# Changelog

## Score

- CAI 55 → 63 (+8.4)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 100 → 100 (+0.0)
- Architecture 92 → 90 (-1.8)
- Maturity 34 → 50 (+16.5)
- Readiness 51 → 55 (+3.3)
- Security 98 → 98 (+0.0)

## Resolved (6)

- Documentation: no project overview (docs/tutorial/database.md)
- Duplicate intent with inconsistent naming and parameter types. Both functions retrieve a related model type from a model instance/type. One uses 'relationship_name' and the other 'field_name'. One takes an instance `model: T`, the other a type `model: Type[BaseModel]`.
- Inconsistent filter class hierarchy and naming. `AdaptiveFilter` and `FilterSpecification` are in the core module, while `AlchemyFilter`, `InternalFilter`, and `MongoFilter` are provider-specific. However, `AdaptiveFilter` seems to be a core abstraction that isn't clearly aligned with `FilterSpecification` (which has `filtering_options_cls`). The naming `AdaptiveFilter` vs `FilterSpecification` is confusing as they likely serve similar composition purposes.
- Inconsistent function signatures and return types for specification application functions. `alchemy_*` functions return `Select` (mostly), `internal_*` functions return `Iterable` or `QueryT` inconsistently (e.g., `internal_order` returns `Iterable`, `internal_join` returns `QueryT`), and `mongo_*` functions return `dict`. The `adaptive` functions take an extra `repository` argument which is not present in the provider-specific functions, breaking the pattern.
- Inconsistent naming and return types for dictionary-to-model conversion. The standalone utils use `dict_to_*_models` (plural models in name, but singular return in Alchemy, dict in Internal). The Repositories use `dict_to_models` (singular return). This creates confusion about whether the function returns a single object, a dict, or a collection.
- Inconsistent serialization/deserialization naming. `loads` is used for deserialization from string, but `json` and `dict` are used for serialization. This is somewhat standard, but `loads` implies a specific format (JSON) while `json` is a method name. It's better to have `from_json`/`to_json` or `serialize`/`deserialize` for clarity.

## Architecture

- Unchanged — 0 containers · 1 contexts · 0 edges
