Kamerplanter is a self-hosted plant lifecycle management system for indoor and outdoor growing — covering everything from seed to harvest.

Whether you're a home grower managing a grow tent, a houseplant owner trying to keep your plants alive, a hobby gardener planning raised beds and crop rotations, or running a community garden with shared responsibilities — Kamerplanter adapts to your experience level and scale. It supports vegetables, herbs, houseplants, and ornamentals with nutrient planning, growth phase tracking, adaptive care reminders, photo-based plant identification, a knowledge assistant, and Home Assistant integration.

• One place for everything — master data, growth tracking, nutrient plans, pest management, and harvest in a single system instead of scattered spreadsheets and notes
• Knows your plants — growth phase state machine with GDD (Growing Degree Days), VPD (Vapor Pressure Deficit), and photoperiod targets guides you through each stage. Perennial cycles and crop rotation built in.
• Takes care of reminders — adaptive care schedules learn from your confirmations, adjust to seasons and hemispheres, and cover 9 care presets from tropical to cactus
• Calculates your nutrients — fertilizer mixing with EC budgets, mixing order safety (CalMag before sulfates), flush protocols, tank management with tap/RO/mixed water sources
• Protects your harvest — integrated pest management with Karenz safety intervals that block premature harvest, resistance tracking to prevent overuse of treatments
• Connects to your smart home — Home Assistant custom integration for sensor data import and actuator control, closing the monitoring-to-action loop
• Answers your questions — RAG-based knowledge assistant with pluggable LLM backends (Anthropic, Ollama, OpenAI-compatible) for plant care advice grounded in your data
• Scales from windowsill to community garden — multi-tenancy with role-based access (admin/grower/viewer), personal and shared gardens, invitation system
• Adapts to your skill level — beginner/intermediate/expert modes control UI complexity, navigation depth, and form field visibility
• Identifies plants from a photo — snap a picture and Kamerplanter names the species (and diagnoses leaf diseases) in seconds, then links it to your master data. Built right into onboarding, so you go from photo to full care setup in under 30 seconds. Privacy-first and optional: a self-hosted recognition engine is the goal, with the Pl@ntNet free tier as a fallback — no cloud account required.
• Enriches your data automatically — GBIF and Perenual adapters fill in botanical details, CSV import for bulk data, iCal export for your calendar app
• Self-hosted and private — runs on your own hardware, no cloud dependency, GDPR-aware design with retention policies

Kamerplanter is composed of the following services. Core services start by default; optional services are enabled via Docker Compose profiles (vectordb, ollama, timescaledb) or their dedicated Skaffold modules.

cp .env.example .env      # configure passwords
docker compose up --build  # start all core services

Open http://localhost:8080 and log in with demo@kamerplanter.local / demo-passwort-2024.

Optional services (AI assistant, time-series) can be enabled via Docker Compose profiles. For Kubernetes-based development with hot-reload, see the Skaffold setup below.

Prerequisites: Docker, Skaffold, a local Kubernetes cluster (Kind/k3s/minikube), Task (optional), Node.js 25+, Python 3.14+.

task setup                                        # create Kind cluster
skaffold dev --trigger=manual --port-forward      # full stack dev loop

Common tasks via Taskfile:

task test:backend       # pytest (821+ tests)
task test:frontend      # vitest (198+ tests)
task test:e2e           # Selenium E2E
task lint:backend       # ruff
task lint:frontend      # ESLint
task ha:deploy          # deploy HA integration to pod
task docs:serve         # MkDocs local preview
task claude             # Claude Code + nolte portfolio plugins loaded

task claude starts Claude Code with the nolte-shared + nolte-engineering portfolio plugins loaded from a local claude-shared checkout (override its path with NOLTE_CLAUDE_SHARED). See CLAUDE.md §"Claude Code plugin adoption".

Full documentation (architecture, API reference, guides) is available at nolte.github.io/kamerplanter.

Kamerplanter is complemented by companion projects, each with its own focus:

• kamerplanter-ha — Home Assistant custom integration (HACS). Surfaces plant phases, nutrient dosages, tank levels, and care tasks as native HA entities, services, and Lovelace cards, so dashboards and automations can build on your grow without leaving Home Assistant.
• kamerplanter-android — Native Android (Kotlin) companion app. Connects a USB (UVC) microscope for sharp pest macro shots and feeds them into Kamerplanter's upload and pest-identification flow — external UVC cameras aren't reachable from web or cross-platform camera APIs.
• kamerplanter-goose — Shareable Goose recipes that wire the Kamerplanter and Home Assistant MCP servers into repeatable, parameterized plant-care agent runs ("what needs watering today") invoked by name.

Feature branches from develop, PRs against develop. Prefixes: feature/, fix/, chore/, docs/.

This project is a vibe coding experiment — built almost entirely through conversational AI prompting with Claude Code. The specifications, architecture, domain models, backend, frontend, Helm charts, E2E tests, and documentation were all developed this way.

MIT License. See LICENSE for details.