Mshi-dev15/wca-ai-tool-s11-afric

Shamba Advisor: A Streamlit-based AI agricultural tool. It uses a two-step LLM pipeline to analyze farmer queries with conversation context and generate safe, localized, step-by-step farming advice with downloadable reports. # 🌱 Shamba Advisor An AI-powered decision assistant for smallholder farmers in Kenya — covering the whole farming cycle, from land preparation and planting, through pest and disease problems, to harvest and storage. **Group:** AFRIC **Members:** Faith Mshiki, Simon Gatuku **Course:** WeCan Academy — AI & Chatbot Course, Season 11 --- ### 🚀 Live Demo Click the badge below to try the live application: --- ## What it does A farmer types a question in plain language — misspellings, mixed English/Swahili, and incomplete sentences are all fine. Shamba Advisor then: 1. **Understands the question** using conversation memory — follow-up messages like "what should I plant?" are correctly understood using earlier context (location, farm size, crop) without the farmer repeating themselves (Call 1) 2. **Pulls in live weather data** automatically, when relevant to the question 3. If multiple causes are plausible (e.g. a pest problem), **asks the farmer to pick one** from quick-select buttons, or type their own description if none fit 4. **Generates a practical, low-cost, week-by-week action plan** based on the analysis (Call 2) 5. **Saves the advice to a file**, downloadable directly from the chat 6. **Remembers the farmer** — returning to the same link signs them back in automatically, with recent chats available in the sidebar to reopen at any time ## How it works — the pipeline ``` Farmer types a message │ ▼ CALL 1 (ai/first_call.py) Analyzes intent, crop, location, symptoms, possible causes, etc. Uses conversation history to understand follow-up questions Returns structured JSON │ ▼ Weather lookup (weather/open_meteo.py) Only runs if the analysis says weather matters │ ▼ [If ambiguous] Farmer picks a cause via buttons, or types their own │ ▼ CALL 2 (ai/second_call.py) Turns the analysis + weather + selected cause into a real action plan │ ▼ Saved to a .txt file + shown in the chat Conversation stored in Supabase (encrypted at rest) ``` Both AI calls use Open …