Reasoning content field routing and display issues

NousResearch/hermes-agent · cluster 22949 · agent

Data as of

7 open items. Fix attempted, not merged.

The short answer

5open issues
2open PRs
11items grouped
Known?
Yes. 11 related items grouped here, 3 issues already closed.
Fix event
No fix has been merged to main. 1 PR closed without merging; see the attempts below.
Tracker
The maintainers' tracker marks this unresolved. That is separate from the fix event.
Still open
5 issues and 2 PRs.

Multiple bugs and fixes around how reasoning/thinking content is separated from final assistant replies, including models that place responses in the wrong field, empty replies when only reasoning tags are emitted, and ordering of reasoning versus final content in API responses.

Open issues5

  • Issue#28755[Bug] TUI: models emitting only reasoning tags produce silent/empty assistant replies
  • Issue#30436NVIDIA NIM: qwen3.5 model puts response in reasoning_content instead of content
  • Issue#56142[Feature]: Reasoning / thinking content appears after the final answer in API response
  • Issue#56213[Bug]: Reasoning / thinking content appears after the final answer in API response
  • Issue#67035[UX] Streaming reasoning models emit provisional content as final — response text changes after user has already read it

Open PRs2

  • PR#28756fix(tui): surface reasoning-only model output as assistant reply
  • PR#30442fix: fallback to reasoning_content for models that put response there (NVIDIA NIM Qwen 3.5)

1 PR closed without merging

From the close comments: closed without comment 1

  • #23039 fix(agent): fallback to reasoning_content when content is null · closed without comment
Closed issues · 3
  • Closed#22949Kimi K2-6 with chat_template_kwargs.thinking: true returns content: null, reasoning_content not displayed in Telegram
  • Closed#40535Desktop app: reasoning content rendered inline cuts through response text
  • Closed#64854[Bug]: GPT-5.6 puts the user-facing response body in the reasoning channel
How this cluster was grouped
  • Link evidence (partly structural links): some members are connected by structural evidence (shared changed files, or a closing/superseded reference); the rest are joined by an automated relevance judgement over their text.
  • Members outside that structural core may belong to a different underlying problem.
  • Pairwise link density 0.12: the share of item pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36. 1.00 means every pair is linked.

Pick up a Lab task

Check whether a fix covers the open issues here, with your own agent. Bring back evidence for maintainers.

Open the Lab

At a glance

Subsystem
agent
Closed issues
3
Attempts
1 PR closed without merging

For your agent

The same facts, typed.