Dwell — AI Property Management Assistant

Dwell helps triage tenant maintenance requests using an on-device LLM (Ollama llama3) and suggests nearby vendors via the Google Maps Places API.

Highlights

• Ollama llama3 integration for maintenance triage and short tenant responses
• Google Maps Places API integration for nearby vendor suggestions
• FastAPI backend with async background processing
• React frontend for tenant reporting and manager dashboard

Screenshots

Tenant view:

Manager dashboard:

Getting Started

Backend (PowerShell):

cd backend
python -m venv .venv; .\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

# Optional configuration (defaults shown)
$env:OLLAMA_BASE_URL = "http://localhost:11434"
$env:OLLAMA_MODEL = "llama3:latest"
$env:OLLAMA_TIMEOUT_SECONDS = "120"

# Maps: set to "google" to use Places API; default is "mock"
$env:MAPS_MODE = "mock"
# If MAPS_MODE=google, provide your API key
# $env:GOOGLE_MAPS_API_KEY = "<your_key>"

uvicorn backend.app.main:app --reload --port 8000

Frontend (PowerShell):

cd frontend
npm install
npm run dev

Key Endpoints

• POST /api/triage: accept a tenant message; returns a work_order_id
• GET /api/work_orders/{id}: fetch triage results and vendor suggestions
• GET /health/ollama: check Ollama connectivity and resolved model
• GET /debug/ollama_raw: inspect raw LLM output (for debugging)

Notes

• Ensure Ollama is running locally on port 11434 and the llama3 model is installed (e.g., ollama pull llama3)
• To use Google Places, set MAPS_MODE=google and define GOOGLE_MAPS_API_KEY