Prevent sending thinking blocks with no signature with ChatBedrockConverse (#408)

Fixes #401

Claude 3.7 still works because signatures are always returned with those
thinking blocks. DeepSeek thinking blocks do not contain signatures and
the ValidationError indicates a signature must be included with thinking
blocks.

According to Converse API docs:
> "Within the reasoningText field, the text fields describes the
reasoning. The signature field is a hash of all the messages in the
conversation and is a safeguard against tampering of the reasoning used
by the model. You must include the signature and all previous messages
in subsequent Converse requests. If any of the messages are changed, the
response throws an error."


https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference-call.html

@3coins I posted a separate PR
#407 to make reviewing
easier. Happy to combine if preferred.

Author: tahouse

libs/aws/langchain_aws/chat_models/bedrock_converse.py

@@ -460,9 +460,12 @@ class Joke(BaseModel):
     def set_disable_streaming(cls, values: Dict) -> Any:
         model_id = values.get("model_id", values.get("model"))
         model_parts = model_id.split(".")
-        
-        # Extract provider from the model_id (e.g., "amazon", "anthropic", "ai21", "meta", "mistral")
-        provider = values.get("provider") or (model_parts[-2] if len(model_parts) > 1 else model_parts[0])
+
+        # Extract provider from the model_id
+        # (e.g., "amazon", "anthropic", "ai21", "meta", "mistral")
+        provider = values.get("provider") or (
+            model_parts[-2] if len(model_parts) > 1 else model_parts[0]
+        )
         values["provider"] = provider
 
         model_id_lower = model_id.lower()
@@ -471,26 +474,46 @@ def set_disable_streaming(cls, values: Dict) -> Any:
         # Here we check based on the updated AWS documentation.
         if (
             # AI21 Jamba 1.5 models
-            (provider == "ai21" and "jamba-1-5" in model_id_lower) or
+            (provider == "ai21" and "jamba-1-5" in model_id_lower)
+            or
             # Some Amazon Nova models
-            (provider == "amazon" and any(x in model_id_lower for x in ["nova-lite", "nova-micro", "nova-pro"])) or
+            (
+                provider == "amazon"
+                and any(
+                    x in model_id_lower for x in ["nova-lite", "nova-micro", "nova-pro"]
+                )
+            )
+            or
             # Anthropic Claude 3 and newer models
-            (provider == "anthropic" and "claude-3" in model_id_lower) or
+            (provider == "anthropic" and "claude-3" in model_id_lower)
+            or
             # Cohere Command R models
             (provider == "cohere" and "command-r" in model_id_lower)
         ):
             streaming_support = True
         elif (
             # AI21 Jamba-Instruct model
-            (provider == "ai21" and "jamba-instruct" in model_id_lower) or
+            (provider == "ai21" and "jamba-instruct" in model_id_lower)
+            or
             # Amazon Titan Text models
-            (provider == "amazon" and "titan-text" in model_id_lower) or
+            (provider == "amazon" and "titan-text" in model_id_lower)
+            or
             # Anthropic older Claude models (Claude 2, Claude 2.1, Claude Instant)
-            (provider == "anthropic" and any(x in model_id_lower for x in ["claude-v2", "claude-instant"])) or
+            (
+                provider == "anthropic"
+                and any(x in model_id_lower for x in ["claude-v2", "claude-instant"])
+            )
+            or
             # Cohere Command (non-R) models
-            (provider == "cohere" and "command" in model_id_lower and "command-r" not in model_id_lower) or
+            (
+                provider == "cohere"
+                and "command" in model_id_lower
+                and "command-r" not in model_id_lower
+            )
+            or
             # All Meta Llama models
-            (provider == "meta") or
+            (provider == "meta")
+            or
             # All Mistral models
             (provider == "mistral") or
             # DeepSeek-R1 models
@@ -501,8 +524,10 @@ def set_disable_streaming(cls, values: Dict) -> Any:
             streaming_support = False
 
