|
|
--- |
|
|
title: "RAG Chatbot — GPT-5.1 + URLs / Files / Text" |
|
|
emoji: 🧠 |
|
|
colorFrom: indigo |
|
|
colorTo: blue |
|
|
sdk: gradio |
|
|
sdk_version: 5.49.1 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
short_description: GPT-5.1 RAG chatbot for URLs, files, and text sources |
|
|
--- |
|
|
|
|
|
# 🔍 RAG Chatbot — GPT-5.1 + URLs / Files / Text |
|
|
|
|
|
A universal Retrieval-Augmented Generation (RAG) chatbot powered by **OpenAI GPT-5.1** and **Gradio 5.49.1**, enabling users to inject their own knowledge in real time — via **URLs**, **uploaded files**, or **raw text blocks** — and query it conversationally. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🚀 Features |
|
|
|
|
|
- **OpenAI API key input** stored per session (no backend storage) |
|
|
- **GPT-5.1** model for chat; **text-embedding-3-large** for vector retrieval |
|
|
- Accepts: |
|
|
- Live **URLs** (auto-fetched text) |
|
|
- Uploaded **files** (`.txt`, `.md`, `.csv`, `.json`) |
|
|
- Custom **text areas** for notes or raw data |
|
|
- Interactive **presets** that auto-fill working examples (ZEN sites, policy QA, etc.) |
|
|
- **Retrieval debugging info** displayed transparently for testing relevance |
|
|
- **No external DB** — fully in-memory RAG layer for portability |
|
|
- Works in **Hugging Face Space**, **Google Colab**, or **local Python** |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧩 How It Works |
|
|
|
|
|
1. Enter your **OpenAI API key** and click **Save**. |
|
|
2. Add knowledge sources (URLs, uploads, or text). |
|
|
3. Click **Build / Refresh Knowledge Base**. |
|
|
- Text is chunked → embedded via `text-embedding-3-large`. |
|
|
- Stored in memory as a lightweight vector index. |
|
|
4. Ask questions — the system retrieves relevant chunks and constrains GPT-5.1 to answer **only from those** and your **system instructions**. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧱 File Structure |
|
|
|
|
|
|
