0xkhingx/premortem-labs

AI protocol validator for biomedical research labs in Nigeria and West Africa. Catches protocol failures before they cost animals, time, and money using local LLM inference (Ollama + Gemma). # premortem-labs AI protocol validator for biomedical research labs in Nigeria and West Africa. Catches protocol failures before they cost animals, time, and money using local LLM inference (Ollama + Gemma). # PreMortem Labs AI protocol validator for biomedical research labs in Nigeria and West Africa. ## Problem Biomedical research in African universities fails due to constraints Western literature doesn't address: - Variable reagent purity from local suppliers - Power instability affecting storage - Limited equipment calibration access - No pilot testing budgets This tool catches these failures before they cost animals, time, and money. ## Quick Start ### Prerequisites - Python 3.10+ - Node.js 18+ - Ollama (ollama.ai) ### Setup 1. **Pull the Gemma 3 model:** ```bash ollama pull gemma3:4b ``` 2. **Start Ollama:** ```bash ollama serve ``` 3. **Start backend (new terminal):** ```bash cd backend pip install -r requirements.txt python -m uvicorn main:app --host 0.0.0.0 --port 8000 --timeout-keep-alive 600 ``` 4. **Start frontend (new terminal):** ```bash npm install npm run dev ``` 5. **Open browser:** ## Usage 1. Paste your research protocol 2. Click "Validate" 3. Get risk score (CRITICAL/HIGH/MEDIUM/LOW) 4. See specific flags and corrections 5. Ask Copilot clarifying questions ## Example Protocols **Bad Protocol (CRITICAL risk):** "We are inducing Type 2 diabetes in Wistar rats using STZ 95mg/kg and Nicotinamide 65mg/kg administered intraperitoneally..." **Good Protocol (MEDIUM risk):** "We are conducting a pilot study on anti-inflammatory effects of Azadirachta indica extract..." ## Architecture - **Frontend:** React + TypeScript + Vite + Tailwind - **Backend:** FastAPI (Python) - **AI:** Gemma 3 4B via Ollama (local inference, offline-first) ## License MIT