fix(providers): sanitize thinking_blocks by provider and harden content normalization

nanobot/providers/base.py

@@ -78,6 +78,12 @@ class LLMProvider(ABC):
                     result.append(clean)
                     continue
 
+            if isinstance(content, dict):
+                clean = dict(msg)
+                clean["content"] = [content]
+                result.append(clean)
+                continue
+
             result.append(msg)
         return result
 

nanobot/providers/litellm_provider.py

@@ -12,9 +12,9 @@ from litellm import acompletion
 from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
 from nanobot.providers.registry import find_by_model, find_gateway
 
-# Standard OpenAI chat-completion message keys plus reasoning_content for
+# Standard chat-completion message keys.
-# thinking-enabled models (Kimi k2.5, DeepSeek-R1, etc.).
+_ALLOWED_MSG_KEYS = frozenset({"role", "content", "tool_calls", "tool_call_id", "name", "reasoning_content"})
-_ALLOWED_MSG_KEYS = frozenset({"role", "content", "tool_calls", "tool_call_id", "name", "reasoning_content", "thinking_blocks"})
+_ANTHROPIC_EXTRA_KEYS = frozenset({"thinking_blocks"})
 _ALNUM = string.ascii_letters + string.digits
 
 def _short_tool_id() -> str:
@@ -158,11 +158,20 @@ class LiteLLMProvider(LLMProvider):
                     return
 
     @staticmethod
-    def _sanitize_messages(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    def _extra_msg_keys(original_model: str, resolved_model: str) -> frozenset[str]:
+        """Return provider-specific extra keys to preserve in request messages."""
+        spec = find_by_model(original_model) or find_by_model(resolved_model)
+        if (spec and spec.name == "anthropic") or "claude" in original_model.lower() or resolved_model.startswith("anthropic/"):
+            return _ANTHROPIC_EXTRA_KEYS
+        return frozenset()
+
+    @staticmethod
+    def _sanitize_messages(messages: list[dict[str, Any]], extra_keys: frozenset[str] = frozenset()) -> list[dict[str, Any]]:
         """Strip non-standard keys and ensure assistant messages have a content key."""
+        allowed = _ALLOWED_MSG_KEYS | extra_keys
         sanitized = []
         for msg in messages:
-            clean = {k: v for k, v in msg.items() if k in _ALLOWED_MSG_KEYS}
+            clean = {k: v for k, v in msg.items() if k in allowed}
             # Strict providers require "content" even when assistant only has tool_calls
             if clean.get("role") == "assistant" and "content" not in clean:
                 clean["content"] = None
@@ -193,6 +202,7 @@ class LiteLLMProvider(LLMProvider):
         """
         original_model = model or self.default_model
         model = self._resolve_model(original_model)
+        extra_msg_keys = self._extra_msg_keys(original_model, model)
 
         if self._supports_cache_control(original_model):
             messages, tools = self._apply_cache_control(messages, tools)
@@ -203,7 +213,7 @@ class LiteLLMProvider(LLMProvider):
 
         kwargs: dict[str, Any] = {
             "model": model,
-            "messages": self._sanitize_messages(self._sanitize_empty_content(messages)),
+            "messages": self._sanitize_messages(self._sanitize_empty_content(messages), extra_keys=extra_msg_keys),
             "max_tokens": max_tokens,
             "temperature": temperature,
         }