         # Set the disable_streaming flag accordingly:
-        # - If streaming is supported (plain streaming), we want streaming enabled (i.e. disable_streaming == False).
-        # - If the model supports streaming only in non-tool mode ("no_tools"), then we must force disable streaming when tools are used.
+        # - If streaming is supported (plain streaming),
+        #       we want streaming enabled (i.e. disable_streaming == False).
+        # - If the model supports streaming only in non-tool mode ("no_tools"),
+        #       then we must force disable streaming when tools are used.
         # - Otherwise, if streaming is not supported, we set disable_streaming to True.
         if "disable_streaming" not in values:
             if not streaming_support:
@@ -514,7 +539,6 @@ def set_disable_streaming(cls, values: Dict) -> Any:
 
         return values
 
-
     @model_validator(mode="after")
     def validate_environment(self) -> Self:
         """Validate that AWS credentials to and python package exists in environment."""
@@ -811,10 +835,13 @@ def _converse_params(
                 "modelId": modelId or self.model_id,
                 "inferenceConfig": inferenceConfig,
                 "toolConfig": toolConfig,
-                "additionalModelRequestFields": additionalModelRequestFields
-                or self.additional_model_request_fields,
-                "additionalModelResponseFieldPaths": additionalModelResponseFieldPaths
-                or self.additional_model_response_field_paths,
+                "additionalModelRequestFields": (
+                    additionalModelRequestFields or self.additional_model_request_fields
+                ),
+                "additionalModelResponseFieldPaths": (
+                    additionalModelResponseFieldPaths
+                    or self.additional_model_response_field_paths
+                ),
                 "guardrailConfig": guardrailConfig or self.guardrail_config,
                 "performanceConfig": performanceConfig or self.performance_config,
                 "requestMetadata": requestMetadata or self.request_metadata,
@@ -1082,28 +1109,30 @@ def _lc_content_to_bedrock(
         elif block["type"] == "guard_content":
             bedrock_content.append({"guardContent": {"text": {"text": block["text"]}}})
         elif block["type"] == "thinking":
-            bedrock_content.append(
-                {
-                    "reasoningContent": {
-                        "reasoningText": {
-                            "text": block.get("thinking", ""),
-                            "signature": block.get("signature", "")
+            if block.get("signature", ""):
+                bedrock_content.append(
+                    {
+                        "reasoningContent": {
+                            "reasoningText": {
+                                "text": block.get("thinking", ""),
+                                "signature": block.get("signature", ""),
+                            }
                         }
                     }
-                }
-            )
+                )
         elif block["type"] == "reasoning_content":
             reasoning_content = block.get("reasoningContent", {})
-            bedrock_content.append(
-                {
-                    "reasoningContent": {
-                        "reasoningText": {
-                            "text": reasoning_content.get("text", ""),
-                            "signature": reasoning_content.get("signature", "")
+            if reasoning_content.get("signature", ""):
+                bedrock_content.append(
+                    {
+                        "reasoningContent": {
+                            "reasoningText": {
+                                "text": reasoning_content.get("text", ""),
+                                "signature": reasoning_content.get("signature", ""),
+                            }
                         }
                     }
-                }
-            )
+                )
         else:
             raise ValueError(f"Unsupported content block type:\n{block}")
     # drop empty text blocks
@@ -1203,7 +1232,7 @@ def _bedrock_to_lc(content: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
                             "type": "reasoning_content",
                             "reasoning_content": {
                                 "type": "text",
-                                "text": reasoning_dict.get("text")
+                                "text": reasoning_dict.get("text"),
                             },
                         }
                     )
@@ -1213,7 +1242,7 @@ def _bedrock_to_lc(content: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
                             "type": "reasoning_content",
                             "reasoning_content": {
                                 "type": "signature",
-                                "signature": reasoning_dict.get("signature")
+                                "signature": reasoning_dict.get("signature"),
                             },
                         }
                     )