This lane complements the base geospatial building layer by adding non-geospatial enrichment signals and scoring outputs.

• Ingest external city-rate, policy, and corporate signals into data/raw/
• Clean and normalize each source into data/processed/
• Build final building-level enrichment features in src/features/
• Compute water-capture, ROI, and overall viability scores in src/scoring/
• Expose starter API route handlers in src/api/routes/

Run modules in order from repo root using package/module execution:

python -m src.ingest.pull_city_rates
python -m src.ingest.pull_policy_signals
python -m src.ingest.pull_corporate_signals
python -m src.clean.clean_city_rates
python -m src.clean.clean_policy_signals
python -m src.clean.clean_corporate_signals
python -m src.features.build_final_building_features

Scoring helpers are imported from src/scoring/ and route handler starters live in src/api/routes/.

• Prefer python -m ... from repo root.
• Running cleaner scripts by direct file path (for example python src/clean/clean_city_rates.py) can lead to import resolution issues in some environments.

NexDrop uses Ollama as an external service for generating concise investment_analysis text.
The FastAPI app does not embed Ollama; it calls Ollama by URL.

Set these in your environment or .env:

OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=qwen2.5:7b
OLLAMA_TIMEOUT_SECONDS=60

Optional batch cap:

INVESTMENT_ANALYSIS_BATCH_LIMIT=500

1. Install Ollama on the host that will serve model inference:
   • https://ollama.com/download
2. Start Ollama service:

   ollama serve

3. Pre-pull the configured model before API startup:

   ollama pull qwen2.5:7b

If Ollama runs in another container or internal host, set OLLAMA_BASE_URL accordingly, for example:

• http://ollama:11434 (same Docker network)
• http://10.0.1.25:11434 (internal host)

From repo root:

PYTHONPATH=backend python -m src.features.generate_investment_analyses

What it does:

• verifies Ollama connectivity and configured model availability (/api/tags)
• fetches rows missing investment_analysis
• generates concise analyses
• updates final_building_features.investment_analysis

Start API:

PYTHONPATH=backend .venv/bin/python -m uvicorn src.api.main:app --host 0.0.0.0 --port 8001

Read current analysis:

curl -sS "http://127.0.0.1:8001/buildings/123/investment-analysis"

Generate/regenerate and persist:

curl -sS -X POST "http://127.0.0.1:8001/buildings/123/investment-analysis/generate"

• Host Ollama separately from the FastAPI app when possible.
• Use OLLAMA_BASE_URL to point FastAPI to the Ollama host.
• Pre-pull the model (ollama pull <model>) before application startup.
• Keep Ollama and API on private/internal networking and expose only the API edge.