Skip to content

Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)

Python

fix(smoke): accept reasoning-only streams and add text placeholder

Alishahryar1f96f541c0a69
diff --git a/providers/openai_compat.py b/providers/openai_compat.py
index 88ff24be..b01f5dad 100644
--- a/providers/openai_compat.py
+++ b/providers/openai_compat.py
@@ -391,6 +391,17 @@ class OpenAIChatTransport(BaseProvider):
for event in sse.ensure_text_block():
yield event
yield sse.emit_text_delta(" ")
+ elif (
+ not has_started_tool
+ and not sse.accumulated_text.strip()
+ and sse.accumulated_reasoning.strip()
+ ):
+ # Some OpenAI-compatible models (e.g. NIM reasoning templates) stream only
+ # ``reasoning_content`` with no ``content``; emit a minimal text block so
+ # clients and smoke ``text_content()`` see a completed assistant message.
+ for event in sse.ensure_text_block():
+ yield event
+ yield sse.emit_text_delta(" ")
for event in self._flush_task_arg_buffers(sse):
yield event
diff --git a/smoke/prereq/test_provider_prereq_live.py b/smoke/prereq/test_provider_prereq_live.py
index ed5d5ffd..c5269f10 100644
--- a/smoke/prereq/test_provider_prereq_live.py
+++ b/smoke/prereq/test_provider_prereq_live.py
@@ -8,6 +8,7 @@ import pytest
from core.anthropic.stream_contracts import (
assert_anthropic_stream_contract,
text_content,
+ thinking_content,
)
from smoke.lib.config import SmokeConfig, auth_headers
from smoke.lib.e2e import ProviderMatrixDriver
@@ -66,7 +67,11 @@ def test_configured_provider_models_stream_successfully(
)
skip_if_upstream_unavailable_events(events)
assert_anthropic_stream_contract(events)
- assert text_content(events).strip(), "provider returned no text"
+ has_text = bool(text_content(events).strip())
+ has_thinking = bool(thinking_content(events).strip())
+ assert has_text or has_thinking, (
+ "provider returned no visible text or thinking content"
+ )
except Exception as exc:
skip_if_upstream_unavailable_exception(exc)
failures.append(
diff --git a/tests/providers/test_streaming_errors.py b/tests/providers/test_streaming_errors.py
index d272c8ed..4bc5ace7 100644
--- a/tests/providers/test_streaming_errors.py
+++ b/tests/providers/test_streaming_errors.py
@@ -302,6 +302,38 @@ class TestStreamingExceptionHandling:
assert '"text_delta"' in event_text
assert "message_stop" in event_text
+ @pytest.mark.asyncio
+ async def test_reasoning_only_stream_emits_placeholder_text(self):
+ """When the model streams only ``reasoning_content`` (no ``content``), add text block.
+
+ NIM / some templates may emit no main ``content``; a minimal text block matches
+ the empty-body placeholder and helps clients that expect a text segment.
+ """
+ provider = _make_provider_with_thinking_enabled(True)
+ request = _make_request()
+ chunk1 = _make_chunk(reasoning_content="reasoning only from provider")
+ chunk2 = _make_chunk(finish_reason="stop")
+ stream_mock = AsyncStreamMock([chunk1, chunk2])
+ with (
+ patch.object(
+ provider._client.chat.completions,
+ "create",
+ new_callable=AsyncMock,
+ return_value=stream_mock,
+ ),
+ patch.object(
+ provider._global_rate_limiter,
+ "wait_if_blocked",
+ new_callable=AsyncMock,
+ return_value=False,
+ ),
+ ):
+ events = await _collect_stream(provider, request)
+ event_text = "".join(events)
+ assert "thinking_delta" in event_text
+ assert '"text_delta"' in event_text
+ assert "message_stop" in event_text
+
@pytest.mark.asyncio
async def test_stream_with_thinking_content(self):
"""Thinking content via think tags is emitted as thinking blocks."""