Predict response
Core frontend sections
Use prediction for headline outcomes, factors for explanation cards, inputs.weather for context, and model_info plus simulation_metadata for technical disclosure.
Product docs
Frontend-facing API reference.
This page mirrors the backend contract already present in the FastAPI app and frames it for the Next.js client.
Endpoint
Method
Purpose
/predict
POST
Parcel-specific risk at a target location, including factor attribution and confidence interval.
/spread
POST
Map-ready GeoJSON contours for probability thresholds at a given horizon.
/batch
POST
JSON batch scoring for up to 500 parcels.
/batch/csv
POST
CSV upload endpoint with ignition coordinates provided as multipart form fields.
/health
GET
Basic health response for deployment and app readiness checks.
Developer tools
Interactive API console
Use the in-app console to inspect request and response payloads against the live frontend contract without leaving the product shell.
Model info
Auditability fields
Risk-scoring responses include compact model provenance so customers can see which model, schema, calibration layer, and validation context produced the score.
Field
Meaning
model_version
Frozen MVP model identifier for customer-visible outputs.
model_family
Serving model family, currently LightGBM for risk scoring.
feature_schema_version
Feature contract used by schema-driven inference.
calibration_version
Calibration or score-normalization layer applied to the model output.
validation_auc_roc
Honest validation AUC when available. Use as triage context, not a guarantee.
validation_brier
Calibration error metric when available.
intended_use
Approved product framing for the score, currently portfolio triage and review prioritization.
Spread response
Map strategy
Each contour arrives as GeoJSON geometry plus threshold metadata, which maps directly to a fill-layer strategy in MapLibre.