CoolFace
Apppublic

LuciferVid/Custom_VoiceBot

sourceHugging Faceupdated 4mo agoView on Hugging Face
0likes
App README

Intelligence Core: Voice RAG System

An enterprise-grade, high-performance RAG (Retrieval-Augmented Generation) ecosystem powered by Google Gemini. This system enables professional-grade document analysis through a sophisticated, minimal voice interface.

System Architecture

🧠 Intelligence & Synthesis

  • —Core Reasoning: Google Gemini 2.0 Flash for context-aware generation and rephrasing.
  • —Vector Embeddings: Cloud-native text-embedding-004 for zero-latency, high-dimensional semantic search.
  • —Voice Pipeline: Bidirectional audio processing using Gemini STT and high-fidelity gTTS.

⚡ Performance & Stability

  • —Multi-User Session Isolation: Advanced UUID-based workspace isolation ensuring total privacy and independent context for every user.
  • —API Resilience Engine: Native exponential backoff and retry logic for high-availability document indexing under peak load.
  • —Automated Maintenance: Intelligent background pruning tasks for session memory and disk space management (24h TTL).
  • —Multi-File Concurrent Sync: Optimized for simultaneous processing of multiple documents without performance degradation.
  • —Thread-Safe Architecture: Implemented system-wide locking for the FAISS engine to prevent context corruption during concurrent writes.
  • —Cloud-Native Design: Replaced local heavy ML dependencies with cloud APIs, reducing memory footprint by 90% (<50MB RAM).
  • —Stateless Stability: Optimized for high-availability deployment on containerized environments like Render.

📁 Data Handling

  • —Multi-Format Support: Native extraction for PDF, DOCX, TXT, and MD.
  • —Context Management: Adaptive sliding-window chunking logic for precise retrieval.
  • —Vector Storage: Isolated, session-specific FAISS indices with thread-safe persistence.

Quick Connection

1. Requirements

Create a .env configuration file:

env
GEMINI_API_KEY=your_production_key

2. Infrastructure Setup

bash
pip install -r requirements.txt

3. Execution

Engine (Backend):

bash
python main.py

Interface (Frontend):

bash
streamlit run frontend/app.py

Production Deployment

The system is architected for zero-configuration deployment:

  • —Backend: Render (Python/Uvicorn)
  • —Frontend: Streamlit Community Cloud

Technical Documentation by [LuciferVid](https://github.com/LuciferVid)