fix: use reasoning_content for Open WebUI streaming

Use DeepSeek R1 format (reasoning_content field) instead of <think>
tags in content. Open WebUI now renders thinking as proper
collapsible blocks instead of broken escaped HTML.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -7,6 +7,60 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [1.8.3] - 2025-12-16
+
+### Fixed
+
+- **Open WebUI streaming rendering** - Use `reasoning_content` field for thinking (DeepSeek R1 format) instead of `<think>` tags in `content`
+- Open WebUI now renders thinking as proper collapsible blocks instead of broken HTML
+
+## [1.8.2] - 2025-12-16
+
+### Fixed
+
+- **HybridRAG keywords schema mismatch** - library-desk now returns `keywords` as dict with `core_keywords`, client now handles both formats
+
+## [1.8.1] - 2025-12-16
+
+### Fixed
+
+#### Ollama Message Sanitization
+- **Fixed `invalid message content type: <nil>` error** from Ollama
+- Created custom `TatlockOllamaProvider` that sanitizes messages before sending to Ollama
+- Ollama rejects assistant messages with `content: null` (tool-only messages from PydanticAI)
+- Provider converts `null` content to empty string `""` for compatibility
+- Updated all agents (Librarian, Biographer, Housekeeper, Tatlock) to use sanitized provider
+- Added `src/ollama/provider.py` with reusable provider pattern
+
+#### Streaming Think Message Accumulation
+- **Fixed repeating think messages in frontend** (e.g., 10x "The Librarian has compiled...")
+- Frontend was accumulating `ReasoningSummaryDelta` events expecting concatenation
+- Added `ReasoningSummaryDone()` signal after each think message to indicate completion
+- Each think slug is now treated as a complete message, not a continuation
+
+## [1.8.0] - 2025-12-15
+
+### Fixed
+
+#### Steward Routing for Web Search
+- Updated Steward guidelines to route web searches, weather, news → Librarian with `search_web`
+- Added URL/article reading → Librarian with `read_url` to routing guidelines
+- Added examples showing `search_web` and `read_url` tool usage
+
+#### Librarian Agent Tool Registration
+- Registered `search_web`, `read_url`, `read_urls_batch` tools with the Librarian PydanticAI agent
+- Updated Librarian system prompt with Web Search & Content Extraction section
+- Fixed tool count in agent logger (11 → 14 tools)
+
+#### Query Enrichment Integration
+- Fixed enriched query (with location/timezone context) not being passed to delegations
+- Response service now uses `enriched_query` from Steward recommendation for all delegations
+- Weather queries now automatically include user's stored location
+
+#### Action Type Detection
+- Added "read", "fetch", "url", "http" keywords to RESEARCH action type for Librarian
+- Ensures proper think messages for URL reading tasks
+
 ## [1.7.0] - 2025-12-15
 
 ### Added

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "tatlock"
-version = "1.7.0"
+version = "1.8.3"
 description = "OpenAI-compatible API with Ollama backend"
 requires-python = ">=3.12"
 dependencies = []

src/agents/biographer/agent.py

@@ -102,18 +102,15 @@ _biographer_agent: Optional[Agent[None, str]] = None
 
 def _create_biographer_agent() -> Agent[None, str]:
     """Create The Biographer PydanticAI agent."""
-    # Import required classes for Ollama configuration
     from pydantic_ai.models.openai import OpenAIChatModel
-    from pydantic_ai.providers.ollama import OllamaProvider
 
-    # PydanticAI expects Ollama base URL to end with /v1
-    clean_host = str(config.OLLAMA_HOST).rstrip('/')
-    base_url = f"{clean_host}/v1"
+    from src.ollama.provider import get_ollama_provider
 
