---
title: Grok-4 Chatbot
emoji: 🗲
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
---
# grok-4-chatbot(phase0)
RAG_Chatbot_Grok
Grok-4 Chatbot (Phase 0)
Live demo: https://dykemanTech-grok-rag-chatbot.hf.space
A real-time streaming chatbot powered by xAI's Grok-4 API, built with Gradio for a clean, responsive user experience.
Current Phase: Phase 0 - Basic Streaming Chatbot (Complete)
This initial version demonstrates core functionality:

Real-time token streaming for natural, typing-effect responses
Multi-turn conversation support with stable context handling
Clean Monochrome dark theme for professional look and high readability
Grok avatar using remote URL
Secure API key management via Hugging Face Secrets
Fully deployed on Hugging Face Spaces with PWA support (installable on mobile/desktop)

Upcoming Features:
Phase 1: Retrieval-Augmented Generation (RAG)

    File upload support (PDF, TXT)
    Document chunking, embedding, and vector storage (Chroma/FAISS)
    Grounded responses using user-uploaded content to reduce hallucinations
    Source citations in responses

Phase 2: AI Agent with Tool Use

    Integration of external tools (web search, code execution, etc.)
    ReAct-style reasoning loop
    Autonomous multi-step task completion

Phase 3: Evaluation Dashboard

    Automated benchmarking and metrics (hallucination detection, relevance, latency)
    Side-by-side model comparison
    Custom test sets and red-teaming

Future Enhancements

    Voice input/output
    Multi-model support (compare Grok, Claude, GPT, etc.)
    Persistent user sessions and chat history

This project is being built iteratively as a hands-on demonstration of modern generative AI engineering skills - from API integration and UI design to advanced RAG and agentic workflows.
Live Demo: https://dykemanTech-grok-rag-chatbot.hf.space
Source Code: https://github.com/ChubbyPandaStudios/RAG_Chatbot_Grok_Clean
Built with passion by a generative AI enthusiast aiming for intermediate roles in 2026. Feedback welcome!