Skip to content

fix: suppress Codex commentary stream deltas - #51429

Closed
davidojedalopez wants to merge 1 commit into
NousResearch:mainfrom
davidojedalopez:fix/codex-suppress-commentary-stream
Closed

fix: suppress Codex commentary stream deltas#51429
davidojedalopez wants to merge 1 commit into
NousResearch:mainfrom
davidojedalopez:fix/codex-suppress-commentary-stream

Conversation

@davidojedalopez

@davidojedalopez davidojedalopez commented Jun 23, 2026

Copy link
Copy Markdown

Summary

  • suppress live streaming of Codex Responses commentary / analysis message-item deltas
  • keep those deltas in collected output_text / transcript state for internal handling
  • add regression coverage proving commentary before tool calls is not streamed while final_answer still streams

Why

Codex can emit a commentary-phase message immediately before a function call in the same Responses stream. Hermes was treating the response.output_text.delta as user-facing stream content, so gateway platforms like Telegram could receive model planning/todo text before final-send suppression kicked in.

@alt-glitch alt-glitch added type/bug Something isn't working comp/agent Core agent runtime: loop, agent_init, prompt builder, context-compression, responses endpoint provider/openai OpenAI / Codex Responses API P2 Medium — degraded but workaround exists duplicate This issue or pull request already exists labels Jun 23, 2026
@alt-glitch

Copy link
Copy Markdown
Collaborator

This was generated by AI during triage.

Duplicate of #25268 — same Codex Responses commentary stream-delta suppression mechanism, both targeting #24933. #25268 is the more comprehensive open fix (suppresses both the commentary message-item promotion and the matching response.output_text.delta stream events). Competing/complementary open cluster for #24933: #21568 (earliest, callback sanitize), #26898 (draft), #41343 (routes commentary to reasoning channel). Maintainer to pick the canonical fix.

@davidojedalopez

Copy link
Copy Markdown
Author

Closing per David's review; work will remain local.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

comp/agent Core agent runtime: loop, agent_init, prompt builder, context-compression, responses endpoint duplicate This issue or pull request already exists P2 Medium — degraded but workaround exists provider/openai OpenAI / Codex Responses API type/bug Something isn't working

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants