fix: group bedrock tool results

Co-authored-by: João Moura <joaomdmoura@gmail.com>

lib/crewai/src/crewai/llms/providers/bedrock/completion.py

@@ -1781,6 +1781,7 @@ class BedrockCompletion(BaseLLM):
 
         converse_messages: list[LLMMessage] = []
         system_message: str | None = None
+        pending_tool_results: list[dict[str, Any]] = []
 
         for message in formatted_messages:
             role = message.get("role")
@@ -1794,53 +1795,56 @@ class BedrockCompletion(BaseLLM):
                     system_message += f"\n\n{content}"
                 else:
                     system_message = cast(str, content)
-            elif role == "assistant" and tool_calls:
-                # Convert OpenAI-style tool_calls to Bedrock toolUse format
-                bedrock_content = []
-                for tc in tool_calls:
-                    func = tc.get("function", {})
-                    tool_use_block = {
-                        "toolUse": {
-                            "toolUseId": tc.get("id", f"call_{id(tc)}"),
-                            "name": func.get("name", ""),
-                            "input": func.get("arguments", {})
-                            if isinstance(func.get("arguments"), dict)
-                            else json.loads(func.get("arguments", "{}") or "{}"),
-                        }
-                    }
-                    bedrock_content.append(tool_use_block)
-                converse_messages.append(
-                    {"role": "assistant", "content": bedrock_content}
-                )
             elif role == "tool":
                 if not tool_call_id:
                     raise ValueError("Tool message missing required tool_call_id")
-                converse_messages.append(
+                pending_tool_results.append(
                     {
-                        "role": "user",
-                        "content": [
-                            {
-                                "toolResult": {
-                                    "toolUseId": tool_call_id,
-                                    "content": [
-                                        {"text": str(content) if content else ""}
-                                    ],
-                                }
-                            }
-                        ],
+                        "toolResult": {
+                            "toolUseId": tool_call_id,
+                            "content": [{"text": str(content) if content else ""}],
+                        }
                     }
                 )
             else:
-                # Convert to Converse API format with proper content structure
-                if isinstance(content, list):
-                    # Already formatted as multimodal content blocks
-                    converse_messages.append({"role": role, "content": content})
-                else:
-                    # String content - wrap in text block
-                    text_content = content if content else ""
+                if pending_tool_results:
                     converse_messages.append(
-                        {"role": role, "content": [{"text": text_content}]}
+                        {"role": "user", "content": pending_tool_results}
                     )
+                    pending_tool_results = []
+
+                if role == "assistant" and tool_calls:
+                    # Convert OpenAI-style tool_calls to Bedrock toolUse format
+                    bedrock_content = []
+                    for tc in tool_calls:
+                        func = tc.get("function", {})
+                        tool_use_block = {
+                            "toolUse": {
+                                "toolUseId": tc.get("id", f"call_{id(tc)}"),
+                                "name": func.get("name", ""),
+                                "input": func.get("arguments", {})
+                                if isinstance(func.get("arguments"), dict)
+                                else json.loads(func.get("arguments", "{}") or "{}"),
+                            }
+                        }
+                        bedrock_content.append(tool_use_block)
+                    converse_messages.append(
+                        {"role": "assistant", "content": bedrock_content}
+                    )
+                else:
+                    # Convert to Converse API format with proper content structure
+                    if isinstance(content, list):
+                        # Already formatted as multimodal content blocks
+                        converse_messages.append({"role": role, "content": content})
+                    else:
+                        # String content - wrap in text block
+                        text_content = content if content else ""
+                        converse_messages.append(
+                            {"role": role, "content": [{"text": text_content}]}
+                        )
+
+        if pending_tool_results:
+            converse_messages.append({"role": "user", "content": pending_tool_results})
 
         # CRITICAL: Handle model-specific conversation requirements
         # Cohere and some other models require conversation to end with user message

lib/crewai/tests/llms/bedrock/test_bedrock.py

@@ -967,3 +967,70 @@ def test_bedrock_agent_kickoff_structured_output_with_tools():
     assert result.pydantic.result == 42, f"Expected result 42 but got {result.pydantic.result}"
     assert result.pydantic.operation, "Operation should not be empty"
     assert result.pydantic.explanation, "Explanation should not be empty"
+
+
+def test_bedrock_groups_three_tool_results():
+    """Consecutive tool results should be grouped into one Bedrock user message."""
+    llm = LLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0")
+
+    test_messages = [
+        {"role": "user", "content": "Use all three tools, then continue."},
+        {
+            "role": "assistant",
+            "content": "",
+            "tool_calls": [
+                {
+                    "id": "tool-1",
+                    "type": "function",
+                    "function": {
+                        "name": "lookup_weather",
+                        "arguments": '{"location": "New York"}',
+                    },
+                },
+                {
+                    "id": "tool-2",
+                    "type": "function",
+                    "function": {
+                        "name": "lookup_news",
+                        "arguments": '{"topic": "AI"}',
+                    },
+                },
+                {
+                    "id": "tool-3",
+                    "type": "function",
+                    "function": {
+                        "name": "lookup_stock",
+                        "arguments": '{"ticker": "AMZN"}',
+                    },
+                },
+            ],
+        },
+        {"role": "tool", "tool_call_id": "tool-1", "content": "72F and sunny"},
+        {"role": "tool", "tool_call_id": "tool-2", "content": "AI news summary"},
+        {"role": "tool", "tool_call_id": "tool-3", "content": "AMZN up 1.2%"},
+    ]
+
+    formatted_messages, system_message = llm._format_messages_for_converse(
+        test_messages
+    )
+
+    assert system_message is None
+    assert [message["role"] for message in formatted_messages] == [
+        "user",
+        "assistant",
+        "user",
+    ]
+    assert len(formatted_messages[1]["content"]) == 3
+
+    tool_results = formatted_messages[2]["content"]
+    assert len(tool_results) == 3
+    assert [block["toolResult"]["toolUseId"] for block in tool_results] == [
+        "tool-1",
+        "tool-2",
+        "tool-3",
+    ]
+    assert [block["toolResult"]["content"][0]["text"] for block in tool_results] == [
+        "72F and sunny",
+        "AI news summary",
+        "AMZN up 1.2%",
+    ]