Disable Claude Code thinking to unblock Opus 4.7 gateway calls#102
Open
jxk590 wants to merge 1 commit into
Open
Disable Claude Code thinking to unblock Opus 4.7 gateway calls#102jxk590 wants to merge 1 commit into
jxk590 wants to merge 1 commit into
Conversation
Claude Code sends `thinking.type=enabled` on outbound requests, which the Databricks AI Gateway rejects for adaptive-only models (Opus 4.7 family) with a 400. The gateway expects `thinking.type=adaptive` + `output_config.effort`. Claude Code does support that shape but only when its model-name pattern matches a known Anthropic id — the gateway exposes models as `databricks-claude-opus-4-7`, so the prefix throws off detection and Claude Code falls back to the legacy `enabled` shape. The result: `ucode claude` validation fails and the config is reverted. Set `CLAUDE_CODE_DISABLE_THINKING=1` in the Claude Code env block so the field is omitted entirely. The gateway still applies adaptive reasoning server-side on Opus 4.7 (where it's mandatory), so reasoning quality is preserved on the new family. Older models lose client-side extended thinking through the gateway, which is the correct tradeoff for a configuration that actually works end-to-end.
Collaborator
|
@jxk590 I'm not able to reproduce the original issue - can you please share a detailed repro of how you came across this issue? |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Claude Code sends
thinking.type=enabledon outbound requests, which the Databricks AI Gateway rejects for adaptive-only models (Opus 4.7 family) with a 400. The gateway expectsthinking.type=adaptive+output_config.effort. Claude Code does support that shape but only when its model-name pattern matches a known Anthropic id — the gateway exposes models asdatabricks-claude-opus-4-7, so the prefix throws off detection and Claude Code falls back to the legacyenabledshape. The result:ucode claudevalidation fails and the config is reverted.Set
CLAUDE_CODE_DISABLE_THINKING=1in the Claude Code env block so the field is omitted entirely. The gateway still applies adaptive reasoning server-side on Opus 4.7 (where it's mandatory), so reasoning quality is preserved on the new family. Older models lose client-side extended thinking through the gateway, which is the correct tradeoff for a configuration that actually works end-to-end.