-    # Create Ollama model with provider
+    # Create Ollama model with sanitized provider
+    # (fixes 'content: null' issue with tool calls)
     model = OpenAIChatModel(
         model_name=config.OLLAMA_DEFAULT_MODEL,
-        provider=OllamaProvider(base_url=base_url)
+        provider=get_ollama_provider(),
     )
 
     agent: Agent[None, str] = Agent(

src/agents/delegation.py

@@ -100,10 +100,13 @@ def _detect_action_type(expert: str, task: str) -> ActionType:
     task_lower = task.lower()
 
     if expert == "librarian":
+        # Web search, URL reading = RESEARCH (fresh external data)
         if any(w in task_lower for w in ["search", "find", "look up", "research"]):
             if any(w in task_lower for w in ["web", "online", "internet"]):
                 return ActionType.RESEARCH
             return ActionType.RETRIEVE
+        if any(w in task_lower for w in ["read", "fetch", "url", "http"]):
+            return ActionType.RESEARCH  # Reading URLs is research
         if any(w in task_lower for w in ["create", "write", "add", "make", "new"]):
             return ActionType.CREATE
         return ActionType.RETRIEVE

src/agents/housekeeper/agent.py

@@ -113,18 +113,15 @@ _housekeeper_agent: Optional[Agent[None, str]] = None
 
 def _create_housekeeper_agent() -> Agent[None, str]:
     """Create the Housekeeper PydanticAI agent."""
-    # Import required classes for Ollama configuration
     from pydantic_ai.models.openai import OpenAIChatModel
-    from pydantic_ai.providers.ollama import OllamaProvider
 
-    # PydanticAI expects Ollama base URL to end with /v1
-    clean_host = str(config.OLLAMA_HOST).rstrip("/")
-    base_url = f"{clean_host}/v1"
+    from src.ollama.provider import get_ollama_provider
 
-    # Create Ollama model with provider
+    # Create Ollama model with sanitized provider
+    # (fixes 'content: null' issue with tool calls)
     model = OpenAIChatModel(
         model_name=config.OLLAMA_DEFAULT_MODEL,
-        provider=OllamaProvider(base_url=base_url),
+        provider=get_ollama_provider(),
     )
 
     agent: Agent[None, str] = Agent(

src/agents/librarian/agent.py

@@ -19,6 +19,9 @@ from src.agents.librarian.tools import (
     get_wiki_page,
     hybrid_search,
     list_dossiers,
+    read_url,
+    read_urls_batch,
+    search_web,
     search_wiki,
     semantic_search,
     smart_create_wiki_page,
@@ -47,8 +50,17 @@ Your role is to help users find, understand, synthesize, and manage information
 
 ## Your Tools
 
-### Research Tools
-- **hybrid_search**: Your primary research tool - searches all sources at once
+### Web Search & Content Extraction
+- **search_web**: Search the internet for current information (weather, news, facts)
+  - Use for: weather forecasts, current events, recent developments, external facts
+  - Returns extracted content from search results, not just snippets
+- **read_url**: Read and extract content from a specific URL
+  - Use when: user provides a URL or you need to read a specific webpage
+- **read_urls_batch**: Read multiple URLs in parallel (up to 20)
+  - Use for: comparing multiple sources, gathering info from several pages
+
+### Internal Research Tools
+- **hybrid_search**: Your primary research tool - searches wiki, graph, and web at once
 - **search_wiki**: Find specific wiki pages by keyword
 - **semantic_search**: Find conceptually similar content
 - **explore_knowledge_graph** / **find_related_entities**: Discover connections
@@ -110,18 +122,15 @@ _librarian_agent: Optional[Agent[None, str]] = None
 
 def _create_librarian_agent() -> Agent[None, str]:
     """Create the Librarian PydanticAI agent."""
-    # Import required classes for Ollama configuration
     from pydantic_ai.models.openai import OpenAIChatModel
-    from pydantic_ai.providers.ollama import OllamaProvider
 
-    # PydanticAI expects Ollama base URL to end with /v1
-    clean_host = str(config.OLLAMA_HOST).rstrip('/')
-    base_url = f"{clean_host}/v1"
+    from src.ollama.provider import get_ollama_provider
 
