oussama-souidi/darija-assistant

Assistant vocal oléicole en darija (vision + RAG) # 🫒 Olive Health Assistant — Voice-Enabled RAG & Vision A specialized AI assistant designed for Tunisian olive farmers. This project combines **Computer Vision** (CNN), **Retrieval-Augmented Generation** (RAG), and **Speech-to-Text** (ASR) to diagnose olive leaf diseases and provide grounded agricultural advice in the Tunisian dialect. --- ## 🌟 Key Features - **📸 Leaf Disease Diagnosis**: Upload a photo of an olive leaf, and our custom CNN model identifies the disease (e.g., Peacock Spot, Verticillium Wilt) with high confidence. - **🎙️ Tunisian Voice Interface**: Ask questions naturally in **Tunisian Darija**. Our optimized Whisper pipeline transcribes local dialect speech with high accuracy. - **📚 Grounded RAG Brain**: Answers are strictly grounded in technical manuals from **FAO**, **EPPO**, and **IOC**. No AI hallucinations—just verified agricultural science. - **🗣️ Natural Voice Responses**: The assistant replies in a Tunisian-accented voice, making technical advice accessible to all farmers. - **🆓 100% Free & Open**: No API keys required! The entire stack runs locally or via free open-source APIs. --- ## 🏗️ Architecture ```mermaid graph TD User[Farmer] -->|Photo| CNN[CNN Server] User -->|Voice| ASR[Whisper ASR Server] CNN -->|Label| RAG[RAG Server] ASR -->|Text| RAG RAG -->|Semantic Search| FAISS[(FAISS Vector Index)] FAISS -->|Context| Trans[Translation & Formatting] Trans -->|Arabic Text| TTS[Edge-TTS Engine] TTS -->|Audio| User ``` --- ## 🚀 Getting Started ### 1. Prerequisites - Python 3.10+ - **FFmpeg** installed and added to your system PATH (required for audio processing). - A GPU is highly recommended for the ASR and CNN servers (but CPU fallback is supported). ### 2. Installation Clone the repository and install dependencies: ```bash pip install -r requirements.txt ``` ### 3. Initialize the Knowledge Base Run the corpus builder to download technical PDFs, chunk the text, and generate the FAISS vector index: ```bash python build_corpus.py ``` …