Skip to content

docs(dte): document thinking effort - #48

Open
easonLiangWorldedtech wants to merge 2 commits into
Zoo-Code-Org:mainfrom
easonLiangWorldedtech:docs/thinking-effort
Open

docs(dte): document thinking effort#48
easonLiangWorldedtech wants to merge 2 commits into
Zoo-Code-Org:mainfrom
easonLiangWorldedtech:docs/thinking-effort

Conversation

@easonLiangWorldedtech

Copy link
Copy Markdown

Documentation for the Dynamic Thinking Effort (DTE) series. Companion to the trial VSIX review build.

New pages

  • Thinking Effort (Features) — the four UI surfaces (composer toggle, task header chip, in-chat display, subtask start effort), effort levels and precedence, the Model Reasoning Effort profile setting, model-driven changes and guardrails, and declaring supported levels for self-hosted models. Persistence semantics are explicit: the task-local override is stored with the task (restored when the task is reopened from history) and is never written to the provider settings.
  • set_thinking_effort (Available Tools) — tool reference following the existing tool-doc skeleton: parameters, behavior, when it is used, guardrails (capability clamp, oscillation refusal, escalation cap of 3 upward changes per task), and usage examples.

Wiring

  • sidebars.ts — both pages added to their categories.
  • experimental-features.md — Dynamic Thinking Effort listed, with a note that the setting gates the model-driven tool only (the chat UI surfaces are capability-gated and work with the setting off).
  • the-chat-interface.md — composer thinking effort selector documented.
  • new-task.md — optional thinking_effort parameter documented.
  • features/index.md — Thinking Effort link added.

Provider pages

  • OpenAI Compatible — declaring supported reasoning effort levels: which levels unlock the controls, and the current scope (the selected level is not yet sent to the server with each request).
  • Ollama — the pre-existing reasoning effort wire through the think parameter, including the Extra High/Max to High mapping.
  • LM Studio — note that the LM Studio page does not expose the declaration, and how to use it via the OpenAI Compatible provider instead.

Images

  • Three gallery captures under static/img/thinking-effort/ (composer menu, header chip, F7 declaration), each with a one-line caption per repo conventions.

Verification

  • pnpm build passes with no warnings (Docusaurus; also re-run by the repo's pre-commit hook).
  • All internal links resolve to existing docs.
  • No version numbers in general docs; no banned words in new content.
  • Every behavioral claim was checked against the trial branch code (task-local state, guardrail constants, i18n strings, provider request builders).

Add the Thinking Effort feature page and the set_thinking_effort tool reference. Wire both into the sidebars, the experimental features list, the chat interface overview, the features index, and the new_task tool docs. Add reasoning effort sections to the OpenAI Compatible, Ollama, and LM Studio provider pages. Include three gallery images under static/img/thinking-effort.
Make the persistence contract explicit: the task-local thinking-effort
override is stored with the task (like the task mode and API config name),
so reopening the task from history restores it; it is never written to the
provider settings, and a new task starts from the profile default. This
aligns the docs with the implemented behavior (taskMetadata HistoryItem
persistence + Task reopen hydration) and removes the ambiguity that a
reviewer could read as a settings-persistence contract.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants