.
+ContextHub is a **Retrieval-Augmented Generation (RAG)** platform hosted on Hugging Face Spaces. It allows you to upload various data sources and query them using natural language. 
+
+Unlike general-purpose AI, ContextHub answers **only from your uploaded content** — ensuring zero hallucinations and total data grounding.
 
 ---
 
@@ -25,174 +37,59 @@ Unlike general-purpose AI, ContextHub answers **only from your uploaded content*
 
 | Mode | Input | Model Used |
 |------|-------|-----------|
-| 📄 PDF Analysis | Upload any PDF | Groq — LLaMA 3.3 70B |
-| 🖼️ Image Intelligence | JPG / PNG | Google Gemini Vision |
-| 🔗 Web Intelligence | Any public URL | Groq — LLaMA 3.3 70B |
-| 🎥 YouTube Intelligence | YouTube URL | Groq — LLaMA 3.3 70B |
+| 📄 **PDF Analysis** | Upload any PDF | Groq — LLaMA 3.3 |
+| 🖼️ **Image Intelligence** | JPG / PNG | Google Gemini Vision |
+| 🔗 **Web Intelligence** | Public URL | Groq — LLaMA 3.3 |
+| 🎥 **YouTube Intelligence** | YouTube URL | Groq — LLaMA 3.3 |
 
-- 🧠 **Nomic Embeddings** — semantic vector search
-- 📦 **Pinecone** — isolated namespaces per source, no data bleed
-- ⚡ **Groq** — ultra-fast LLM inference
-- 👁️ **Gemini** — multimodal image understanding
-- 🖥️ **Streamlit** — clean tabbed UI with session management
+- 🧠 **Nomic Embeddings** — semantic vector search.
+- 📦 **Pinecone** — isolated namespaces per source to prevent data bleed.
+- ⚡ **Groq** — ultra-fast LLM inference for near-instant answers.
+- 🖥️ **Streamlit** — native Hugging Face integration for a smooth UI.
 
 ---
 
-## 🧠 Architecture
+## 🧠 Architecture on Spaces
 
-### PDF & YouTube — RAG Pipeline
-```
-Input → Text Extraction → Chunking → Nomic Embeddings
-     → Pinecone (isolated namespace) → Similarity Search
-     → Top-K Chunks → Groq LLM → Answer
-```
-
-### Image — Vision Pipeline
-```
-Image Upload → Base64 Encode → Gemini Vision API → Answer
-```
-
-### URL — Web Pipeline
-```
-URL → BeautifulSoup Scraping → Groq LLM → Answer
-```
-
----
+Since this is running in a Hugging Face Space, the FastAPI backend has been integrated directly into the Streamlit logic (`app.py`) for optimized performance within a single container.
 
-## 📁 Project Structure
-
-```
-context-hub/
-│
-├── backend/
-│   ├── main.py              # FastAPI server — all routes & RAG logic
-│   └── .env                 # API keys (never commit this)
-│
-├── frontend/
-│   └── app.py               # Streamlit UI — tabs, session state, previews
-│
-├── requirements.txt         # All Python dependencies
-├── run.sh                   # One-command startup script
-├── .gitignore
-└── README.md
-```
+### The Pipeline:
+1. **Ingestion**: Documents/URLs are parsed and chunked.
+2. **Embedding**: Text is converted to vectors via Nomic.
+3. **Storage**: Vectors are stored in Pinecone using unique session namespaces.
+4. **Retrieval**: Relevant context is pulled based on user queries.
+5. **Generation**: Groq (LLaMA 3.3) generates the final response based *only* on retrieved context.
 
 ---
 
-## ⚙️ Setup & Installation
-
-### 1. Clone the Repository
-
-```bash
-git clone https://github.com/varshasharma01/Context-Hub.git
-cd context-hub
-```
-
-### 2. Install Dependencies
-
-```bash
-pip install -r requirements.txt
-```
-
-### 3. Configure Environment Variables
-
-Create a `.env` file inside the `backend/` folder:
-
-```env
-GROQ_API_KEY=your_groq_api_key
-GEMINI_API_KEY=your_gemini_api_key
-PINECONE_API_KEY=your_pinecone_api_key
-NOMIC_API_KEY=your_nomic_api_key
-```
-
-> **Get your keys from:**
-> - Groq → https://console.groq.com
-> - Gemini → https://aistudio.google.com
-> - Pinecone → https://app.pinecone.io
-> - Nomic → https://atlas.nomic.ai
+## 🔐 Environment Variables (Required)
 
