feat(models): add StepFun reasoning model adapter (#3461)

Add PatchedChatStepFun adapter for StepFun reasoning models (step-3.7-flash,
step-3.5-flash). Captures reasoning from both streaming and non-streaming
responses and replays it on historical assistant messages for multi-turn
tool-call conversations.

- New: PatchedChatStepFun adapter with streaming/non-streaming reasoning capture
- Support both reasoning and reasoning_content field names
- 17 unit tests covering all response paths
- Updated: config.example.yaml with StepFun configuration example

.env.example

@@ -21,6 +21,7 @@ INFOQUEST_API_KEY=your-infoquest-api-key
 # DEEPSEEK_API_KEY=your-deepseek-api-key
 # NOVITA_API_KEY=your-novita-api-key  # OpenAI-compatible, see https://novita.ai
 # MINIMAX_API_KEY=your-minimax-api-key  # OpenAI-compatible, see https://platform.minimax.io
+# STEPFUN_API_KEY=your-stepfun-api-key  # OpenAI-compatible, see https://platform.stepfun.com
 # VLLM_API_KEY=your-vllm-api-key  # OpenAI-compatible
 # FEISHU_APP_ID=your-feishu-app-id
 # FEISHU_APP_SECRET=your-feishu-app-secret

backend/packages/harness/deerflow/models/patched_stepfun.py

@@ -0,0 +1,175 @@
+"""Patched ChatOpenAI adapter for StepFun reasoning models.
+
+StepFun returns ``reasoning`` (or ``reasoning_content`` with deepseek-style) in
+both streaming deltas and non-streaming responses. Standard ``ChatOpenAI``
+ignores these non-standard fields, so reasoning content is silently dropped.
+This adapter captures reasoning from all response paths and replays it on
+historical assistant messages for multi-turn tool-call conversations.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from typing import Any
+
+from langchain_core.language_models import LanguageModelInput
+from langchain_core.messages import AIMessage, AIMessageChunk
+from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
+from langchain_openai import ChatOpenAI
+
+from deerflow.models.assistant_payload_replay import (
+    restore_assistant_payloads,
+    restore_reasoning_content,
+)
+
+_MISSING = object()
+
+
+def _extract_reasoning(value: Any) -> str | object:
+    """Return reasoning content from a dict/Pydantic object.
+
+    StepFun may return reasoning via ``reasoning`` (default) or
+    ``reasoning_content`` (deepseek-style). Check both fields.
+    """
+    if isinstance(value, Mapping):
+        # Check reasoning_content first (deepseek-style), then reasoning (default)
+        for field in ("reasoning_content", "reasoning"):
+            if field in value and value[field] is not None:
+                return value[field]
+        return _MISSING
+
+    # Pydantic / SDK object attributes
+    for field in ("reasoning_content", "reasoning"):
+        attr = getattr(value, field, _MISSING)
+        if attr is not _MISSING and attr is not None:
+            return attr
+
+    # Some SDK versions store extra fields in model_extra
+    model_extra = getattr(value, "model_extra", None)
+    if isinstance(model_extra, Mapping):
+        for field in ("reasoning_content", "reasoning"):
+            if field in model_extra and model_extra[field] is not None:
+                return model_extra[field]
+
+    return _MISSING
+
+
+def _with_reasoning_content(message: AIMessage | AIMessageChunk, reasoning: str) -> AIMessage | AIMessageChunk:
+    """Return a copy of *message* with reasoning_content stored in additional_kwargs."""
+    additional_kwargs = dict(message.additional_kwargs)
+    if additional_kwargs.get("reasoning_content") != reasoning:
+        additional_kwargs["reasoning_content"] = reasoning
+    return message.model_copy(update={"additional_kwargs": additional_kwargs})
+
+
+def _get_typed_choice_message(response: Any, index: int) -> Any:
+    """Extract the SDK-typed choice message at *index*, if available."""
+    choices = getattr(response, "choices", None)
+    if choices is None:
+        return None
+    try:
+        return choices[index].message
+    except (AttributeError, IndexError, TypeError):
+        return None
+
+
+class PatchedChatStepFun(ChatOpenAI):
+    """ChatOpenAI with full reasoning support for StepFun models.
+
+    Captures ``reasoning`` / ``reasoning_content`` from both streaming and
+    non-streaming responses and replays it on historical assistant messages in
+    multi-turn tool-call conversations.
+    """
+
+    @classmethod
+    def is_lc_serializable(cls) -> bool:
+        return True
+
+    @property
+    def lc_secrets(self) -> dict[str, str]:
+        return {"api_key": "STEPFUN_API_KEY", "openai_api_key": "STEPFUN_API_KEY"}
+
+    # --- Request payload replay ---
+
+    def _get_request_payload(
+        self,
+        input_: LanguageModelInput,
+        *,
+        stop: list[str] | None = None,
+        **kwargs: Any,
+    ) -> dict:
+        """Restore ``reasoning_content`` on historical assistant messages."""
+        original_messages = self._convert_input(input_).to_messages()
+        payload = super()._get_request_payload(input_, stop=stop, **kwargs)
+
+        restore_assistant_payloads(
+            payload.get("messages", []),
+            original_messages,
+            restore_reasoning_content,
+        )
+
+        return payload
+
+    # --- Streaming reasoning capture ---
+
+    def _convert_chunk_to_generation_chunk(
+        self,
+        chunk: dict,
+        default_chunk_class: type,
+        base_generation_info: dict | None,
+    ) -> ChatGenerationChunk | None:
+        """Capture ``reasoning`` / ``reasoning_content`` from streaming deltas."""
+        generation_chunk = super()._convert_chunk_to_generation_chunk(
+            chunk,
+            default_chunk_class,
+            base_generation_info,
+        )
+        if generation_chunk is None:
+            return None
+
+        choices = chunk.get("choices", [])
+        if choices:
+            delta = choices[0].get("delta") or {}
+            reasoning = _extract_reasoning(delta)
+            if reasoning is not _MISSING and isinstance(generation_chunk.message, AIMessageChunk):
+                generation_chunk = ChatGenerationChunk(
+                    message=_with_reasoning_content(generation_chunk.message, reasoning),
+                    generation_info=generation_chunk.generation_info,
+                )
+
+        return generation_chunk
+
+    # --- Non-streaming reasoning capture ---
+
+    def _create_chat_result(
+        self,
+        response: dict | Any,
+        generation_info: dict | None = None,
+    ) -> ChatResult:
+        """Extract ``reasoning`` / ``reasoning_content`` from non-streaming responses."""
+        result = super()._create_chat_result(response, generation_info)
+        response_dict = response if isinstance(response, dict) else response.model_dump()
+        choices = response_dict.get("choices", [])
+
+        patched_generations: list[ChatGeneration] | None = None
+        for index, generation in enumerate(result.generations):
+            choice = choices[index] if index < len(choices) else {}
+            choice_message = choice.get("message", {}) if isinstance(choice, Mapping) else {}
+            reasoning = _extract_reasoning(choice_message)
+
+            if reasoning is _MISSING and not isinstance(response, dict):
+                reasoning = _extract_reasoning(_get_typed_choice_message(response, index))
+
+            message = generation.message
+            if reasoning is not _MISSING and isinstance(message, AIMessage):
+                if patched_generations is None:
+                    patched_generations = list(result.generations)
+                patched_generations[index] = ChatGeneration(
+                    message=_with_reasoning_content(message, reasoning),
+                    generation_info=generation.generation_info,
+                )
+
+        return ChatResult(
+            generations=patched_generations or result.generations,
+            llm_output=result.llm_output,
+        )

