feat: display model name in Feishu card footer

[qianhaizhang]

Summary

Add support for displaying the AI model name in the Feishu card footer. This feature allows users to see which model (including provider and think mode) was used to generate responses directly in the Feishu message card.

Changes

• Added footer.model config option to enable/disable model display
• Display format: provider/model (e.g., "anthropic/claude-sonnet-4-6")
• Filter out models ending with "off" (think mode indicator)
• Show thinkLevel as suffix when not "off" (e.g., "model (extended)")

Files Changed

• src/card/builder.ts - Main footer rendering logic
• src/card/reply-dispatcher.ts - Pass model info to controller
• src/card/streaming-card-controller.ts - Handle model selection
• src/core/config-schema.ts - Config schema update
• src/core/footer-config.ts - Footer config type
• src/card/__tests__/footer-model.test.ts - 30 unit tests

Test plan

• Verify model name displays correctly in Feishu card footer
• Test with different provider/model combinations
• Test think mode filtering (models ending with "off")
• Test thinkLevel suffix display when not "off"
• Run existing unit tests to ensure no regression
• Test with footer.model disabled (should not show model)

🤖 Generated with Claude Code

[HanShaoshuai-k]

Nice feature! The test coverage is solid and the default-off behavior is a good design choice. A few things to address:

Must Fix:

reply-dispatcher.ts L305-L320: The onModelSelectedWrapper breaks the original callback semantics — when ctx is undefined, prefixContext.onModelSelected is never called, which may break prefix context state reset. The forward should be unconditional:

const onModelSelectedWrapper = (
  ctx: { provider: string; model: string; thinkLevel: string | undefined } | undefined,
) => {
  if (ctx && controller) {
    controller.setModel(ctx);
  }
  prefixContext.onModelSelected(ctx);
};

Should Fix:

1. The model type { provider: string; model: string; thinkLevel?: string } is duplicated inline in 4+ places. Consider extracting a shared ModelInfo interface in src/core/types.ts.
2. Missing vitest.config.ts — the PR adds vitest as a dependency but no config file. vitest run may not behave as expected without it.
3. provider is typed as required string but tests pass objects without it (e.g. { model: 'claude-haiku-4-5' }). Should be provider?: string.

Minor:

• Missing newline at end of footer-model.test.ts
• setModel tests only assert expect(controller).toBeDefined() — not very meaningful

Question: The "filter models ending with off" logic relies on SDK internal naming conventions. Is this a stable contract, or would it be safer to rely solely on the thinkLevel field?