fix: preserve empty-string reasoning_content instead of coercing to None

Custom providers (e.g. DeepSeek) may return reasoning_content as an
empty string "" to explicitly indicate no reasoning occurred. The
previous truthiness checks (, ) treated "" as falsy
and converted it to None, which caused the field to be dropped from
the message history entirely. Providers that require reasoning_content
on all assistant messages then rejected subsequent requests.

Replace truthiness checks with identity checks () so that
empty-string reasoning_content is preserved as-is. The streaming path
is unchanged since an empty join genuinely means no chunks received.

Fixes #4105

nanobot/providers/openai_compat_provider.py

@@ -999,7 +999,7 @@ class OpenAICompatProvider(LLMProvider):
             if not content and msg0.get("reasoning") and self._spec and self._spec.reasoning_as_content:
                 content = self._extract_text_content(msg0.get("reasoning"))
             reasoning_content = msg0.get("reasoning_content")
-            if not reasoning_content and msg0.get("reasoning"):
+            if reasoning_content is None and msg0.get("reasoning"):
                 reasoning_content = self._extract_text_content(msg0.get("reasoning"))
             for ch in choices:
                 ch_map = self._maybe_mapping(ch) or {}
@@ -1011,7 +1011,7 @@ class OpenAICompatProvider(LLMProvider):
                         finish_reason = str(ch_map["finish_reason"])
                 if not content:
                     content = self._extract_text_content(m.get("content"))
-                if not reasoning_content:
+                if reasoning_content is None:
                     reasoning_content = m.get("reasoning_content")
 
             parsed_tool_calls = []
@@ -1074,8 +1074,8 @@ class OpenAICompatProvider(LLMProvider):
                 function_provider_specific_fields=fn_prov,
             ))
 
-        reasoning_content = getattr(msg, "reasoning_content", None) or None
-        if not reasoning_content and getattr(msg, "reasoning", None):
+        reasoning_content = getattr(msg, "reasoning_content", None)
+        if reasoning_content is None and getattr(msg, "reasoning", None):
             reasoning_content = msg.reasoning
 
         return LLMResponse(

tests/providers/test_reasoning_content.py

@@ -9,7 +9,6 @@ from unittest.mock import patch
 
 from nanobot.providers.openai_compat_provider import OpenAICompatProvider
 
-
 # ── _parse: non-streaming ─────────────────────────────────────────────────
 
 
@@ -52,6 +51,32 @@ def test_parse_dict_reasoning_content_none_when_absent() -> None:
     assert result.reasoning_content is None
 
 
+def test_parse_dict_reasoning_content_empty_string_preserved() -> None:
+    """reasoning_content=\"\" is preserved, not coerced to None.
+
+    Some providers (e.g. DeepSeek) require the reasoning_content key to
+    be present in subsequent requests even when empty.  Coercing \"\" to
+    None drops the key downstream and causes API errors.
+    """
+    with patch("nanobot.providers.openai_compat_provider.AsyncOpenAI"):
+        provider = OpenAICompatProvider()
+
+    response = {
+        "choices": [{
+            "message": {
+                "content": "answer",
+                "reasoning_content": "",
+            },
+            "finish_reason": "stop",
+        }],
+        "usage": {"prompt_tokens": 5, "completion_tokens": 3, "total_tokens": 8},
+    }
+
+    result = provider._parse(response)
+
+    assert result.reasoning_content == ""
+
+
 # ── _parse_chunks: streaming dict branch ─────────────────────────────────