backend/tests/test_patched_stepfun.py

@@ -0,0 +1,305 @@
+"""Tests for deerflow.models.patched_stepfun.PatchedChatStepFun."""
+
+from __future__ import annotations
+
+from unittest.mock import MagicMock, patch
+
+from langchain_core.messages import AIMessage, AIMessageChunk, HumanMessage
+
+
+def _make_model(**kwargs):
+    from deerflow.models.patched_stepfun import PatchedChatStepFun
+
+    return PatchedChatStepFun(
+        model="step-3.7-flash",
+        api_key="test-key",
+        base_url="https://api.stepfun.com/v1",
+        **kwargs,
+    )
+
+
+# ---------------------------------------------------------------------------
+# Basic properties
+# ---------------------------------------------------------------------------
+
+
+def test_is_lc_serializable_returns_true():
+    from deerflow.models.patched_stepfun import PatchedChatStepFun
+
+    assert PatchedChatStepFun.is_lc_serializable() is True
+
+
+def test_lc_secrets_contains_stepfun_api_key_mapping():
+    model = _make_model()
+    assert model.lc_secrets["api_key"] == "STEPFUN_API_KEY"
+    assert model.lc_secrets["openai_api_key"] == "STEPFUN_API_KEY"
+
+
+# ---------------------------------------------------------------------------
+# _extract_reasoning helper
+# ---------------------------------------------------------------------------
+
+
+def test_extract_reasoning_from_dict_with_reasoning():
+    from deerflow.models.patched_stepfun import _extract_reasoning
+
+    assert _extract_reasoning({"reasoning": "thinking..."}) == "thinking..."
+
+
+def test_extract_reasoning_from_dict_with_reasoning_content():
+    from deerflow.models.patched_stepfun import _extract_reasoning
+
+    assert _extract_reasoning({"reasoning_content": "thinking..."}) == "thinking..."
+
+
+def test_extract_reasoning_prefers_reasoning_content_over_reasoning():
+    from deerflow.models.patched_stepfun import _extract_reasoning
+
+    result = _extract_reasoning({"reasoning_content": "deepseek", "reasoning": "native"})
+    assert result == "deepseek"
+
+
+def test_extract_reasoning_missing_returns_sentinel():
+    from deerflow.models.patched_stepfun import _MISSING, _extract_reasoning
+
+    assert _extract_reasoning({}) is _MISSING
+    assert _extract_reasoning({"reasoning": None}) is _MISSING
+
+
+# ---------------------------------------------------------------------------
+# Request payload replay (_get_request_payload)
+# ---------------------------------------------------------------------------
+
+
+def test_reasoning_content_injected_into_assistant_tool_call_message():
+    model = _make_model()
+
+    human = HumanMessage(content="Check Beijing weather.")
+    ai = AIMessage(
+        content="",
+        additional_kwargs={"reasoning_content": "I need to call the weather tool."},
+    )
+    payload_message = {
+        "role": "assistant",
+        "content": "",
+        "tool_calls": [
+            {
+                "id": "call_weather",
+                "type": "function",
+                "function": {"name": "get_weather", "arguments": '{"location":"Beijing"}'},
+            }
+        ],
+    }
+    base_payload = {
+        "messages": [
+            {"role": "user", "content": "Check Beijing weather."},
+            payload_message,
+        ]
+    }
+
+    with patch.object(type(model).__bases__[0], "_get_request_payload", return_value=base_payload):
+        with patch.object(model, "_convert_input") as mock_convert:
+            mock_convert.return_value = MagicMock(to_messages=lambda: [human, ai])
+            payload = model._get_request_payload([human, ai])
+
+    assert payload["messages"][1]["reasoning_content"] == "I need to call the weather tool."
+
+
+def test_reasoning_content_is_noop_when_missing():
+    model = _make_model()
+
+    human = HumanMessage(content="hello")
+    ai = AIMessage(content="hi", additional_kwargs={})
+    base_payload = {
+        "messages": [
+            {"role": "user", "content": "hello"},
+            {"role": "assistant", "content": "hi"},
+        ]
+    }
+
+    with patch.object(type(model).__bases__[0], "_get_request_payload", return_value=base_payload):
+        with patch.object(model, "_convert_input") as mock_convert:
+            mock_convert.return_value = MagicMock(to_messages=lambda: [human, ai])
+            payload = model._get_request_payload([human, ai])
+
+    assert "reasoning_content" not in payload["messages"][1]
+
+
+# ---------------------------------------------------------------------------
+# Streaming reasoning capture (_convert_chunk_to_generation_chunk)
+# ---------------------------------------------------------------------------
+
+
+def test_convert_chunk_captures_reasoning_field():
+    """StepFun default format: delta.reasoning."""
+    model = _make_model()
+
+    chunk = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"role": "assistant", "reasoning": "I need "}}]},
+        AIMessageChunk,
+        {},
+    )
+
+    assert chunk is not None
+    assert chunk.message.additional_kwargs["reasoning_content"] == "I need "
+
+
+def test_convert_chunk_captures_reasoning_content_field():
