Tendril: a garden assistant you never have to open

Description and use

Tendril delivers a weather‑aware, low‑friction garden schedule that runs locally, costs nothing, and requires no account, enabling users to focus on gardening rather than app usage.

Problem

Garden owners need a low‑friction assistant that plans watering and care based on weather, without forcing them to open an app.

Outcome

Tendril delivers a weather‑aware, low‑friction garden schedule that runs locally, costs nothing, and requires no account, enabling users to focus on gardening rather than app usage.

Workflow

• 1. User sets city and adds plants via QR code or manual entry.
• 2. Every Sunday at 7 AM the server reads the Open‑Meteo forecast and uses Qwen 2.5 (3B) via Ollama to generate a JSON plan of up to two tasks per day, each under six words.
• 3. The plan is rendered as a single image using Pillow, formatted for a phone lock screen.
• 4. Each morning the app checks the overnight weather; if rain or heat spikes occur it rewrites only that day’s tasks.
• 5. The user can tap "Done" or speak a note; faster‑whisper converts speech to text, a keyword checker verifies watering and plant names, and the plan is updated accordingly.
• 6. If a task remains pending after 6 PM the app sends one gentle reminder line.
• 7. The system validates every plan with Pydantic; if validation fails it retries or falls back to a rule‑based planner.
• 8. All data (plants, history, city) stays on the local machine; only coordinates are sent to Open‑Meteo.
• 9. Docker Compose runs the entire stack with a single command.
• 10. The user can swap the local model by changing the OLLAMA_MODEL environment variable.
• 11. The app runs quietly every morning, costing nothing beyond local compute.
• 12. The user receives a concise, weather‑aware schedule that sometimes advises "don't water" when rain is forecast.

Evidence

It plans the week from the weather. Every Sunday at 7 AM it reads the forecast, and a small open model running on my own machine writes a plan: at most two tiny tasks a day, each under six words.

Source

It checks the sky every morning. If rain shows up overnight, or a heat spike, it quietly rewrites just today's tasks.

Source

The plan is one picture. It's made to fit a phone's lock screen, with the top third left empty for the clock.

Source

Tell it what you did. Tap Done, or hold the mic and say something like "Watered the chillies, there are aphids under the tulsi leaves."

Source

No accounts. Set your city, add your plants, scan a QR code, add it to your home screen. Done.

Source

Python and FastAPI for the app, with SQLite for the garden's memory.

Source

Qwen 2.5 (3B), an open-weight model, running locally through Ollama. It writes the weekly plan and reads my notes.

Source

faster-whisper, an open speech model, turning voice notes into text right on the CPU.

Source

Open-Meteo for the forecast and city search. Free, and no API key.

Source

Pillow to draw the plan picture. Every icon in it, the sun, the clouds, the little leaf, is drawn in code.

Source

Docker Compose to run it all with one command.

Source