AgentIQX-Multiagentic-Gradio-App
AgentIQX 🚀
Modern Multi-Agent Audio/Video/PDF Transcription & Summarization App
💻 Tech Stack
| Layer | Technology | |——-|———–| | Frontend/UI | Gradio Blocks, Custom CSS (Glassmorphic, Dark/Cyan theme) | | Transcription | OpenAI Whisper (local GPU/CPU) | | Summarization / LLM | Ollama (TinyLlama, DeepSeek, Phi3) | | Retrieval / Search | SentenceTransformers (all-MiniLM-L6-v2), FAISS | | PDF Handling | PyPDF2, PyMuPDF | | Text-to-Speech | gTTS (local), pyttsx3 | | Email Automation | SMTP, dotenv credentials | | Utilities | Python 3.11+, Logging, Chunking, FAISS Index Management |
APP_SNAPSHOTS
✨ Key Features
1. Transcribe Anything
- Upload PDFs, audio (mp3/wav), or video (mp4/m4a/mov).
- Multi-method extraction ensures accurate transcripts via Whisper, PyPDF2, and PyMuPDF.
2. Flexible Summarization
- Generate concise, detailed, bulleted, numbered, or paragraph-style summaries.
- Focus on custom sections or areas of interest using local LLMs (Ollama).
3. Retrieval-Augmented QA (RAG)
- Ask questions about your documents.
- Relevant chunks retrieved via embeddings & FAISS.
- Answers generated by LLM with full context grounding.
4. Step-by-Step Explanations
- Automatically generate pedagogical rationales for any answer.
- Ideal for teaching, learning, or auditing LLM decisions.
5. Semantic Embedding & Chunk Management
- Text is split into word-overlapped chunks, embedded, and indexed with FAISS.
- Supports fast similarity search and RAG pipelines.
6. Text-to-Speech (TTS)
- Convert any summary or transcript to speech locally.
- Audio is playable in-browser or downloadable as MP3.
7. Email Automation
- Send summaries or transcripts via SMTP with .env-based secure credentials.
- Live validation and status feedback in UI.
8. Glassmorphic, Responsive UI
- Smooth, intuitive tab navigation.
- Dark/cyan neon-glass theme with glowing buttons.
- Forward/backward navigation between workflow tabs.
9. Robust Logging & Error Handling
- Detailed logs for each agent.
- Error propagation surfaced to UI for easy debugging.
📊 Workflow Overview
- Upload & Extract → PDF / Audio / Video → Raw text via Whisper & multi-method extraction.
- Chunk & Embed → Text split into chunks, embedded (SentenceTransformers), indexed (FAISS).
- Summarize → Select style, level of detail, focus areas → Local LLM returns summary.
- TTS & Email → Speak summary using gTTS → Email via SMTP with secure validation.
- RAG Q&A → Ask questions → Relevant chunks retrieved → Answer generated.
- Explain Agent → Generate step-by-step rationale for any answer/context.
⚡ Advanced Features
- Cached embeddings & indices for speed.
- Lazy agent loading for efficient memory usage.
- Full error propagation to UI.
- Drop-in alternative models by editing a single config line.
- Privacy-first design — no cloud calls; all processing local.
🚀 Getting Started
```bash
Create a fresh virtual environment and install dependencies
pip install -r requirements.txt
Set up .env for email credentials
echo “SENDER_EMAIL=your_email@domain.com\nSENDER_PASSWORD=yourpassword” > .env
Launch AgentIQX
python app.py