+    """StepFun deepseek-style format: delta.reasoning_content."""
+    model = _make_model()
+
+    chunk = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"role": "assistant", "reasoning_content": "I need "}}]},
+        AIMessageChunk,
+        {},
+    )
+
+    assert chunk is not None
+    assert chunk.message.additional_kwargs["reasoning_content"] == "I need "
+
+
+def test_convert_chunk_streams_reasoning_then_content():
+    """Full streaming flow: reasoning deltas followed by content."""
+    model = _make_model()
+
+    first = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"role": "assistant", "reasoning": "I need "}}]},
+        AIMessageChunk,
+        {},
+    )
+    second = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"reasoning": "a tool."}}]},
+        AIMessageChunk,
+        {},
+    )
+    answer = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"content": "Done."}, "finish_reason": "stop"}], "model": "step-3.7-flash"},
+        AIMessageChunk,
+        {},
+    )
+
+    assert first is not None
+    assert second is not None
+    assert answer is not None
+
+    combined = first.message + second.message + answer.message
+    assert combined.additional_kwargs["reasoning_content"] == "I need a tool."
+    assert combined.content == "Done."
+
+
+def test_convert_chunk_noop_when_no_reasoning():
+    model = _make_model()
+
+    chunk = model._convert_chunk_to_generation_chunk(
+        {"choices": [{"delta": {"content": "Hello."}, "finish_reason": "stop"}], "model": "step-3.7-flash"},
+        AIMessageChunk,
+        {},
+    )
+
+    assert chunk is not None
+    assert "reasoning_content" not in chunk.message.additional_kwargs
+
+
+# ---------------------------------------------------------------------------
+# Non-streaming reasoning capture (_create_chat_result)
+# ---------------------------------------------------------------------------
+
+
+def test_create_chat_result_extracts_reasoning_field():
+    """StepFun default format: message.reasoning."""
+    model = _make_model()
+    response = {
+        "choices": [
+            {
+                "message": {
+                    "role": "assistant",
+                    "content": "The weather is sunny.",
+                    "reasoning": "The tool returned sunny weather.",
+                },
+                "finish_reason": "stop",
+            }
+        ],
+        "model": "step-3.7-flash",
+    }
+
+    result = model._create_chat_result(response)
+    message = result.generations[0].message
+
+    assert message.content == "The weather is sunny."
+    assert message.additional_kwargs["reasoning_content"] == "The tool returned sunny weather."
+
+
+def test_create_chat_result_extracts_reasoning_content_field():
+    """StepFun deepseek-style format: message.reasoning_content."""
+    model = _make_model()
+    response = {
+        "choices": [
+            {
+                "message": {
+                    "role": "assistant",
+                    "content": "The weather is sunny.",
+                    "reasoning_content": "The tool returned sunny weather.",
+                },
+                "finish_reason": "stop",
+            }
+        ],
+        "model": "step-3.7-flash",
+    }
+
+    result = model._create_chat_result(response)
+    message = result.generations[0].message
+
+    assert message.content == "The weather is sunny."
+    assert message.additional_kwargs["reasoning_content"] == "The tool returned sunny weather."
+
+
+def test_create_chat_result_reads_reasoning_from_sdk_object():
+    """When the response is a Pydantic model, reasoning is an attribute."""
+    model = _make_model()
+
+    class FakeMessage:
+        reasoning = "Reasoning stored on the SDK message object."
+        reasoning_content = None
+        model_extra = None
+
+    class FakeChoice:
+        message = FakeMessage()
+
+    class FakeResponse:
+        choices = [FakeChoice()]
+
+        def model_dump(self, **kwargs):
+            return {
+                "choices": [
+                    {
+                        "message": {
+                            "role": "assistant",
+                            "content": "Answer.",
+                        },
+                        "finish_reason": "stop",
+                    }
+                ],
+                "model": "step-3.7-flash",
+            }
+
+    result = model._create_chat_result(FakeResponse())
+    assert result.generations[0].message.additional_kwargs["reasoning_content"] == "Reasoning stored on the SDK message object."
+
+
+def test_create_chat_result_noop_when_no_reasoning():
+    model = _make_model()
+    response = {
+        "choices": [
+            {
+                "message": {
+                    "role": "assistant",
+                    "content": "Hello!",
+                },
+                "finish_reason": "stop",
+            }
+        ],
+        "model": "step-3.7-flash",
+    }
+
+    result = model._create_chat_result(response)
+    assert "reasoning_content" not in result.generations[0].message.additional_kwargs

