[avoidwork]

Description

Resolves the Copilot model context window from the /models API. Two gaps are closed:

• resolveCopilotModel (src/provider/openai.js) now decorates the /models discovery request with the Copilot bearer token via createCopilotFetch, so the token is read fresh from memory/auth.json and 401 re-auth is handled consistently.
• getModelContextLength (src/provider/modelInfo.js) now reads Copilot's usable prompt budget (limits.max_prompt / max_prompt_tokens, falling back to limits.context_window / max_context_window_tokens), preferring max_prompt over context_window.

The OpenSpec change copilot-model-context-window was proposed, implemented, and archived, syncing the model-context-length and github-copilot-provider specs.

Type of Change

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactor (no functional changes)
• Performance improvement
• CI / build / tooling

Testing

• Added 5 Copilot-shaped resolver cases to tests/unit/provider/modelInfo.test.js (limits.max_prompt, limits.context_window, max_prompt_tokens, max_context_window_tokens, no-usable-fields).
• Added 1 case to tests/unit/provider/openai.test.js asserting the Authorization: Bearer header on the /models request.
• Full suite: 4355 tests pass. npm run lint clean. npm run coverage maintained (modelInfo.js 100% lines).

Coverage

• Line coverage maintained

Checklist

• npm run lint passes
• Tests pass with maintained line coverage
• No forbidden patterns used
• Conventional Commit style applied

Closes #1539

[avoidwork]

Audit Results: copilot-model-context-window

Goal Fulfillment

• Authenticate model discovery — resolveCopilotModel now decorates the /models request with createCopilotFetch(), injecting the Copilot bearer token and handling 401 re-auth consistently. ✅
• Copilot context window — getModelContextLength reads limits.max_prompt / max_prompt_tokens, falling back to limits.context_window / max_context_window_tokens. Prefers max_prompt (usable budget) over context_window. ✅
• Guard new fields — Only returns a number when present and numeric; falls through to undefined otherwise. Never throws. ✅
• Unit tests — 5 Copilot-shaped cases in modelInfo.test.js; 1 Authorization header case in openai.test.js. ✅
• Verify — 4355 tests pass, lint clean, coverage maintained. ✅

Spec Compliance

• model-context-length delta: all Copilot shape scenarios covered (limits.max_prompt, limits.context_window, max_prompt_tokens, max_context_window_tokens, no-usable-fields). ✅
• github-copilot-provider delta: discovery request authenticated scenario covered. ✅

Task Completion

All 6 tasks complete (1.1, 2.1, 2.2, 3.1, 3.2, 4.1). ✅

Quality Check

• modelInfo.js at 100% line coverage (all new branches covered).
• openai.js at 96.35% lines / 80% functions — uncovered lines (62-64, 171, 199-200, 212-215) are pre-existing retry-delay and prototype-patch paths, not from this change.
• No forbidden patterns; conventional commit style applied.