-    # Create Ollama model with provider
+    # Create Ollama model with sanitized provider
+    # (fixes 'content: null' issue with tool calls)
     model = OpenAIChatModel(
         model_name=config.OLLAMA_DEFAULT_MODEL,
-        provider=OllamaProvider(base_url=base_url)
+        provider=get_ollama_provider(),
     )
 
     agent: Agent[None, str] = Agent(
@@ -130,7 +139,7 @@ def _create_librarian_agent() -> Agent[None, str]:
         retries=2,
     )
 
-    # Register research tools
+    # Register research tools (internal knowledge)
     agent.tool_plain(hybrid_search)
     agent.tool_plain(search_wiki)
     agent.tool_plain(semantic_search)
@@ -139,6 +148,11 @@ def _create_librarian_agent() -> Agent[None, str]:
     agent.tool_plain(explore_knowledge_graph)
     agent.tool_plain(find_related_entities)
 
+    # Register web search & content extraction tools
+    agent.tool_plain(search_web)
+    agent.tool_plain(read_url)
+    agent.tool_plain(read_urls_batch)
+
     # Register wiki read tools
     agent.tool_plain(get_wiki_page)
 
@@ -150,7 +164,7 @@ def _create_librarian_agent() -> Agent[None, str]:
     logger.info(
         "librarian_agent_created",
         model=config.OLLAMA_DEFAULT_MODEL,
-        tool_count=11,
+        tool_count=14,  # 7 research + 3 web + 1 wiki read + 3 wiki write
     )
 
     return agent

src/agents/librarian/client.py

@@ -281,9 +281,16 @@ class LibraryDeskClient:
                 metadata=r.get("metadata", {}),
             ))
 
+        # Handle keywords being either a list or a dict with core_keywords
+        raw_keywords = data.get("keywords", [])
+        if isinstance(raw_keywords, dict):
+            keywords = raw_keywords.get("core_keywords", [])
+        else:
+            keywords = raw_keywords
+
         return HybridRAGResponse(
             results=results,
-            keywords=data.get("keywords", []),
+            keywords=keywords,
             synonyms=data.get("synonyms", []),
             related_dossiers=data.get("related_dossiers", []),
             formatted_context=data.get("formatted_context", ""),

src/agents/steward/agent.py

@@ -58,8 +58,9 @@ GUIDELINES:
 - Simple greetings/chat → no capabilities needed (conversational response only)
 - Questions about prior conversation ("what did I say", "my name", "what we discussed") → no capabilities (Tatlock has full history)
 - Math/calculations → tatlock_core
-- Quick web searches → tatlock_core
 - Time/date queries → tatlock_core
+- Web searches, weather, news, current information → librarian with search_web
+- Read a URL or article → librarian with read_url
 - Wiki creation ("create a page about X", "add X to wiki") → librarian with smart_create
 - Wiki updates ("update the page", "add to dossier") → librarian with update
 - Research queries ("find info", "what do we know about", "search for") → librarian with hybrid_search
@@ -74,8 +75,10 @@ COMPLEXITY: [simple/moderate/complex]
 CONTEXT: [any relevant conversation context, or "none"]
 
 EXAMPLES:
+- "DELEGATE: librarian to search_web for tomorrow's weather forecast"
 - "DELEGATE: librarian to create a wiki page about CI/CD pipelines"
-- "DELEGATE: librarian to search for information about Docker networking"
+- "DELEGATE: librarian to hybrid_search for information about Docker networking"
+- "DELEGATE: librarian to read_url https://example.com/article"
 - "DELEGATE: tatlock_core to calculate the result"
 - "DELEGATE: none (conversational response only)"
 

src/agents/tatlock.py

@@ -143,7 +143,7 @@ class TatlockAgent(AgentInterface):
 
         # Import required classes for Ollama configuration
         from pydantic_ai.models.openai import OpenAIChatModel
-        from pydantic_ai.providers.ollama import OllamaProvider
+        from src.ollama.provider import get_ollama_provider
 
