Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Garden Trials is an AI gardening app designed to stop guessing, not to diagnose plants. You write a dated observation, a local Gemma model turns it into plain facts and suggests one thing to check, then you watch the plant or make one controlled change. The app records counts and a photo, calculates a verdict and an evidence level, and gives you a next step. Its most important output is behavioral: close the laptop and look at the garden.
Built with Streamlit, Ollama, a local JSON file and the gemma3:1b model, the project runs on an ordinary computer without an API key or cloud AI call. That makes it private and reproducible, but also exposes the limits of a small local model: in real use, readings were rejected, a question was blocked, and some surviving questions were too generic to be useful.
As an Amazon Associate I earn from qualifying purchases.
What is Garden Trials?
Garden Trials is a local plant-observation and experiment journal created by Shubhra Pokhariya. Its guiding loop is: “Look first. Change one thing. Let the garden answer.” The app is for gardeners learning to read plants and for anyone who wants to test an assumption before changing a care routine.
It deliberately avoids the usual promise of an AI plant-identification or disease-diagnosis tool. The project’s rule is explicit: it does not say what is wrong with a plant and does not claim what caused a result. Instead, it helps you create a small, auditable observation.
#1 Best Overall
- Spacious Design: Measuring 21.1" wide and 14.1" deep, our lap desk comfortably fits most laptops up to 15.6". Extra room for accessories ensures convenience.
- Enhanced Functionality: Packed with handy features, including a 5x9" precision tracking mouse pad and a built-in phone slot for seamless work or video calls. Plus, enjoy ergonomic support with the integrated cushioned wrist rest.
- Cool Comfort: Enjoy a stable surface with our lap desk's dual bolster cushion, designed for comfort and airflow, keeping your lap cool during extended use.
- Durable Surface: Work with confidence on our lap desk's solid surface, featuring a sleek black carbon color, ensuring optimal air circulation to prevent your laptop from overheating.
- On-the-Go Convenience: With an integrated handle and lightweight design (2.8 lbs), our lap desk is portable for travel or moving around the house, offering flexibility in any space.
- Write what you actually saw, with a date.
- Translate the note into plain, checkable facts.
- Choose one question or one change.
- Keep a similar plant unchanged when possible.
- Count the result, add a photo and let ordinary Python calculate the report.
The laptop supplies structure and reminders. The evidence still has to come from the garden.
How does the app turn a note into a garden trial?
The workflow has a strict order so that an attractive AI explanation cannot replace an observation.
- Record a dated note. Describe visible facts such as open flowers, buds, stem position or leaf count. The date matters because later entries may be backfilled and marked as such.
- Extract plain facts. Gemma reads the note and returns a constrained interpretation. If the model drops a number, the app rejects its reading and keeps the writer’s exact words.
- Receive one check. The model proposes one specific thing to inspect. It is not asked to name a disease or infer a cause.
- Watch or choose one change. You may observe without intervening, or select one option from a short, plant-specific menu. The menu can exclude unsuitable actions; for example, “water more” is not offered for bougainvillea.
- Keep a comparison. A similar plant should remain unchanged whenever practical. This gives the trial a reference instead of treating every change in one plant as proof.
- Record counts and a photo. The app stores the observation in a local folder and produces a table and chart.
- Read the report. Verdicts, evidence levels from 0 to 3 and the next step are calculated in code rather than invented by the language model.
This design separates language tasks from decisions. Gemma handles reading and suggesting; Python decides whether the data is complete and how much evidence the trial earned.
What does the app refuse to let the model do?
Reliability-critical behavior stays outside Gemma because model output is treated as untrusted input.
Numbers cannot silently disappear
A number check compares the model’s reading with the writer’s note. If a count is missing, the interpretation is rejected and the original wording is retained. This prevents a polished summary from quietly changing “four buds” to “some buds.”
Added details must match the note
Details that Gemma introduces are filtered through a word-match check. The guard is intended to stop the model from adding observations that were never written.
Free tools Windows power users keep installed
One-click scans. No signup required.
Causes stay unknown
The cause field is forced to unknown. Even when a plant improves after a change, the app does not permit the model to declare that the change caused the improvement.
