orenh57/test2

Purpose and runtime entry points

CereBI is a tenant-scoped operational digital-twin and intelligence platform. Operational source data is ingested, mapped into a MongoDB-backed twin, then queried by console APIs and bounded calculation workflows. Production runtime starts through the FastAPI application factory at src/cerebi/app/backend/server.py, with composition/lifecycle wiring in src/cerebi/app/backend/composition/application_factories.py and src/cerebi/app/backend/composition/lifespan.py. The browser UI is a React/Vite SPA rooted at src/cerebi/app/frontend/src/main.jsx.

Application boundary and UI

The backend is the sole browser-to-data HTTP boundary; browsers do not access MongoDB. Route modules expose domains including analytics, twin/object browsing, tenant administration, configuration, and task execution: src/cerebi/app/backend/routes/analytics.py, src/cerebi/app/backend/routes/object_browser.py, and src/cerebi/app/backend/routes/task_execution.py. The frontend consumes these APIs through src/cerebi/app/frontend/src/api/client.js; major operational views include src/cerebi/app/frontend/src/pages/DigitalTwinExplorerPage.jsx and src/cerebi/app/frontend/src/pages/WorkbenchPage.jsx. Built frontend assets are served by the backend, while Vite is a separate development runtime.

Composition, tenancy, and authorization

app owns dependency composition and tenant scoping, keeping data and intelligence implementation-independent. Tenant access and configuration are composed in src/cerebi/app/backend/composition/tenant_access.py and src/cerebi/app/backend/configuration.py. Backend workflows translate HTTP operations into domain reads or commands, notably src/cerebi/app/backend/workflows/tenant_catalog.py, src/cerebi/app/backend/workflows/data_operations.py, and src/cerebi/app/backend/workflows/task_execution.py. Analytics-specific and loopback security controls sit at src/cerebi/app/backend/security/analytics.py and src/cerebi/app/backend/security/loopback.py.

Digital twin ingestion and persistence

The data subsystem transforms external sources into normalized twin objects and relationships. Ingestion supports spreadsheet/Drive CSV synchronization and snapshot loading through src/cerebi/data/ingestion/sync_worker.py, src/cerebi/data/ingestion/spreadsheet_ingestion.py, and src/cerebi/data/ingestion/twin_snapshot_loader.py. Object-model mapping and validation form the source-to-twin boundary in src/cerebi/data/object_model/loader.py, src/cerebi/data/object_model/config_mapper.py, and src/cerebi/data/object_model/validation.py. Twin documents and relationships are stored and queried via src/cerebi/data/storage/twin_storage.py, src/cerebi/data/storage/twin_relationships.py, and src/cerebi/data/relationships/twin_map_mongo.py.

External inputs and infrastructure

MongoDB is the operational system of record for tenant configuration, twins, registry state, and persisted intelligence results. The only production Mongo driver-construction boundary is src/cerebi/foundation/mongo/provider.py, configured through src/cerebi/foundation/mongo/configuration.py. Aurora/RDS support is an explicit read-only source adapter in src/cerebi/data/storage/aurora.py, not a normal request-path dependency. Weather enrichment integrates NWS and Open-Meteo through src/cerebi/data/inputs/weather/nws.py and src/cerebi/data/inputs/weather/openmeteo.py.

Intelligence execution pipeline

Intelligence is configured by a model/task registry, then executes capabilities against twin-derived inputs. Bootstrap assembles tenant/client frameworks and validates registry configuration in src/cerebi/intelligence/bootstrap/container.py, src/cerebi/intelligence/bootstrap/registry_config.py, and src/cerebi/intelligence/registry/service.py. The calculation runner resolves inputs, guards execution, invokes adapters, finalizes results, and supports caching/locking: src/cerebi/intelligence/calculation/runner.py, src/cerebi/intelligence/calculation/resolution.py, and src/cerebi/intelligence/calculation/result_finalization.py. Capability adapters include route-delay scoring at src/cerebi/intelligence/capabilities/route_delay/adapter.py and relationship ranking at src/cerebi/intelligence/capabilities/relationship_ranking/adapter.py.

Results, impact, and recommendations

Completed calculations can persist results, write twin-facing outputs, evaluate impact configuration, and generate recommendations. These boundaries are represented by src/cerebi/intelligence/persistence/mongo/store.py, src/cerebi/intelligence/twin/result_sink.py, src/cerebi/intelligence/impact/engine.py, and src/cerebi/intelligence/recommendation/service.py. Memory and no-op implementations across these subsystems provide non-production or isolated execution modes; Mongo implementations are the durable runtime path.