         # PydanticAI expects Ollama base URL to end with /v1
         # Remove trailing slash from ollama_host if present
@@ -153,7 +153,7 @@ class TatlockAgent(AgentInterface):
         # Create Ollama model with provider
         ollama_model = OpenAIChatModel(
             model_name=self.model_name,
-            provider=OllamaProvider(base_url=base_url)
+            provider=get_ollama_provider()
         )
 
         # Create PydanticAI agent with Ollama model
@@ -448,7 +448,7 @@ class TatlockAgent(AgentInterface):
             ... )
         """
         from pydantic_ai.models.openai import OpenAIChatModel
-        from pydantic_ai.providers.ollama import OllamaProvider
+        from src.ollama.provider import get_ollama_provider
 
         logger.info(
             "tatlock_run_with_scoped_tools",
@@ -464,7 +464,7 @@ class TatlockAgent(AgentInterface):
 
         ollama_model = OpenAIChatModel(
             model_name=self.model_name,
-            provider=OllamaProvider(base_url=base_url)
+            provider=get_ollama_provider()
         )
 
         # Create agent with scoped tools
@@ -542,7 +542,7 @@ class TatlockAgent(AgentInterface):
             Text chunks from the streaming response
         """
         from pydantic_ai.models.openai import OpenAIChatModel
-        from pydantic_ai.providers.ollama import OllamaProvider
+        from src.ollama.provider import get_ollama_provider
 
         logger.info(
             "tatlock_run_with_scoped_tools_stream",
@@ -557,7 +557,7 @@ class TatlockAgent(AgentInterface):
 
         ollama_model = OpenAIChatModel(
             model_name=self.model_name,
-            provider=OllamaProvider(base_url=base_url)
+            provider=get_ollama_provider()
         )
 
         # Create agent with scoped tools
@@ -637,7 +637,7 @@ class TatlockAgent(AgentInterface):
                 - raw_output: The agent's raw text output
         """
         from pydantic_ai.models.openai import OpenAIChatModel
-        from pydantic_ai.providers.ollama import OllamaProvider
+        from src.ollama.provider import get_ollama_provider
         from pydantic_ai.settings import ModelSettings
         from pydantic_ai.messages import (
             ModelRequest,
@@ -661,7 +661,7 @@ class TatlockAgent(AgentInterface):
 
         ollama_model = OpenAIChatModel(
             model_name=self.model_name,
-            provider=OllamaProvider(base_url=base_url)
+            provider=get_ollama_provider()
         )
 
         # Create agent with scoped tools
@@ -759,7 +759,7 @@ class TatlockAgent(AgentInterface):
             str: Butler-toned response synthesized from all results
         """
         from pydantic_ai.models.openai import OpenAIChatModel
-        from pydantic_ai.providers.ollama import OllamaProvider
+        from src.ollama.provider import get_ollama_provider
         from pydantic_ai.messages import ModelRequest, ModelResponse, UserPromptPart, TextPart
 
         logger.info(
@@ -802,7 +802,7 @@ class TatlockAgent(AgentInterface):
 
         ollama_model = OpenAIChatModel(
             model_name=self.model_name,
-            provider=OllamaProvider(base_url=base_url)
+            provider=get_ollama_provider()
         )
 
         # Synthesis agent uses butler prompt but no tools

src/chat/schemas.py

@@ -55,6 +55,7 @@ class ChatCompletionChunkDelta(CustomBaseModel):
     """Delta in streaming chunk."""
     role: str | None = None
     content: str | None = None
+    reasoning_content: str | None = None  # For thinking/reasoning (DeepSeek R1 format)
 
 
 class ChatCompletionChunkChoice(CustomBaseModel):