Rank #2
- 5-in-1 Connectivity: Equipped with a 4K HDMI port, a 5 Gbps USB-C data port, two 5 Gbps USB-A ports, and a USB C 100W PD-IN port. Note: The USB C 100W PD-IN port supports only charging and does not support data transfer devices such as headphones or speakers.
- Powerful Pass-Through Charging: Supports up to 85W pass-through charging so you can power up your laptop while you use the hub. Note: Pass-through charging requires a charger (not included). Note: To achieve full power for iPad, we recommend using a 45W wall charger.
- Transfer Files in Seconds: Move files to and from your laptop at speeds of up to 5 Gbps via the USB-C and USB-A data ports. Note: The USB C 5Gbps Data port does not support video output.
- HD Display: Connect to the HDMI port to stream or mirror content to an external monitor in resolutions of up to 4K@30Hz. Note: The USB-C ports do not support video output.
- What You Get: Anker 332 USB-C Hub (5-in-1), welcome guide, our worry-free 18-month warranty, and friendly customer service.
Pest and disease speculation is blocked
Questions about pests or disease are blocked unless the gardener raised those subjects. This narrows a common route to confident but unsupported diagnoses.
Choices come from a fixed menu
The model cannot invent an open-ended treatment. It selects from options filtered for the plant. That makes trials more comparable and avoids offering an obviously unsuitable action.
Files are protected during saves
Saves use a temporary file and a swap process. If a garden file is unreadable, it is moved aside instead of overwritten, preserving the damaged file for recovery.
Recommended Free Tools
What do the evidence levels mean?
The app reports an evidence level from 0 through 3. The implementation clearly defines the strongest and weakest cases; the exact thresholds for the middle levels are implementation details rather than universal scientific grades.
| Level | What the project establishes | How to use it |
|---|---|---|
| 0 | No comparison plant and no controlled change, as in a simple backfilled observation. | Treat the entry as a baseline or record, not as a test of an intervention. |
| 1–2 | Intermediate evidence levels calculated by the app; the project description does not publish a separate threshold for each level. | Use the report as a prompt for another, better-controlled observation rather than a final causal claim. |
| 3 | Two finished trials of the same change, each with a comparison plant, no other change and the same direction of result. | This is the app’s strongest repeatable signal, still not a guarantee that one factor caused the result. |
A high level means the observation design met the app’s rules. It does not turn a small home-garden trial into a controlled scientific experiment with every confounder removed.
What happened in the author’s real garden?
The project reports both automated tests and imperfect live use. That distinction is important: passing software tests does not mean every prompt is useful in a garden.
Petunia: a clear baseline, not a treatment result
From October 7 through October 10, the petunia record showed:
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Rank #3
- Note: Not suitable for MacBooks released after 2023 or devices with a protruding front camera; Not applicable to full-screen or notch-style tempered glass screen protectors; Do not use on the rear camera of the phone.
- 💻 Why Do You Need a Webcam Cover Slide? — Safeguard your privacy by covering your webcam with our reliable webcam cover when not in use. Don't let anyone secretly watch you. Stay protected!
- ✅ Thin & Stylish — Enhance your laptop's functionality and aesthetics with our 0.027" ultra-thin webcam covers. Seamlessly close your laptop while adding a touch of sophistication.
- ✅ Fits Most Devices — Compatible with laptops, phones, tablets, desktops! Keep your privacy intact on Ap/ple, Mac/Book, iPh/one, iP/ad, H/P, L/novo, De/ll, Ac/er, As/us, Sa/msung devices.
- ✅ 365 Days Protection — Our upgraded 3.0 adhesive ensures a strong hold that won't damage your equipment. Experience reliable, long-term privacy protection day in and day out.
| Date | Open flowers | Buds |
|---|---|---|
| October 7 | 2 | 4 |
| October 8 | 2 | 4 |
| October 9 | 2 | 4 |
| October 10 | 4 | 2 |
The app marked the entries as backfilled and assigned evidence level 0 because there was no comparison plant and no controlled change. The final count is useful as a record of what happened, but it cannot show why the numbers changed.
