Sanamuni123/MandiMitra

AI tool-calling agent for mandi price discovery - checks real Agmarknet price data and flags favorable selling windows. # 📈 MandiMitra An AI agent that checks real mandi (market) prices and tells a farmer or small trader whether now looks like a favorable window to sell — instead of them having to track prices manually or guess based on gut feel. ## What this project is Small farmers and traders often sell as soon as they harvest, without visibility into whether prices are currently high, low, or trending in a useful direction — they don't have time to track multiple markets themselves. **MandiMitra** is an agent that, given a question like *"Should I sell tomatoes in Guntur this week?"*, decides to look up real price data, computes a trend summary, and answers with a specific recommendation grounded in that data — not a guess. ## What it's related to - **Agentic AI / GenAI portfolio project** — unlike its siblings AnnadataAI and VaradaAI, which are RAG systems answering from static PDFs, MandiMitra is a **tool-calling agent** working with live, structured data — a deliberately different technique to show range beyond RAG. - **Domain**: agricultural market price discovery, using India's public Agmarknet dataset (mandi prices) via data.gov.in. - **Roadmap**: - **v1 (this)** — mandi price agent for agricultural commodities, English, single tool - **v2** — shrimp/aquaculture export price tracking (a second data source + tool), multi-commodity comparison, price alerts over time (not just point-in-time) - **v3** — scheduled monitoring with push notifications when a tracked commodity crosses a favorable threshold, rather than only answering on request ## Flow of the code **`prices.py`** — the data layer, no LLM involved: ``` fetch_mandi_prices(commodity, state, district) │ ▼ GET api.data.gov.in... (Agmarknet dataset) │ ▼ raw price records (date, market, min/max/modal price) │ ▼ compute_trend(records) → sort by arrival_date, pull modal prices → latest_price, avg_price, pct_vs_avg → favorable_window = latest price >= 5% above the recent average ``` **`agen …