src/chat/service.py

@@ -172,24 +172,9 @@ async def create_chat_completion_stream(
 
     async for event in stream_generator:
         if event.event == StreamEventType.REASONING_SUMMARY_DELTA:
-            # Start <think> block if needed
-            if not in_reasoning:
-                yield ChatCompletionChunk(
-                    id=completion_id,
-                    object=constants.CHAT_COMPLETION_CHUNK_OBJECT,
-                    created=created_at,
-                    model=request.model,
-                    choices=[
-                        ChatCompletionChunkChoice(
-                            index=0,
-                            delta=ChatCompletionChunkDelta(content="<think>\n"),
-                            finish_reason=None,
-                        )
-                    ],
-                )
-                in_reasoning = True
-
-            # Stream reasoning delta
+            # Stream reasoning via reasoning_content field (DeepSeek R1 format)
+            # Open WebUI renders this as collapsible thinking block
+            in_reasoning = True
             yield ChatCompletionChunk(
                 id=completion_id,
                 object=constants.CHAT_COMPLETION_CHUNK_OBJECT,
@@ -198,29 +183,15 @@ async def create_chat_completion_stream(
                 choices=[
                     ChatCompletionChunkChoice(
                         index=0,
-                        delta=ChatCompletionChunkDelta(content=event.delta),
+                        delta=ChatCompletionChunkDelta(reasoning_content=event.delta),
                         finish_reason=None,
                     )
                 ],
             )
 
         elif event.event == StreamEventType.REASONING_SUMMARY_DONE:
-            # Close <think> block
-            if in_reasoning:
-                yield ChatCompletionChunk(
-                    id=completion_id,
-                    object=constants.CHAT_COMPLETION_CHUNK_OBJECT,
-                    created=created_at,
-                    model=request.model,
-                    choices=[
-                        ChatCompletionChunkChoice(
-                            index=0,
-                            delta=ChatCompletionChunkDelta(content="</think>\n\n"),
-                            finish_reason=None,
-                        )
-                    ],
-                )
-                in_reasoning = False
+            # Signal end of reasoning block (no content needed)
+            in_reasoning = False
 
         elif event.event == StreamEventType.OUTPUT_TEXT_DELTA:
             # Stream message content

src/ollama/provider.py

@@ -0,0 +1,130 @@
+"""
+PydanticAI provider for Ollama with message sanitization.
+
+Ollama's OpenAI-compatible API rejects messages with `content: null`,
+which PydanticAI sends for assistant messages that only contain tool calls.
+This provider sanitizes messages to use empty strings instead of null.
+"""
+from typing import Any
+
+from openai import AsyncOpenAI
+from pydantic_ai.providers.ollama import OllamaProvider
+
+from src.core.config import config
+from src.core.logging_config import get_logger
+
+logger = get_logger(__name__)
+
+
+class TatlockOllamaProvider(OllamaProvider):
+    """
+    Custom OllamaProvider with message sanitization for Tatlock agents.
+
+    Fixes the 'invalid message content type: <nil>' error that occurs
+    when assistant messages have `content: null` with tool calls.
+    """
+
+    def __init__(self, base_url: str | None = None):
+        """
+        Initialize provider with Ollama base URL.
+
+        Args:
+            base_url: Ollama API URL (defaults to config.OLLAMA_HOST/v1)
+        """
+        if base_url is None:
+            clean_host = str(config.OLLAMA_HOST).rstrip("/")
+            base_url = f"{clean_host}/v1"
+
+        super().__init__(base_url=base_url)
+
+        # Override the client with our sanitized version
+        self._openai_client = _SanitizedAsyncOpenAI(base_url=base_url)
+
+        logger.debug("tatlock_ollama_provider_created", base_url=base_url)
+
+
+class _SanitizedAsyncOpenAI(AsyncOpenAI):
+    """AsyncOpenAI client that sanitizes messages before sending."""
+
+    def __init__(self, **kwargs: Any):
+        # Ollama doesn't need an API key
+        super().__init__(api_key="ollama", **kwargs)
+
+    @property
+    def chat(self) -> "_SanitizedChat":
+        """Return sanitized chat interface."""
+        return _SanitizedChat(self)
+
+
+class _SanitizedChat:
+    """Chat interface wrapper with sanitized completions."""
+
+    def __init__(self, client: _SanitizedAsyncOpenAI):
+        self._client = client
+        self._original_chat = AsyncOpenAI.chat.fget(client)  # type: ignore
+
+    @property
+    def completions(self) -> "_SanitizedCompletions":
+        """Return sanitized completions interface."""
+        return _SanitizedCompletions(self._original_chat.completions)
+
+
+class _SanitizedCompletions:
+    """Completions wrapper that sanitizes messages before API calls."""
+
+    def __init__(self, original_completions: Any):
+        self._original = original_completions
+
+    async def create(self, **kwargs: Any) -> Any:
+        """
+        Create chat completion with sanitized messages.
+
+        Converts `content: null` to `content: ""` in assistant messages
+        to prevent Ollama's 'invalid message content type: <nil>' error.
+        """
+        if "messages" in kwargs:
+            kwargs["messages"] = _sanitize_messages(kwargs["messages"])
+
+        return await self._original.create(**kwargs)
+
+
+def _sanitize_messages(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    """
+    Sanitize messages to fix null content issues.
+
+    When an assistant message has tool_calls but no text content,
+    PydanticAI sets content to None. Ollama rejects this.
+    We convert None to empty string.
+
+    Args:
+        messages: List of chat messages
+
+    Returns:
+        Sanitized messages with null content replaced by empty strings
+    """
+    sanitized = []
+    for msg in messages:
+        msg_copy = dict(msg)
+
+        # Fix null content in assistant messages with tool calls
+        if msg_copy.get("role") == "assistant":
+            if msg_copy.get("content") is None and msg_copy.get("tool_calls"):
+                msg_copy["content"] = ""
+                logger.debug(
+                    "sanitized_null_content",
+                    tool_call_count=len(msg_copy["tool_calls"]),
+                )
+
+        sanitized.append(msg_copy)
+
+    return sanitized
+
+
+def get_ollama_provider() -> TatlockOllamaProvider:
+    """
+    Get a configured Ollama provider for PydanticAI agents.
+
+    Returns:
+        TatlockOllamaProvider configured with sanitization
+    """
+    return TatlockOllamaProvider()

