A crop season brings together land, water, weather and many small decisions. AgriSense keeps that context in one place, with the evidence and uncertainty in view.
Register a field, track its crop season and journal what happened. Explore planning, water and spending screens without turning missing observations into confident numbers. This is Team Underdawgs' agriculture prototype, preserved here as a personal contribution case study with a runnable synthetic local demo.
Try it locally · My contribution · Development · Original team repository
The real application, Auth Emulator, API and PostgreSQL running together. All records in these captures are synthetic. Unavailable recommendations remain visible.
| Need | What the prototype does |
|---|---|
| Keep the field context | Progressive setup with location, area, crop and separate consent choices; switch between owned fields |
| Remember what happened | Season journal with notes, quantities, spending and owner-authorized photo uploads/viewing |
| Check a soil card | Upload, extraction job and editable human-confirmation flow; actual extraction requires a configured model |
| Explore a decision | Planning, readiness, water and money screens carry evidence and missing-data reasons |
| Use a familiar language | Shared navigation and selected feature controls in English, Hindi, Marathi, Punjabi and Telugu; coverage/review is partial |
 |  |
Install Git and Docker with Compose v2, using Linux containers. No cloud account or provider key is needed.
git clone https://github.com/Muneer320/AgriSense.git
cd AgriSense
git switch codex/cleanup-demo
docker compose -p agrisense-demo up -d --build --waitOpen http://localhost:5176. The first build downloads dependencies. Create a local account with the fictional number +16505550123, choose Send code, and read its generated verification code:
docker compose -p agrisense-demo logs --tail=30 authAfter sign-in, add a field using synthetic details, defer the optional soil card, add a crop season and save a journal entry. Reload to see it persist. A permanent notice identifies the synthetic demo; no SMS or WhatsApp message is sent.
See the walkthrough and troubleshooting guide for emulator URLs, storage, shutdown and the actual browser integration suite. Named volumes retain local records, media and gracefully exported emulator accounts across container recreation.
Important
This prototype is not validated agricultural advice. Spray windows require reviewed evidence; soil OCR and grounded assistant answers need a configured provider. Missing data must remain missing. The local demo does not verify the former public deployment, cloud access, real farmer use or outbound WhatsApp.
flowchart LR
Farmer[Field and season records] --> Web[Next.js web app]
Web --> Auth[Firebase Auth Emulator]
Web --> API[FastAPI authenticated API]
Auth --> API
API --> DB[(PostgreSQL)]
API --> Media[Private media with expiring links]
API --> Science[Evidence and missing-data checks]
API --> Outbox[WhatsApp outbox - no local delivery]
The local demo uses a static production web export, real bearer-token verification and real database migrations. Python owns the scientific calculations; the browser renders their results. Authenticated API responses are not cached by the service worker.
Verification and development
The maintenance pass checks contracts, backend behavior, PostgreSQL migrations, the web export and a real emulator/API browser journey. The older contract-fixture browser suite is separate and is not counted as integration evidence. Exact checks, remaining lint/advisory debt and Windows commands are maintained in DEVELOPMENT.md and cleanup progress.
Use Python 3.12 with uv and Node >=22.12 (CI uses Node 24). Generated Python/TypeScript contracts come from the shared contract definitions; do not patch them manually. Keep all credentials, account exports, media and browser authentication state outside Git.
Scope, languages and integrations
- Non-English terminology remains pending human review. Some explanatory copy, journal and agronomist screens remain English; WhatsApp conversation handling defaults to English.
- Gemini OCR/assistant, forecast services, Google Cloud infrastructure and Meta delivery have implementation paths but need their own configuration and verification. The Compose setup does not load a team credential file.
- Reference review, real labeled training outcomes and model promotion require external domain work. No invented coefficients or adoption claims are used to fill these gaps.
- A production operational/security review and farmer-facing privacy policy are still needed before collecting real records. The included local-only trust-authentication setup must not be exposed publicly.
The case study explains the tradeoffs and contribution evidence. Historical integration notes and team setup are retained for their original context.
Built by Team Underdawgs: Muneer Alam, Akasha Prasad, Anoushka Wayangankar, Jaideep and Hina. Original team history and contributor identities are preserved. This fork documents Muneer's work on farmer-facing integration, authentication, conversational workflows, delivery support and presentation with specific commit evidence.
Issues and focused pull requests are welcome in this fork. Read the development guide and preserve scientific review flags and team attribution. No repository-wide license was supplied in the original project; no blanket permission to reuse team code is asserted here. Third-party dependencies and retained skill/reference files keep their own licenses.