zulyadainimuhammad/poultryguard-ai

Offline AI Poultry Health & Climate Advisory Assistant for African Farmers built for the Africa Deep Tech Challenge 2026. # PoultryGuard AI Offline AI-powered poultry health, vaccination, climate, biosecurity, and farm management assistant for the Africa Deep Tech Challenge (ADTC) 2026. PoultryGuard AI runs entirely on the ADTC Standard Laptop without cloud APIs, internet access, or dedicated GPU hardware. It combines a curated Markdown knowledge base, a local FAISS vector store, and a quantised GGUF language model to deliver accurate, grounded advisory responses to poultry farmers across Africa. > **Status:** Sprint 1.5 complete — architecture refined with Query Classification Layer and Emergency Advisory Module; knowledge base schema and validator implemented. Application implementation begins in Sprint 2. --- ## Project Overview PoultryGuard AI supports poultry farmers, extension officers, students, and farm managers with local-first guidance across: - Poultry disease awareness and early symptom triage - Vaccination schedule assistance - Farm climate and housing recommendations - Biosecurity checklists - Feeding and flock management guidance - Market and operational record support --- ## Architecture PoultryGuard AI follows a clean layered architecture with strict offline-first constraints. ```mermaid graph TD subgraph UI["Frontend — Streamlit Desktop App"] Q[Question Input] --> R[Response Display] E[Emergency Alert Panel] end subgraph Backend["Backend — Application Services"] QC[Query Classifier] O[Query Orchestrator] EM[Emergency Advisory Module] end subgraph RAG["RAG Pipeline"] EMB[Embedder] --> RET[FAISS Retrieval] --> PB[Prompt Builder] end subgraph Inference["Local Inference"] LC[llama.cpp] --> GGUF[Qwen2.5-1.5B Q4_K_M] end subgraph Storage["Local Storage"] KB[Markdown Knowledge Base] VS[(FAISS Index)] end UI --> Backend Backend --> RAG RAG --> VS RAG --> Inference KB --> VS ``` **Key boundaries:** | Layer | Path | Responsibility | |---|---|---| | Frontend | `app/frontend/` | Streamlit pages and UI components | | Backend | `app/backend/` | Query orchestration …