src/responses/service.py

@@ -515,15 +515,18 @@ async def create_response_with_steward(request: ResponseRequest) -> Response:
     from src.agents.tatlock import TatlockAgent
     tatlock = TatlockAgent()
 
+    # Use enriched query (with location/timezone context) if available
+    effective_query = enriched.recommendation.enriched_query or user_message
+
     if delegation_only:
         # Direct delegation path - collect results then synthesize
         orchestration_results = await _direct_delegation_with_results(
-            user_message, enriched.recommendation, tracker, conversation_id
+            effective_query, enriched.recommendation, tracker, conversation_id
         )
     else:
         # Phase 1: Orchestrate tool calls
         orchestration_results = await tatlock.orchestrate_tool_calls(
-            user_message=user_message,
+            user_message=effective_query,
             steward_note=enriched.steward_note,
             scoped_tools=enriched.scoped_tools,
             message_history=conversation_history,

src/responses/streaming.py

@@ -211,8 +211,10 @@ class StreamingCoordinator:
                 )
 
                 # Stream think slugs that were collected during delegation
+                # Each think message is complete, so we signal done after each
                 for think_msg in orchestration_results.get("think_messages", []):
                     yield ReasoningSummaryDelta(delta=think_msg)
+                    yield ReasoningSummaryDone()
                     await asyncio.sleep(0.05)
 
             else:

tests/chat/test_streaming_wrapper.py

@@ -4,7 +4,7 @@ Tests for chat completions streaming wrapper.
 Tests that the wrapper correctly:
 - Wraps Responses API
 - Enables reasoning automatically
-- Converts reasoning to <think> tags
+- Streams reasoning via reasoning_content field (DeepSeek R1 format)
 - Streams both reasoning and content
 """
 import json
@@ -17,7 +17,7 @@ from src.chat import constants
 @pytest.mark.unit
 @pytest.mark.asyncio
 async def test_streaming_wrapper_enables_reasoning(async_client: AsyncClient):
-    """Test that streaming wrapper automatically enables reasoning."""
+    """Test that streaming wrapper automatically enables reasoning via reasoning_content."""
     request_data = {
         "model": "lorem-tester",
         "messages": [
@@ -27,7 +27,7 @@ async def test_streaming_wrapper_enables_reasoning(async_client: AsyncClient):
     }
 
