From b3b1e58ad60c4874fb491c181b5ae10d0f6d15cc Mon Sep 17 00:00:00 2001 From: teknium1 <127238744+teknium1@users.noreply.github.com> Date: Sun, 5 Jul 2026 06:05:07 -0700 Subject: [PATCH] fix(codex): stream commentary deltas through the reasoning channel Follow-up to the salvaged #58696 (devatnull) + #41343 (annguyenNous) commits: instead of fully suppressing commentary/analysis-phase stream deltas, fire on_reasoning_delta so the CLI/gateway display them like thinking text. Matches Codex CLI semantics where commentary is never the turn's final answer, while keeping the narration visible in the reasoning display. Adds devatnull to AUTHOR_MAP. --- agent/codex_runtime.py | 17 +++++++++++++---- scripts/release.py | 1 + .../run_agent/test_run_agent_codex_responses.py | 7 ++++++- 3 files changed, 20 insertions(+), 5 deletions(-) diff --git a/agent/codex_runtime.py b/agent/codex_runtime.py index 38ba98f5835..1cf48ec1705 100644 --- a/agent/codex_runtime.py +++ b/agent/codex_runtime.py @@ -605,9 +605,10 @@ def _consume_codex_event_stream( _raise_stream_error(event) # Track the phase of the active streamed message item. Codex/Harmony - # ``commentary``/``analysis`` text is protocol/tool preamble, not the - # final answer. We still collect completed output items for replay, but - # suppress those deltas from live user-visible callbacks. + # ``commentary``/``analysis`` text is mid-turn preamble/progress + # narration, never the final answer. We still collect completed output + # items for replay, but route those deltas to the reasoning callback so + # they display like thinking text instead of assistant content. if event_type == "response.output_item.added": item = _event_field(event, "item") item_type = _item_field(item, "type", "") @@ -623,7 +624,15 @@ def _consume_codex_event_stream( if "output_text.delta" in event_type or event_type == "response.output_text.delta": delta_text = _event_field(event, "delta", "") is_commentary_delta = active_message_phase in {"commentary", "analysis"} - if delta_text and not is_commentary_delta: + if delta_text and is_commentary_delta: + # Commentary streams through the reasoning channel, not the + # visible answer stream (and stays out of output_text). + if on_reasoning_delta is not None: + try: + on_reasoning_delta(delta_text) + except Exception: + logger.debug("Codex stream on_reasoning_delta raised", exc_info=True) + elif delta_text: collected_text_deltas.append(delta_text) if not has_tool_calls: if not first_delta_fired: diff --git a/scripts/release.py b/scripts/release.py index d6f0d1b3085..a908eb92d82 100755 --- a/scripts/release.py +++ b/scripts/release.py @@ -489,6 +489,7 @@ AUTHOR_MAP = { "Rivuza@users.noreply.github.com": "Rivuza", "annguyenNous@users.noreply.github.com": "annguyenNous", "285874597+annguyenNous@users.noreply.github.com": "annguyenNous", + "perkintahmaz50@gmail.com": "devatnull", "kylekahraman@users.noreply.github.com": "kylekahraman", "130975919+kylekahraman@users.noreply.github.com": "kylekahraman", "seppe@fushia.be": "seppegadeyne", diff --git a/tests/run_agent/test_run_agent_codex_responses.py b/tests/run_agent/test_run_agent_codex_responses.py index 0b9f7daa06f..0c24adc4ed6 100644 --- a/tests/run_agent/test_run_agent_codex_responses.py +++ b/tests/run_agent/test_run_agent_codex_responses.py @@ -629,7 +629,7 @@ def test_run_codex_stream_returns_collected_items_when_stream_ends_without_termi assert response.output == [output_item] -def test_consume_codex_stream_suppresses_commentary_phase_deltas(monkeypatch): +def test_consume_codex_stream_routes_commentary_phase_deltas_to_reasoning(monkeypatch): from agent.codex_runtime import _consume_codex_event_stream commentary_item = SimpleNamespace( @@ -646,6 +646,7 @@ def test_consume_codex_stream_suppresses_commentary_phase_deltas(monkeypatch): arguments="{}", ) streamed = [] + reasoning_streamed = [] response = _consume_codex_event_stream( _FakeCreateStream([ @@ -665,9 +666,11 @@ def test_consume_codex_stream_suppresses_commentary_phase_deltas(monkeypatch): ]), model="gpt-5-codex", on_text_delta=streamed.append, + on_reasoning_delta=reasoning_streamed.append, ) assert streamed == [] + assert reasoning_streamed == ["I’ll call the tool now."] assert response.output == [commentary_item, function_item] assert response.output_text == "" @@ -1635,6 +1638,7 @@ def test_normalize_codex_response_marks_commentary_only_message_as_incomplete(mo assert finish_reason == "incomplete" assert (assistant_message.content or "") == "" + assert "inspect the repository" in (assistant_message.reasoning or "") assert assistant_message.codex_message_items assert assistant_message.codex_message_items[0]["phase"] == "commentary" assert "inspect the repository" in assistant_message.codex_message_items[0]["content"][0]["text"] @@ -1651,6 +1655,7 @@ def test_normalize_codex_response_does_not_fallback_to_output_text_for_commentar assert finish_reason == "incomplete" assert (assistant_message.content or "") == "" + assert "call the tool" in (assistant_message.reasoning or "") assert assistant_message.codex_message_items[0]["phase"] == "commentary" def test_normalize_codex_response_final_answer_overrides_top_level_incomplete(monkeypatch):