Marigold: an interesting before-and-after
On October 5, the marigold had about three open flowers and three buds. On October 9, after it was moved from a sheltered corner into open sun, the count was seven open flowers and five buds. The photos and conditions do not establish that the move caused flowering; weather, plant variation and timing could also matter.
Bougainvillea cuttings: an unfair comparison
Five cuttings shared soil, bags and a bench. One flowered and four did not. However, the cuttings were different varieties and came from different mother plants, so the result was not a fair test of a treatment. A stronger design would use same-variety cuttings taken at the same time from different mother plants.
How reliable was the local model in practice?
The project reports 85 tests in 2026. The suite includes deliberately bad model answers—garbage output, dropped numbers, invented pests and off-menu choices—as well as end-to-end tests that drive the real pages. None of those tests requires Ollama, so they test the application’s guards and workflow without depending on model availability.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Live garden use was more mixed. Gemma’s readings for three notes were rejected by the number check, and one of three questions was blocked. The two questions that passed were generic; one asked whether a stem was straight even though that did not match the author’s observation. The guards stopped invented pests, but they could not guarantee that a small model’s surviving question would be useful.
The checks cover anticipated failures, not every failure. They also rely on English word matching, which does not provide the same protection for Hindi notes. A gardener writing in Hindi should treat the safety layer as less complete than the English path.
How do I run Garden Trials on my own laptop?
The documented quickstart targets Python 3.11 or newer, Ollama, Streamlit and the project’s requirements file. The model runs locally, so the first setup needs internet access to download packages, Ollama and Gemma; normal use does not require a cloud AI call.
Rank #4
- Anti-Slip Surface - Transform your laptop into a mobile workstation with the AboveTEK portable laptop lap desk. The anti-slip surface provides a strong grip for laptops up to 15.6 inches(Diagonal), while the double rubber strip on the bottom ensures a stable display or typing experience on your lap, couch, or bed.
- Retractable Mouse Pad - Retractable laptop mouse pad extends on both directions for the left/right handed with elevation along the edges for stopping mouse from falling off. The size of laptop tray is 14" X 9.7" and the size of mouse pad is 7.4" X 6.1".
- Effective Heat Shield - The effective heat shield made of sturdy and thick material protects your laptop from overheating. Prioritizes your comfort and safety, an ideal lap pad or board for working anywhere.
- EASY to Carry and Store - With an ergonomic and simplistic design, the lap desk is portable to store in a backpack. Only 15" in size, 2.2 lb of weight and with slim 0.6 inch thickness, it is ready to be easily carried around.
- Widely Applicable - The smooth platform accommodates laptops and tablets up to 15.6 inches(Diagonal), making it a versatile accessory and one of the best gifts for mom, dad, students and professionals. Perfect for use as a laptop bed tray or tablet holder anywhere at home, library, or park.
- Install Python 3.11 or newer.
- Install Ollama for your operating system.
- Install the dependencies from the project’s requirements file.
- Download the local model with
ollama pull gemma3:1b. - Start the app with
python -m streamlit run app.py. - On Windows, use the included
run.bat; on macOS or Linux, userun.shif you prefer the project’s launch scripts. - To open the sample garden, set
GARDEN_DIR=examples/my-gardenbefore launching.
The author selected the 1B Gemma model to test whether the idea could work on an ordinary laptop without a cloud account. On a laptop CPU, an answer can take up to a minute, so the interaction is intentionally slower than a typical hosted chatbot.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →What does local operation mean for privacy and cost?
Ollama runs Gemma on the user’s own laptop. There is no API key, usage bill or cloud AI call. Notes and photos stay in a local folder, and the author says photo location data is removed. Streamlit anonymous usage statistics were disabled.
Local does not mean “no setup.” You still need a one-time internet connection for the software and model downloads, enough disk space for them and a machine capable of running the model. Once installed, the garden records can be kept without sending them to an AI service.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should I design a one-change gardening test?
Start with a baseline, change one variable and leave everything else as consistent as you can. The comparison plant should be as similar as practical, and both plants should be observed on the same schedule.
