lsm / neokai / 25467051357 / 13
81%
dev: 81%

DEFAULT BRANCH: dev
Ran
Files 276
Run time 7s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 31.86% (+0.03%) from 31.826%
25467051357.13

push

github

web-flow 
Improve thinking level UI: rename Auto→Off, add think24k, provider-aware options (#1804)

* Rename Auto→Off in thinking level UI, add think24k, provider-aware options

- Replace 'auto' with 'off' across ThinkingLevel type and all UI labels
- Add think24k option (24000 tokens) to thinking level selector
- Introduce ProviderCapabilities.thinkingModes ('off'|'on'|'granular')
- Update all 9 providers with correct thinkingModes capability
- Filter thinking options in SessionStatusBar based on provider capability
- Daemon query-options-builder disables thinking when mode is 'off'
- For 'on' mode, any enabled level maps to enabled/31999 tokens
- Backward compatibility: normalize legacy 'auto' → 'off' at RPC boundaries

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Address review feedback: shared helpers, tests, dead code

- Replace inline thinking-options switch in SessionStatusBar with
  getThinkingOptionsForProvider from @neokai/shared
- Use normalizeThinkingLevel in query-options-builder and both
  session-handlers RPC endpoints instead of inline normalization
- Add missing think24k test in session-handlers.test.ts
- Replace stale ultrathink test with proper think24k assertion
- Add provider-aware thinking mode tests for 'off' (minimax) and
  'on' (kimi) in query-options-builder.test.ts

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Accept legacy 'auto' thinking level during space import

- Update thinkingLevelSchema to preprocess 'auto' → 'off' so older
  v1 space exports with thinkingLevel: 'auto' remain importable
- Add round-trip test verifying legacy 'auto' is normalized to 'off'

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Fix CI: avoid provider instantiation just to read thinkingModes capability

- `addSessionStateOptions` now reads thinkingModes from the static
  `PROVIDER_THINKING_MODES` map instead of calling `getProviderContextManager()`,
  which instantiates real providers and fails in CI when API keys ... (continued)

22482 of 70566 relevant lines covered (31.86%)

16.27 hits per line

Source Files on job daemon-online-space-1 - 25467051357.13