timber-common · import commonRuntime-built models, persistence, shared identity records, and domain service discovery.
Shared platform component. Domain applications are documented separately.
Timber supplies the shared model and service infrastructure. It builds models from YAML, provides persistence interfaces and shared identity records, and discovers domain models and services through plugin entry points. Financial providers, caches and forecasters are registered by the Investments plugin, rather than imposed on every application.
Loads table, column and relationship definitions from YAML and registers domain-owned models.
Topic-scoped confirmation, consent-checked delivery, unsubscribe, and verified newsletter privacy operations. Each domain supplies its edition content.
Applications store and query records through shared database and session mechanisms; domain adapters own their schemas.
Stores users and organization membership so identity and resource permissions can be enforced consistently by serving services.
Exposes typed decisions and probability-bearing answers through a replaceable provider interface.
An installed domain contributes models and services without copying the shared substrate.
Timber supplies runtime-built models, persistence interfaces, shared identity records and service discovery. Model definitions can be loaded from YAML; domain packages register additional models and services through explicit plugin entry points. A new domain extends the substrate without copying its core.
Cambium owns generic prediction, scoring and blend contracts. Timber provides the database and registry mechanisms through which applications store evidence and resolve the services they need. The application adapter owns the mapping between those contracts and its records. Neither a generic model declaration nor a service registration proves that a particular learning loop has been deployed.
The release candidate includes a bounded NumPy gradient-boosted stump learner with positive recency weights for batch refits. Domains choose their own targets, features and retraining windows. This is not an online adaptive-forest algorithm.
Immutable execution snapshots require user and organization ownership. Public results and scientist diagnostics are stored separately; domain workers append observed outcomes with their availability times. Reading a past execution does not rerun today's model over yesterday's forecast.
Identity and enterprise membership are shared models. Authentication policy is enforced by Canopy; resource and tenant permissions are also checked where an operation is served. Storage does not grant permission to read a record.
Stock-provider adapters, financial caches, return forecasters and financial analytics are supplied by the Investments plugin. They are examples of services registered into Timber, not capabilities every domain must inherit or install.
A domain registers its models and services with Timber, its HTTP routes and configuration with Grove, its permitted agent tools with Acorn, and its widgets with Sky. Dependencies and schemas remain explicit. Workflow and page behavior belongs in Grove configuration; reusable intelligence belongs in Cambium.
The scientist workbench uses this same structure. Its investment prediction and outcome records are owned by the Investments plugin and read through Timber's persistence interface. A separate tenant's private evidence is not made visible by assigning a platform scientist role.
Cambium documents the shared intelligence contracts and their implementation status. Investments documents the financial models and their application-specific evaluation.