config.example.yaml

@@ -274,6 +274,32 @@ models:
   #       thinking:
   #         type: disabled
 
+  # Example: StepFun (阶跃星辰) reasoning models
+  # StepFun provides OpenAI-compatible API with reasoning models.
+  # With reasoning_format: deepseek-style, the API returns reasoning_content
+  # (same field as DeepSeek), which must be replayed on historical assistant
+  # messages in multi-turn tool-call conversations.
+  # Use PatchedChatStepFun instead of plain ChatOpenAI.
+  # Docs: https://platform.stepfun.com/docs/api-reference/chat-completions
+  # - name: step-3.7-flash
+  #   display_name: Step 3.7 Flash
+  #   use: deerflow.models.patched_stepfun:PatchedChatStepFun
+  #   model: step-3.7-flash
+  #   api_key: $STEPFUN_API_KEY
+  #   base_url: https://api.stepfun.com/v1
+  #   request_timeout: 600.0
+  #   max_retries: 2
+  #   max_tokens: 4096
+  #   supports_thinking: true
+  #   supports_reasoning_effort: true
+  #   supports_vision: true
+  #   when_thinking_enabled:
+  #     extra_body:
+  #       reasoning_format: deepseek-style
+  #   when_thinking_disabled:
+  #     extra_body:
+  #       reasoning_format: deepseek-style
+
   # Example: MiniMax (OpenAI-compatible) - International Edition
   # MiniMax provides high-performance models with 512K context window and 128K max output
   # Docs: https://platform.minimax.io/docs/api-reference/text-openai-api