Photo-to-diagnosis flow
Photograph a plant or pick one from the gallery, add a typed or spoken note, and get a structured Arabic diagnosis from Gemini 2.5 Flash.
- AI Product Engineering
- In-house Product
Mashtal AI
An Arabic AI plant doctor. Photograph a sick plant, add a typed or spoken note, and get a structured Arabic diagnosis with a treatment plan and care schedule.
Launching — Google Play submission
Mobile app interface
Snapshot
Production status & delivery details.
Verified production status, stack, and maintenance model.
- Client / Context
- Mashtal AI (تشخيص النبات بالذكاء الاصطناعي) — Arabic AI plant doctor app, built by GenAILabs
- Engagement
- In-house GenAILabs product — no client engagement
- Status
- Launching. In Google Play submission; iOS not confirmed
- Technology
- Flutter/Dart, GetX, Hive · Firebase (Remote Config, Firestore) · Google Gemini 2.5 Flash · Microsoft Clarity · speech_to_text, camera, image_picker
The challenge
What problem had to be solved?
The context, constraints, and risk surface that shaped every decision.
Arabic-speaking plant owners rarely find Arabic-first help when a plant is sick. Most identification apps are English-first, so users cannot tell what is wrong or what to do about it.
Mashtal AI had to turn one photo into a clear Arabic diagnosis and a plan the owner can act on. AI diagnosis can be wrong, so results are advisory and users can challenge them. It also had to stay light on data: no accounts, with scan history kept on the device.
Contributions
What GenAILabs built.
Everything in-house except the Gemini API, Firebase and Clarity.
Structured Arabic report
Plant ID, disease and severity, a treatment plan, care schedule, pet safety and seasonal tips — 41 structured fields per scan, from one Gemini prompt and JSON schema.
Wrong-result feedback
A feedback flow lets users challenge a diagnosis and get a corrected re-analysis.
Arabic-first Flutter app
Fully Arabic and right-to-left, with voice input and an on-device scan history in Hive that works offline.
Firebase config
Remote Config delivers the Gemini key and runtime settings, so the key can be rotated without a release.
Branding & store compliance
Logo, feature graphic, store listing, Data Safety form and privacy documentation, prepared for Google Play submission.
System / solution
How the product works.
A Flutter client talking directly to Gemini and Firebase, with history stored on the device.
Screenshot placeholder
Screenshot placeholder
Screenshot placeholder
System architecture
Flutter app (GetX, Hive) → photo and optional typed or spoken note sent to Google Gemini 2.5 Flash with a fixed prompt and JSON schema → structured Arabic diagnosis (41 fields) → saved to on-device history. A feedback flow re-runs the analysis on request. Remote Config delivers the key and settings; Microsoft Clarity records sessions.
Flywheel impulse
Deployment, integration, and real-world use.
How the product is set up for launch, and what changes before it scales.
Mashtal AI is in Google Play submission, with the listing, Data Safety form and store assets prepared. It is an in-house GenAILabs product, so there is no client engagement. iOS is not confirmed.
- Integration pattern
- Photo and notes sent to the Google Gemini 2.5 Flash API; Remote Config for the key and settings; Microsoft Clarity for session insight; scan history stored on the device only
- Scale protocol
- The Gemini key is fetched through Remote Config but still reaches the client app, so Gemini calls move behind a backend before scaling
- Maintenance mode
- Maintained in-house. Monetization (freemium, with the full breakdown behind a paywall) has been discussed but is not decided or built
Outcomes / evidence
What the product does today.
Pre-launch, so only what the product does today. Traction figures follow after launch.
Per scan
41 structured fields
One photo returns 41 structured Arabic data fields
Feedback loop
Challenge & re-analyze
Users can flag a wrong diagnosis and get a corrected re-analysis
Pre-launch
Offline history, no sign-up
Scan history works offline and no sign-up is needed. Store assets and compliance documents are prepared for Play submission
Capabilities involved
How core competencies were applied.
Built and prepared for launch end to end by the GenAILabs team.
AI Product Engineering
The whole product: Flutter app, a Gemini prompt and JSON schema for structured Arabic diagnoses, voice input, on-device history and a feedback loop.
In-house delivery
Branding, store listing and compliance documents prepared by the same team, ready for Google Play submission.
Related work
Other production builds.
Two more apps from the same team.
AI Product EngineeringProduction
Smart Teacher
Arabic AI tutor that explains homework from a photo — step-by-step answers, an AI teacher chat, lessons and a quiz mode.
AI Product EngineeringProduction
Kaya Health Coach
AI functional medicine coach live on both app stores — Flutter app, FastAPI backend, real-time LLM streaming, structured health journeys.
Let's talk
Have a similar challenge?
Let's discuss the problem you're trying to solve.
Talk to our team
