How pricingfrontier/haute works
Haute is an open-source insurance pricing engine: teams compose a visual or code-backed pipeline, prepare data, build rating logic, train and score models, explore results, optimise prices, then trace and serve pricing results. The map includes the documented major capabilities, the browser-to-server pipeline path, execution, and project history/storage. Several capability modules are represented by their existing enclosing directory because the README documents them but their implementation was not sampled.
Main entry points: haute Rating pipeline · pipeline.py Pipeline API · App.tsx Visual editor · usePipelineAPI.ts Pipeline client
Components 21
Pipeline authoring
- Visual editor
frontend/src/App.tsx - Pipeline client
frontend/src/hooks/usePipelineAPI.ts - Pipeline document
frontend/src/types/pipelineDocument.ts - Node editors
frontend/src/panels/editors - Graph commits
frontend/src/hooks/useGraphCommitController.ts - Document recovery
src/haute/_pipeline_recovery.py
Pricing workflow
- Data inputs
src/haute/_input_providers.py - Rating pipeline
src/haute - Explore and profile
src/haute - Model training
src/haute - Model scoring
src/haute - Price optimisation
src/haute - Price trace
src/haute
Execution and serving
- Pipeline API
src/haute/routes/pipeline.py - Lazy execution
src/haute/_execute_lazy.py
Project state
- Git history
src/haute/_git.py - Project storage
src/haute/_project_storage.py
Other components
- Pricing team
- Polars
- MLflow
- Pricing API consumer
Connections 20
- Pricing team → Visual editor: authors
- Visual editor → Pipeline client: requests pipeline
- Pipeline client → Pipeline document: validates document
- Visual editor → Node editors: configures nodes
- Visual editor → Graph commits: commits edits
- Pipeline client → Pipeline API: calls endpoints
- Pipeline API → Document recovery: recovers documents
- Pipeline API → Lazy execution: runs pipeline
- Lazy execution → Data inputs: loads inputs
- Lazy execution → Polars: executes plans
- Rating pipeline → Explore and profile: profiles data
- Rating pipeline → Model training: trains models
- Model training → MLflow: logs runs
- Rating pipeline → Model scoring: scores models
- Model scoring → MLflow: loads models
- Rating pipeline → Price optimisation: optimises prices
- Rating pipeline → Price trace: explains calculations
- Pipeline API → Git history: saves history
- Git history → Project storage: publishes project
- Pricing API consumer → Pipeline API: requests pricing