EcoFeast is my Python‑based open‑source project on GitHub that aims to autonomously redistribute surplus food from donors to NGOs.
The EcoFeast repository is written in Python and has 2 forks.
EcoFeast is production‑ready with 91/91 tests passing and CI integration.
- Freshness scoring uses XGBoost because it handles tabular features well and offers fast inference.
For a visual overview, see the observability dashboard built with Django templates and Chart.js, which shows pipeline status, agent latency, and decision trails.
I host the services via Docker Compose: Django (http://localhost:8000), PostgreSQL (localhost:5432), Redis (localhost:6379), and Qdrant (localhost:6333).
The repository details also appear on my github profile page.
- SHAP is added for explainability so NGOs can see why a freshness score was assigned.
- Groq Llama Vision handles image intake; Groq Llama 3 70B handles conversational text intake.
- SentenceTransformers (
all‑MiniLM‑L6‑v2) embeds NGO profiles; embeddings are stored in Qdrant for fast vector search combined with Haversine distance filtering. - LangGraph orchestrates the multi‑agent pipeline, providing conditional routing, self‑loops, and escalation on timeouts or rejections.
- Docker + Docker‑Compose are used for reproducible development and production deployment.