test(copilot): add dedicated test suite for Glossary OS Copilot

[paollacq]

Summary

This PR adds a dedicated Vitest suite for Glossary Copilot inside apps/glossary-os, turning Copilot quality into something visible and verifiable instead of implicit. The suite covers the Copilot core from multiple angles: concept detection, domain classification, error pattern matching, context building, prompt construction, structured response parsing, API input validation, and end-to-end Copilot action flows with mocked Gemini responses. In total, the PR adds 67 passing tests across 13 test files, plus a few runtime fixes discovered while building the suite. The goal is straightforward: make the Copilot behavior defensible for reviewers and judges.

What This PR Adds

• a dedicated test suite under apps/glossary-os/src/__tests__
• fixtures for realistic Solana / Anchor code and common Solana errors
• unit coverage for Copilot helper modules
• integration coverage for generateCopilotAnswer(...) in all action modes:
  • explain
  • debug
  • generate
  • plan
• contract tests for /api/copilot
• runtime hardening based on failures surfaced by the tests

Test Coverage

Current suite size:

• 13 test files
• 67 passing tests

Covered modules:

• src/lib/copilot/concept-detector.ts
• src/lib/copilot/domain-classifier.ts
• src/lib/copilot/error-patterns.ts
• src/lib/copilot/prompt-builder.ts
• src/lib/copilot/response-parser.ts
• src/lib/copilot.ts
• src/lib/glossary.ts
• src/app/api/copilot/route.ts

Test layers:

• Unit
  • concept detection
  • domain classification
  • error pattern matching
  • response parsing
  • prompt building
  • context building
  • glossary shared utilities
• Integration
  • explain flow
  • debug flow
  • generate flow
  • plan flow
• Contract
  • API input validation
  • Copilot structured output shape

Why This Matters

This PR does not add another Copilot surface area. It makes the existing Copilot reliable.

That matters because Glossary Copilot is not a static page feature. It includes:

• dynamic glossary context assembly
• prompt construction
• structured JSON parsing
• mode-specific behavior
• API validation
• Gemini integration boundaries

Without a dedicated suite, regressions in any of those layers would be easy to miss. With this PR, the Copilot behavior becomes much easier to review, maintain, and trust.

Validation

Commands run:

npm run test --workspace @stbr/glossary-os
npm run typecheck:web

Result:

Test Files  13 passed (13)
Tests       67 passed (67)

Screenshots

Copilot Workspace

Local Run

From the repository root:

npm install
npm run test --workspace @stbr/glossary-os
npm run typecheck:web