Skip to content

docs: clarify think tool is for models without native reasoning#2191

Merged
dgageot merged 1 commit intodocker:mainfrom
dgageot:board/fix-docker-agent-issue-2190-bfa77c62
Mar 20, 2026
Merged

docs: clarify think tool is for models without native reasoning#2191
dgageot merged 1 commit intodocker:mainfrom
dgageot:board/fix-docker-agent-issue-2190-bfa77c62

Conversation

@dgageot
Copy link
Member

@dgageot dgageot commented Mar 20, 2026

The documentation previously recommended including the think tool in every agent. In reality, it is only useful for models that lack built-in reasoning capabilities (e.g., smaller or older models). Models with native thinking — such as Claude with extended thinking, OpenAI o-series, or Gemini with a thinking budget — do not need it.

Changes

  • docs/tools/think: reword overview and tip to explain when to use it
  • docs/guides/tips: rename section and clarify it targets non-reasoning models
  • docs/concepts/multi-agent: qualify the best-practice bullet as conditional

Fixes #2190

The documentation previously recommended including the think tool in every
agent. In reality, it is only useful for models that lack built-in
reasoning capabilities (e.g., smaller or older models). Models with
native thinking — such as Claude with extended thinking, OpenAI o-series,
or Gemini with a thinking budget — do not need it.

Update docs across three files to reflect this:
- docs/tools/think: reword overview and tip to explain when to use it
- docs/guides/tips: rename section and clarify it targets non-reasoning models
- docs/concepts/multi-agent: qualify the best-practice bullet as conditional

Fixes docker#2190

Assisted-By: docker-agent
@dgageot dgageot requested a review from a team as a code owner March 20, 2026 08:24
Copy link

@docker-agent docker-agent bot left a comment

Choose a reason for hiding this comment

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

Assessment: 🟡 NEEDS ATTENTION

This documentation PR clarifies when to use the think tool, which is a valuable improvement. However, there are terminology inconsistencies between this PR and the existing provider documentation that should be addressed.

Summary:

  • 3 confirmed issues related to model terminology inconsistencies
  • All issues are in documentation (no code bugs)
  • The core message of the PR is correct, but the terminology needs alignment with provider docs

The think tool is a reasoning scratchpad that lets agents think step-by-step before acting. The agent can write its thoughts without producing visible output to the user — ideal for planning complex tasks, breaking down problems, and reasoning through multi-step solutions.

This is a lightweight tool with no side effects. It's recommended for all agents — it improves the quality of reasoning on complex tasks at minimal cost.
This is a lightweight tool with no side effects. It is most useful for models that lack built-in reasoning or thinking capabilities (e.g., smaller or older models). For models that already support native thinking — such as Claude with extended thinking, OpenAI o-series, or Gemini with a thinking budget — this tool is unnecessary since the model can reason internally.
Copy link

Choose a reason for hiding this comment

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

Terminology inconsistency with provider documentation

The term "extended thinking" is not used in the Anthropic provider documentation. The docs at docs/providers/anthropic/index.md refer to this feature as "Thinking Budget" and "Interleaved Thinking".

Similarly, "OpenAI o-series" is not documented anywhere in docs/providers/openai/index.md. The OpenAI provider docs list models like gpt-5, gpt-5-mini, gpt-4o, and gpt-4o-mini with thinking budget support, but no "o-series".

Suggestion: Use terminology consistent with the provider documentation:

  • For Claude: "Claude with Thinking Budget" or "Claude Sonnet 4.5"
  • For OpenAI: Refer to specific models like "gpt-5" or "gpt-5-mini" instead of "o-series"
  • Consider linking to the provider docs for clarity

```

The agent uses it as a scratchpad for planning and decision-making.
The agent uses it as a scratchpad for planning and decision-making. If your model already supports a [thinking budget]({{ '/configuration/models/#thinking-budget' | relative_url }}) (e.g., Claude with extended thinking, OpenAI o-series, Gemini with thinking enabled), you don't need this tool — the model can reason internally.
Copy link

Choose a reason for hiding this comment

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

Same terminology issue in tips documentation

This line also uses "extended thinking" and "o-series" which are not found in the provider documentation. Please use the same terminology as documented in docs/providers/anthropic/index.md and docs/providers/openai/index.md.

@dgageot dgageot merged commit b729670 into docker:main Mar 20, 2026
8 checks passed
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.

Misleading doc about the think toolset

3 participants