fadelkora21/allo-nutri-adtc-2026

Offline AI-powered least-cost poultry feed formulator for smallholder farmers in Benin # ALLO NUTRI Feed Formulator — ADTC 2026 ALLO NUTRI is an offline-first poultry-feed formulation prototype designed for smallholder farmers in Benin. It combines: - a deterministic least-cost feed optimizer; - poultry nutritional and ingredient-safety constraints; - a compact local GGUF language model; - offline explanations in French, English and Hausa. The language model explains verified optimizer results. It does not calculate or modify ingredient quantities. ## Submission information - Project ID: `allo-nutri-feed-formulator` - Domain: Agriculture - Runtime: `llama.cpp` - Model: Qwen2.5 1.5B Instruct - Quantization: GGUF Q4_K_M - Operation: 100% offline after model download ## Repository structure ```text . ├── web/ │ └── index.html ├── app.py ├── optimizer.py ├── benchmark.py ├── download_model.sh ├── setup_model.sh ├── metadata.json ├── REPORT.md ├── EVALUATION_PROMPTS.md ├── requirements.txt ├── test_optimizer.py └── validate_submission.py ``` The `model/` directory is created automatically by `download_model.sh`. Model weights are not committed to Git. ## System requirements Recommended target environment: - Ubuntu 22.04 or later - 4 CPU cores - 8 GB RAM - Python 3.10 or later - `curl` - `git` - `cmake` - a C/C++ compiler ## Install Python dependencies ```bash python3 -m pip install -r requirements.txt ``` ## Download the model Run: ```bash bash download_model.sh ``` The script downloads the public GGUF model to: ```text model/allo-nutri-qwen2.5-1.5b-q4_k_m.gguf ``` The script: - requires no private credentials; - creates the `model/` directory automatically; - supports resumed downloads; - verifies the GGUF file header; - avoids downloading the model again when a valid file already exists. ## Install llama.cpp For automatic local setup, run: ```bash bash setup_model.sh ``` Then apply the environment variables displayed by the script, or define them manually: ```bash export LLAMA_CLI="$PWD/llama.cpp/build/bin/llama-cli" export AL …