Skip to content

[Feature] unify reasoning effort on one shared scale - #5

Merged
jason-fxz merged 2 commits into
mainfrom
effort-dialect
Aug 16, 2026
Merged

[Feature] unify reasoning effort on one shared scale#5
jason-fxz merged 2 commits into
mainfrom
effort-dialect

Conversation

@jason-fxz

Copy link
Copy Markdown
Collaborator

Clients/Agent speak different reasoning-effort dialects; each checkpoint's template only accepts its own. To better unify the experience, every named effort projects onto one shared scale

name none minimal low medium high xhigh max
scale 0.0 0.1 0.2 0.7 0.9 0.99 0.99

Selection: the checkpoint's vocabulary is probe-rendered from its own template at load, and a foreign value lands on the nearest supported gear on this scale.

Effort/thinking behavior is no longer hardcoded per model family: the think_spec registry is deleted, protocol toggles broadcast every spelling templates read, and the TUI's gears on /v1/cache/status are derived from the same probe.

Copilot AI lite review requested due to automatic review settings August 16, 2026 06:40

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

@jason-fxz
jason-fxz merged commit d7cd6ed into main Aug 16, 2026
@jason-fxz
jason-fxz deleted the effort-dialect branch August 16, 2026 06:41
andy-yang-1 added a commit that referenced this pull request Aug 16, 2026
Main's #5 replaced the per-family thinking-gear registry with checkpoint
probing (tokenizer/effort.py), which supersedes the muse entries this branch
had added to model_meta -- resolved by taking main's side and mapping muse
onto the new machinery where it belongs: the render layer now broadcasts
reasoning_effort as reasoning_strength too (the same every-spelling rule the
thinking toggles use; Jinja ignores undeclared variables and an explicit
caller spelling wins). Through the broadcast the probe sees muse's template
grade effort (validating nothing, defaulting high), so /v1/cache/status
derives the OpenAI gear triple with default high, and muse's native xhigh
still passes quantization untouched. The double-bos fix is re-applied to the
reworked tokenize path: template-rendered prompts encode with
add_special_tokens=False, raw strings and the dsv4 encoder keep the default.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants