Reasoning-only content leaking into assistant history/display
Cluster 111761 · apps/desktop. Data as of
1 open item. Fix merged in #133729 on 6 Oct 2026. Check it covers your case.
The short answer
- Known?
- Yes. 8 related items grouped here, 1 issue already closed.
- Fix event
- A fix was merged to main in PR #133729. Check it covers your case. No release data is tracked.
- Tracker
- The maintainers' tracker marks this
resolved, deciding PR #115596. That is separate from the fix event. - Still open
- 0 issues and 1 PR.
A bug where reasoning-only model outputs were incorrectly promoted into assistant message content, polluting conversation history and transcripts. Multiple fixes address gating promotion, preventing persistence, dropping such rows from transcripts, and preserving reasoning provenance.
Open issues0
Open PRs1
- PR#114463fix(desktop): drop reasoning-only assistant rows from the hydrated transcript (#111761)
Fixes on record
-
#113188 fix(agent): reasoning promoted on a reasoning-only stop is never persisted as an ordinary reply (#111761)
files:
agent/turn_final_response.py,gateway/run.py,tests/agent/test_empty_terminal_reasoning_surface.py,tests/agent/test_reasoning_only_stop_persistence.py,tests/gateway/test_replay_entry_fields.py,website/docs/developer-guide/session-storage.md -
#113796 fix(agent): reasoning-only stop keeps chain-of-thought out of rows and no longer fakes a completion on a planning-monologue stall (closes remaining atoms of #111761)
files:
agent/agent_runtime_helpers.py,agent/turn_final_response.py,tests/agent/test_reasoning_only_stop_persistence.py,tests/agent/test_stall_guards.py,tests/agent/test_thinking_prefill_trailing_turn.py -
#115596 fix(desktop): /reasoning hide|show hides Thinking blocks immediately (closes remaining atom of #111761)
files:
apps/desktop/src/app/session/hooks/use-prompt-actions/slash.ts,apps/desktop/src/lib/desktop-slash-commands.ts,apps/desktop/src/lib/desktop-slash-registry.json,apps/desktop/src/lib/reasoning-slash.test.ts,apps/desktop/src/lib/reasoning-slash.ts,hermes_cli/commands.py,website/docs/user-guide/desktop.md -
#133729 fix(agent): a reasoning model's private chain-of-thought is no longer shown as the reply when it stops with empty content (#114082, salvage #112833)
files:
agent/agent_init.py,agent/reasoning_promotion.py,agent/turn_empty_response.py,agent/turn_failure_copy.py,agent/turn_final_response.py,tests/agent/test_empty_terminal_reasoning_surface.py,tests/agent/test_reasoning_only_stop_persistence.py,tests/agent/test_run_agent.py,website/docs/user-guide/configuring-models.md
2 PRs closed without merging
From the close comments: done elsewhere 1 · superseded by another PR 1
Closed issues · 1
- Closed#111761[Bug]: reasoning is promoted into assistant content on reasoning-only clean stop — pollutes history, not just display
How this cluster was grouped
-
Link evidence (
structural links throughout): every member is connected by structural evidence: shared changed files, or a closing/superseded reference. -
Pairwise link density
0.25: the share of item pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36.1.00means 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