# ai.matey.react.hooks

## 0.2.2

### Patch Changes

- 73aa9f1: Fix broken CJS entry points across the whole package family. Every package declares
  `"type": "module"` for ESM subpath resolution, but shipped `dist/cjs/` builds with no nested
  override - Node walked up to the package root, saw `"type": "module"`, and misinterpreted the
  compiled CommonJS as ESM, so `require("ai.matey.x")` failed with `Cannot find module './y.js'`
  on every package in the family (ESM `import` was unaffected). Each package's build now emits a
  `dist/cjs/package.json` containing `{"type":"commonjs"}` (via a new
  `scripts/fix-cjs-package-json.js` post-build step) to correctly scope the CJS build's module
  type. No source or `exports` map changes - verified via `npm pack` + fresh install against the
  exact repro in #23, both direct `require()` and the `require` export condition on subpaths (e.g.
  `ai.matey.backend.browser/chrome-ai`).

  (#23)

- Updated dependencies [73aa9f1]
  - ai.matey.react.core@0.2.2
  - ai.matey.types@0.5.1
  - ai.matey.utils@0.4.2

## 0.2.1

### Patch Changes

- f227db2: Lint hardening: previously-unlinted packages (cli, react-\*) now pass the strict ESLint config;
  fixed floating/misused promises in React hooks and CLI, case-block declarations, and unused
  variables. require-await and no-redundant-type-constituents re-enabled repo-wide.
- 2912b7d: Introduce a shared, data-driven model registry in `ai.matey.utils` as the single source of truth
  for model metadata (pricing, context windows, capabilities, quality/latency). The registry ships
  with a mid-2026 seed (GPT-5.x/o-series, Claude 4.x, Gemini 2.5/3, Grok, current Mistral/DeepSeek,
  plus embedding models) and is runtime-extensible via `registerModels()` / `overrideModelPricing()`,
  with alias and longest-prefix fallback so new dated snapshots of known families still resolve.

  `ai.matey.core`'s model-pricing API is now a thin delegate over the registry (no API break; legacy
  models keep their prices, marked `deprecated`). Capability inference recognizes current families.
  Cost-tracking middleware consults the registry before provider-level defaults. `useTokenCount`
  consults the registry for context windows. Backend default models updated:
  `claude-3-haiku-20240307` → `claude-sonnet-4-5-20250929` (Anthropic), `gpt-3.5-turbo` →
  `gpt-5-mini` (OpenAI) — note this changes behavior for requests that omit a model. `estimateCost()`
  on both backends now prices the actual requested model. Refreshed `deepseek-chat` pricing.

- Updated dependencies [dae4d01]
- Updated dependencies [e7df1d0]
- Updated dependencies [f227db2]
- Updated dependencies [2912b7d]
- Updated dependencies [78731bb]
- Updated dependencies [b7e2312]
- Updated dependencies [58ebc03]
  - ai.matey.types@0.3.0
  - ai.matey.utils@0.3.0
  - ai.matey.react.core@0.2.1