tkmind/tkmind_go
1
0
Fork 0
Code Issues Pull Requests Actions 4 Packages Projects Releases Wiki Activity

claude adaptive thinking (#7944)

Browse Source 
Signed-off-by: Angie Jones <angie@Angies-MBP.lan>
Co-authored-by: Angie Jones <angie@Angies-MacBook-Pro.local>
Co-authored-by: Angie Jones <angie@Angies-MBP.lan>
This commit is contained in:
Angie Jones
2026-03-16 23:00:53 -05:00
committed by GitHub
parent 4604354bc2
commit 28cab714b6
3 changed files with 108 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
documentation/docs/getting-started/providers.md
+1 -1
View File
@@ -1262,7 +1262,7 @@ Some models expose their internal reasoning or "chain of thought" as part of the
| **DeepSeek-R1** (via OpenAI, Ollama, OpenRouter, OVHcloud, etc.) | Reasoning captured from the `reasoning_content` field in the API response |
| **Kimi** (via Groq or other OpenAI-compatible endpoints) | Reasoning captured from the `reasoning_content` field in the API response |
| **Gemini CLI** (Google Gemini models with thinking enabled) | Thinking blocks captured from the streaming response |
| **Claude** (Anthropic, with [extended thinking](/docs/guides/environment-variables#claude-extended-thinking) enabled) | Thinking blocks captured from the API response |
| **Claude** (Anthropic, with [Claude thinking](/docs/guides/environment-variables#claude-thinking-configuration) enabled) | Thinking blocks captured from the API response |
<Tabs groupId="interface">
<TabItem value="ui" label="goose Desktop" default>
documentation/docs/guides/environment-variables.md
+15 -12
View File
@@ -164,28 +164,32 @@ export GOOSE_LEAD_FAILURE_THRESHOLD=3
export GOOSE_LEAD_FALLBACK_TURNS=2
```
### Claude Extended Thinking
### Claude Thinking Configuration
These variables control Claude's [extended thinking](https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking) feature, which allows the model to reason through complex problems before generating a response. Supported on Anthropic and Databricks providers.
These variables control Claude's reasoning behavior. Supported on Anthropic and Databricks providers.
| Variable | Purpose | Values | Default |
|----------|---------|---------|---------|
| `CLAUDE_THINKING_ENABLED` | Enables extended thinking for Claude models | Set to any value to enable | Disabled |
| `CLAUDE_THINKING_BUDGET` | Maximum tokens allocated for Claude's internal reasoning process | Positive integer (minimum 1024) | 16000 |
| `CLAUDE_THINKING_TYPE` | Controls Claude reasoning mode | `adaptive`, `enabled`, `disabled` | `adaptive` for Claude 4.6+ models, otherwise `disabled` |
| `CLAUDE_THINKING_BUDGET` | Maximum tokens allocated for Claude's internal reasoning process when `CLAUDE_THINKING_TYPE=enabled` | Positive integer (minimum 1024) | 16000 |
**Examples**
```bash
# Enable extended thinking with default budget (16000 tokens)
export CLAUDE_THINKING_ENABLED=1
# Claude 4.6 adaptive thinking
export GOOSE_PROVIDER=anthropic
export GOOSE_MODEL=claude-sonnet-4-6
export CLAUDE_THINKING_TYPE=adaptive
# Enable with a larger budget for complex tasks
export CLAUDE_THINKING_ENABLED=1
# Explicit extended thinking with the default budget
export CLAUDE_THINKING_TYPE=enabled
# Explicit extended thinking with a larger budget for complex tasks
export CLAUDE_THINKING_TYPE=enabled
export CLAUDE_THINKING_BUDGET=32000
# Enable with a smaller budget for faster responses
export CLAUDE_THINKING_ENABLED=1
export CLAUDE_THINKING_BUDGET=8000
# Disable Claude thinking entirely
export CLAUDE_THINKING_TYPE=disabled
```
:::tip Viewing Thinking Output
@@ -734,4 +738,3 @@ When deploying goose in enterprise environments, administrators might need to co
- For security-sensitive variables (like API keys), consider using the system keyring instead of environment variables.
- Some variables may require restarting goose to take effect.
- When using the planning mode, if planner-specific variables are not set, goose will fall back to the main model configuration.