-### 4. Run the Project (One Command)
+To run this Space, you must add the following **Secrets** in your Space Settings:
 
-```bash
-chmod +x run.sh    # first time only
-./run.sh
-```
-
-This starts both servers together:
-- ⚙️ FastAPI backend → `http://localhost:8000`
-- 🖥️ Streamlit frontend → `http://localhost:8501`
-
-> **Or run manually in two terminals:**
-> ```bash
-> # Terminal 1
-> uvicorn backend.main:app --reload
->
-> # Terminal 2
-> streamlit run frontend/app.py
-> ```
-
----
-
-## 🔗 API Endpoints
-
-| Method | Endpoint | Description |
-|--------|----------|-------------|
-| `POST` | `/upload` | Upload & index a PDF |
-| `POST` | `/query` | Query the indexed PDF |
-| `POST` | `/process-image` | Upload & store an image |
-| `POST` | `/query-image` | Query the image via Gemini |
-| `POST` | `/process-url` | Scrape & store a webpage |
-| `POST` | `/query-url` | Query the webpage content |
-| `POST` | `/process-youtube` | Fetch & index YouTube transcript |
-| `POST` | `/query-youtube` | Query the video transcript |
+| Secret Key | Source |
+|----------|----------|
+| `GROQ_API_KEY` | [Groq Console](https://console.groq.com) |
+| `GEMINI_API_KEY` | [Google AI Studio](https://aistudio.google.com) |
+| `PINECONE_API_KEY` | [Pinecone Dashboard](https://app.pinecone.io) |
+| `NOMIC_API_KEY` | [Nomic Atlas](https://atlas.nomic.ai) |
 
 ---
 
 ## 🛠️ Tech Stack
 
-| Layer | Technology |
-|-------|-----------|
-| Frontend | Streamlit |
-| Backend | FastAPI |
-| LLM (Text) | Groq — LLaMA 3.3 70B Versatile |
-| LLM (Vision) | Google Gemini |
-| Embeddings | Nomic Embed Text v1 / Vision v1.5 |
-| Vector Database | Pinecone |
-| PDF Parsing | PyMuPDF (fitz) |
-| Web Scraping | BeautifulSoup4 |
-| Transcripts | youtube-transcript-api |
-
----
-
-## ⚡ Key Highlights
-
-- **True multimodal RAG** — 4 input types in one platform
-- **Namespace isolation** — each source gets its own Pinecone namespace, zero data bleed between uploads
-- **Session-aware UI** — answers clear on new uploads, no stale responses
-- **Dual AI providers** — Groq for speed, Gemini for vision
-- **One-command startup** — `run.sh` boots both servers together
-
----
-
-## 🚀 Future Improvements
-
-- [ ] Chat history with memory
-- [ ] Multi-document support (query across multiple PDFs)
-- [ ] YouTube timestamp-based answers
-- [ ] Streaming responses
-- [ ] Docker containerization
-- [ ] Cloud deployment (Render + Streamlit Cloud)
+- **UI Framework:** Streamlit
+- **Orchestration:** LangChain
+- **LLMs:** Groq (LLaMA 3.3 70B), Google Gemini 1.5 Pro
+- **Vector Database:** Pinecone
+- **Embeddings:** Nomic-Embed-Text-v1
+- **Parsing:** BeautifulSoup4, PyPDF, youtube-transcript-api
 
 ---
 
 ## 👩‍💻 Author
 
-**Varsha Sharma**  
-
-[![GitHub](https://img.shields.io/badge/GitHub-varshasharma01-black?style=flat&logo=github)](https://github.com/varshasharma01)
+**Varsha Sharma** [![GitHub](https://img.shields.io/badge/GitHub-varshasharma01-black?style=flat&logo=github)](https://github.com/varshasharma01)
 
 ---
 
-## ⭐ Support
-
-If you found this useful, consider giving it a ⭐ on GitHub !
+*Note: This Space is strictly for educational purposes and adheres to data grounding principles.*