fix: Empty thinking text in thought chunks on current Claude models#766
Open
pavelpenev wants to merge 1 commit into
Open
fix: Empty thinking text in thought chunks on current Claude models#766pavelpenev wants to merge 1 commit into
pavelpenev wants to merge 1 commit into
Conversation
Recent Claude models default the API's `thinking.display` field to
"omitted". Thinking blocks still stream, but signature-only: the block
opens, a single signature_delta arrives, and no thinking_delta ever
carries text. The agent then emits agent_thought_chunk notifications
whose text is always empty, so ACP clients render blank thought
sections — regardless of effort level or MAX_THINKING_TOKENS.
Request `display: "summarized"` explicitly so thought chunks carry
visible summary text again:
- unset MAX_THINKING_TOKENS → {type: "adaptive", display: "summarized"}
(adaptive matches the SDK default on current models, but is now sent
explicitly where previously no thinking config was sent at all)
- positive budget → display added to the enabled config
- 0 (disabled) unchanged
A thinking config provided via _meta.claudeCode.options still overrides
this default wholesale, so embedders retain full control.
Verified over ACP traffic with an Emacs agent-shell client on
claude-fable-5, claude-opus-4-8, and claude-sonnet-4-6: before, every
agent_thought_chunk had empty text; after, summarized reasoning
streams and renders.
Related: agentclientprotocol#297
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Recent Claude models default the API's
thinking.displayfield to "omitted". Thinking blocks still stream, but signature-only: the block opens, a single signature_delta arrives, and no thinking_delta ever carries text. The agent then emits agent_thought_chunk notifications whose text is always empty, so ACP clients render blank thought sections — regardless of effort level or MAX_THINKING_TOKENS.Request
display: "summarized"explicitly so thought chunks carry visible summary text again:A thinking config provided via _meta.claudeCode.options still overrides this default wholesale, so embedders retain full control.
Verified over ACP traffic with an Emacs agent-shell client on claude-fable-5, claude-opus-4-8, and claude-sonnet-4-6: before, every agent_thought_chunk had empty text; after, summarized reasoning streams and renders.
Related: #297