Richard-Gidi/One-Africa-Markets

# OneAfrica Market Pulse — Automated Market Intelligence (Streamlit Demo) A lightweight, executive-ready demo that **scans news sources**, **summarizes insights**, and **ranks relevance** for **tree commodities** (cashew, shea, cocoa, palm kernel) and policy updates across Africa. > “Ask your data why, until it has nothing else to say.” — Richard Gidi --- ## Features - Curated RSS feeds (Reuters commodities, AllAfrica Agriculture, FAO, etc.) - Keyword-based **relevance scoring** using TF‑IDF (no API keys required) - **Extractive summaries** (2–6 sentences) with a simple TF‑IDF centroid method - **Impact tags** via rule-based classifier (Supply Risk, Price Upside, FX Pressure, Logistics, General) - Export **CSV** and **Markdown Digest** - Fully configurable in the sidebar (feeds, keywords, filters) ## Quickstart ```bash # 1) Create a virtual env (recommended) python -m venv .venv source .venv/bin/activate # on Windows: .venv\Scripts\activate # 2) Install dependencies pip install -r requirements.txt # 3) Run Streamlit streamlit run streamlit_app.py ``` ## Customization - **Add/Remove Feeds:** Edit `DEFAULT_SOURCES` inside `streamlit_app.py`. - **Keywords:** Use the sidebar to modify keywords (comma-separated) during runtime. - **Impact Rules:** Adjust `IMPACT_RULES` in `streamlit_app.py` to tune the tags. - **Lookback Window:** Slider in the sidebar controls how far back to fetch from each feed. ## Notes - This demo intentionally avoids heavyweight models and paid APIs, so it will run anywhere. - For production, you can: - Replace the summarizer with an LLM (OpenAI, Azure, etc.). - Add **n8n** to schedule daily runs and push digests to email/WhatsApp/Slack. - Plug outputs into **Power BI** or a **FastAPI** backend for multi-user access. - Store articles and embeddings in a vector DB (FAISS/Chroma) for semantic search. - Add entity extraction and country/commodity tagging for deeper analytics. ## Branding The app is titled **OneAfrica Market Pulse …