FasalNama is a farm record investigator powered by open-weight AI. Farmers log irrigation, fertilizer, spraying, weather and crop observations by voice (Urdu), text (Urdu, Roman Urdu or English) or photo. When a crop problem appears, FasalNama builds an evidence-cited timeline of what changed in the field beforehand, flags missing records, and prepares questions for an agricultural expert. It never diagnoses: a code-based verifier rejects any uncited or cause-as-fact claim before a report is shown. Open-weight models (Whisper, Qwen or Llama, and a vision model) handle speech, language and photos, while plain code handles dates, gaps and numeric checks. It ships as a reusable Agent Skill plus an original harness and a web app that runs fully locally or on Render.