Use a fair comparison
For cuttings, match variety, timing and source as closely as possible. The bougainvillea example shows why: different varieties and mother plants make a one-versus-four result uninterpretable.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Define the count before you begin
Decide whether you will count open flowers, buds, leaves or another visible measure. Record the same measure each time instead of changing the definition after seeing the result.
Do not stack interventions
If you move a plant into sun, do not also change watering, fertilizer and pot position during the same trial. A result after several simultaneous changes cannot identify which change mattered.
Best Value
- Spacious Design: Measuring 21.1" wide and 12" deep, our lap desk comfortably fits most laptops up to 15.6". Extra room for accessories ensures convenience.
- Enhanced Functionality: Packed with handy features, including a 5x9" precision tracking mouse pad and a built-in phone slot for seamless work or video calls. Plus, enjoy laptop support with the integrated device ledge.
- Cool Comfort: Enjoy a stable surface with our lap desk's dual bolster cushion, designed for comfort and airflow, keeping your lap cool during extended use.
- Durable Surface: Work with confidence on our lap desk's solid surface, featuring a blush pink color, ensuring optimal air circulation to prevent your laptop from overheating.
- On-the-Go Convenience: With an integrated handle and lightweight design (2.14 lbs), our lap desk is portable for travel or moving around the house, offering flexibility in any space.
Repeat in the same direction
The app’s level-3 rule requires two finished trials of the same change, each with a comparison plant, no other change and the same direction of result. Repetition is what separates a promising observation from a one-off coincidence.
What are the project’s limits?
- It cannot verify behavior. The app can prompt you to go outside, but it cannot prove that you actually inspected the plant.
- A small model may ask a useless question. Guards can reject bad structure and unsupported details, not guarantee horticultural insight.
- Evidence is local and incremental. A home trial can show that a pattern repeated under recorded conditions; it cannot establish a universal gardening rule.
- Language coverage is uneven. English word matching is not equivalent protection for Hindi notes.
- CPU latency affects the experience. A response may take up to a minute on a laptop CPU, which is a deliberate trade-off for local processing.
These limits are part of the app’s argument. A tool that admits “unknown” and sends you back to the plant is more honest than one that produces a confident diagnosis from a single photograph.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesWho should use an app like this?
Garden Trials fits gardeners who enjoy keeping records, beginners who want a repeatable way to observe plants and privacy-conscious users who do not want notes or photos uploaded. It is less suitable for someone seeking instant pest identification, a guaranteed treatment or a fully automated care schedule.
Its distinctive value is not the 1B model alone. The value is the surrounding discipline: fixed choices, comparison plants, count checks, explicit uncertainty and a report that ends with a next observation. The laptop is useful only until it has helped you decide what to look at next.
Frequently Asked Questions
Can Garden Trials prove that sunlight caused the marigold to flower?
No. The move into open sun happened before the higher count, but the recorded photos and conditions do not rule out weather, timing or plant differences. A repeated trial with a comparable plant and no other change would provide stronger evidence, not absolute proof.
What happens to a garden file if it becomes unreadable?
The save process moves the unreadable file aside instead of overwriting it. That preserves the damaged copy for inspection or recovery while allowing a fresh save to proceed.
Can I use Hindi notes with the same safety guarantees as English notes?
Not according to the project’s reported limitations. The word-match guard was designed around English and does not provide the same protection for Hindi, so Hindi entries deserve extra manual checking.
Why does the app sometimes keep my exact wording instead of a cleaner AI summary?
When the model drops a number or fails a validation check, preserving the original note prevents an apparently tidy summary from changing your observation. The unedited wording becomes the audit trail for the trial.
What should I do when the app gives a generic question?
Treat it as a prompt to inspect the plant, not as a conclusion. Write a more specific observation, keep the comparison plant unchanged and start another entry rather than forcing an intervention based on an irrelevant question.
The Bottom Line
Garden Trials is best understood as a local experiment notebook with AI-assisted prompts. Its strongest feature is the refusal to diagnose or invent causes: it records what you saw, tests one change and makes uncertainty visible before sending you back outside.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
