vultr/openclaw
1
0
Fork 0
mirror of https://github.com/openclaw/openclaw.git synced 2026-05-06 15:00:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: centralize provider thinking profiles

Browse Source
This commit is contained in:
Peter Steinberger
2026-04-21 09:04:37 +01:00
parent 1cc2fc82ca
commit f1805ab54d
57 changed files with 718 additions and 572 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
docs/tools/thinking.md
View File

@@ -21,8 +21,9 @@ title: "Thinking Levels"
- `x-high`, `x_high`, `extra-high`, `extra high`, and `extra_high` map to `xhigh`.
- `highest` maps to `high`.
- Provider notes:
- `adaptive` is only advertised in native command menus and pickers for providers/models that declare adaptive thinking support. It remains accepted as a typed directive for compatibility with existing configs and aliases.
- `max` is only advertised in native command menus and pickers for providers/models that declare max thinking support. Existing stored `max` settings are remapped to the largest supported level for the selected model when the model does not support `max`.
- Thinking menus and pickers are provider-profile driven. Provider plugins declare the exact level set for the selected model, including labels such as binary `on`.
- `adaptive`, `xhigh`, and `max` are only advertised for provider/model profiles that support them. Typed directives for unsupported levels are rejected with that model's valid options.
- Existing stored unsupported levels, including old `max` values after switching models, are remapped to the largest supported level for the selected model.
- Anthropic Claude 4.6 models default to `adaptive` when no explicit thinking level is set.
- Anthropic Claude Opus 4.7 does not default to adaptive thinking. Its API effort default remains provider-owned unless you explicitly set a thinking level.
- Anthropic Claude Opus 4.7 maps `/think xhigh` to adaptive thinking plus `output_config.effort: "xhigh"`, because `/think` is a thinking directive and `xhigh` is the Opus 4.7 effort setting.
@@ -38,7 +39,7 @@ title: "Thinking Levels"
2. Session override (set by sending a directive-only message).
3. Per-agent default (`agents.list[].thinkingDefault` in config).
4. Global default (`agents.defaults.thinkingDefault` in config).
5. Fallback: `adaptive` for Anthropic Claude 4.6 models, `off` for Anthropic Claude Opus 4.7 unless explicitly configured, `low` for other reasoning-capable models, `off` otherwise.
5. Fallback: provider-declared default when available, `low` for other catalog models marked reasoning-capable, `off` otherwise.
## Setting a session default
@@ -111,10 +112,13 @@ title: "Thinking Levels"
- The web chat thinking selector mirrors the session's stored level from the inbound session store/config when the page loads.
- Picking another level writes the session override immediately via `sessions.patch`; it does not wait for the next send and it is not a one-shot `thinkingOnce` override.
- The first option is always `Default (<resolved level>)`, where the resolved default comes from the active session model: `adaptive` for Claude 4.6 on Anthropic, `off` for Anthropic Claude Opus 4.7 unless configured, `low` for other reasoning-capable models, `off` otherwise.
- The picker stays provider-aware:
- most providers show `off | minimal | low | medium | high`
- Anthropic/Bedrock Claude 4.6 shows `off | minimal | low | medium | high | adaptive`
- Anthropic Claude Opus 4.7 shows `off | minimal | low | medium | high | xhigh | adaptive | max`
- Z.AI shows binary `off | on`
- The first option is always `Default (<resolved level>)`, where the resolved default comes from the active session model's provider thinking profile.
- The picker uses `thinkingOptions` returned by the gateway session row. The browser UI does not keep its own provider regex list; plugins own model-specific level sets.
- `/think:<level>` still works and updates the same stored session level, so chat directives and the picker stay in sync.
## Provider profiles
- Provider plugins can expose `resolveThinkingProfile(ctx)` to define the model's supported levels and default.
- Each profile level has a stored canonical `id` (`off`, `minimal`, `low`, `medium`, `high`, `xhigh`, `adaptive`, or `max`) and may include a display `label`. Binary providers use `{ id: "low", label: "on" }`.
- Published legacy hooks (`supportsXHighThinking`, `isBinaryThinking`, and `resolveDefaultThinkingLevel`) remain as compatibility adapters, but new custom level sets should use `resolveThinkingProfile`.
- Gateway rows expose `thinkingOptions` and `thinkingDefault` so ACP/chat clients render the same profile that runtime validation uses.