     chunks_received = []
-    think_tags_found = False
+    reasoning_content_found = False
 
     async with async_client.stream(
         "POST",
@@ -51,12 +51,12 @@ async def test_streaming_wrapper_enables_reasoning(async_client: AsyncClient):
                     chunk = json.loads(data_str)
                     chunks_received.append(chunk)
 
-                    # Check for <think> tags in delta content
+                    # Check for reasoning_content in delta (DeepSeek R1 format)
                     if "choices" in chunk and len(chunk["choices"]) > 0:
                         delta = chunk["choices"][0].get("delta", {})
-                        content = delta.get("content")
-                        if content and ("<think>" in content or "</think>" in content):
-                            think_tags_found = True
+                        reasoning = delta.get("reasoning_content")
+                        if reasoning:
+                            reasoning_content_found = True
 
                 except json.JSONDecodeError:
                     pass
@@ -64,14 +64,14 @@ async def test_streaming_wrapper_enables_reasoning(async_client: AsyncClient):
     # Should have received chunks
     assert len(chunks_received) > 0
 
-    # Should have found <think> tags (reasoning enabled automatically)
-    assert think_tags_found, "Expected <think> tags in streaming output"
+    # Should have found reasoning_content (reasoning enabled automatically)
+    assert reasoning_content_found, "Expected reasoning_content in streaming output"
 
 
 @pytest.mark.unit
 @pytest.mark.asyncio
 async def test_streaming_wrapper_reasoning_before_content(async_client: AsyncClient):
-    """Test that reasoning (<think> tags) comes before actual content."""
+    """Test that reasoning_content comes before regular content."""
     request_data = {
         "model": "lorem-tester",
         "messages": [
@@ -80,10 +80,7 @@ async def test_streaming_wrapper_reasoning_before_content(async_client: AsyncCli
         "stream": True
     }
 
-    all_content = []
-    found_think_opening = False
-    found_think_closing = False
-    found_content_after_think = False
+    chunk_types = []  # Track order: 'reasoning' or 'content'
 
     async with async_client.stream(
         "POST",
@@ -106,28 +103,22 @@ async def test_streaming_wrapper_reasoning_before_content(async_client: AsyncCli
                     chunk = json.loads(data_str)
                     if "choices" in chunk and len(chunk["choices"]) > 0:
                         delta = chunk["choices"][0].get("delta", {})
-                        content = delta.get("content", "")
-                        if content:
-                            all_content.append(content)
+                        reasoning = delta.get("reasoning_content")
+                        content = delta.get("content")
 
-                            if "<think>" in content:
-                                found_think_opening = True
-                            if "</think>" in content:
-                                found_think_closing = True
-                            # Content after closing think tag
-                            if found_think_closing and content.strip() and "<think>" not in content and "</think>" not in content:
-                                found_content_after_think = True
+                        if reasoning:
+                            chunk_types.append("reasoning")
+                        if content:
+                            chunk_types.append("content")
 
                 except json.JSONDecodeError:
                     pass
 
-    # Verify ordering
-    full_text = "".join(all_content)
-    if found_think_opening and found_think_closing:
-        # Reasoning should come before main content
-        think_start = full_text.index("<think>")
-        think_end = full_text.index("</think>")
-        assert think_start < think_end, "Opening <think> should come before closing </think>"
+    # Verify reasoning comes before content
+    if "reasoning" in chunk_types and "content" in chunk_types:
+        first_reasoning = chunk_types.index("reasoning")
+        first_content = chunk_types.index("content")
+        assert first_reasoning < first_content, "reasoning_content should come before content"
 
 
 @pytest.mark.unit