feat: add message format converter and improve tool call handling

- Add msg_converter.py for bidirectional OpenAI<->AgentScope format conversion
- Support tool_call_id in basic_tracker context serialization
- Update multiagent_tracking to use msg_converter utilities
- Update schema comments to English
- Improve workflow_metadata documentation in base_tracker

Author: TaoShuchang

ajet/context_tracker/base_tracker.py

@@ -136,7 +136,7 @@ def __init__(self, config, tokenizer, **kwargs):
         self.already_mad_flag: bool = False
         self.round_cnt = 0
         self.generation_prompt_token = None
-        self.workflow_metadata: Optional[Dict[str, Any]] = None  # 初始化 workflow_metadata 以存储工具统计信息
+        self.workflow_metadata: Optional[Dict[str, Any]] = None  # Initialize workflow_metadata to store tool statistics
 
         assert (
             self.config.ajet.data.max_prompt_length

ajet/context_tracker/basic_tracker.py

@@ -192,6 +192,8 @@ def to_role_content(self, ext_msg_array: List[ExtendedMessage]) -> List:
             }
             if ext_msg.tool_calls:
                 d.update({"tool_calls": ext_msg.tool_calls})
+            if ext_msg.tool_call_id:
+                d.update({"tool_call_id": ext_msg.tool_call_id})
             result.append(d)
         return result
 

ajet/context_tracker/multiagent_tracking.py

@@ -20,7 +20,15 @@
 from ajet.utils.color_hsl import adjust_color_hsl
 from ajet.utils.compute_madness import compute_string_madness
 from ajet.utils.tokenizer import ajet_apply_chat_template
-
+#  
+from ajet.utils.msg_converter import (
+    convert_grouped_steps_to_openai_format,
+    convert_ext_msg_to_openai_format,
+    agentscope_to_openai,
+    openai_to_agentscope,
+    agentscope_to_openai_grouped,
+    openai_to_agentscope_grouped,
+)
 @dataclass
 class TimelineMergingPolicyConfig:
     timeline_compare_level: str = "text"
@@ -101,27 +109,43 @@ def step_spawn_timeline(self, messages: List[dict], tools: List = [], disable_to
                 author = "env"
                 ignore = False
                 str_content = ""
-
-                # fix msg content
-                if msg["content"] is None:
-                    msg["content"] = ""
-                elif isinstance(msg["content"], list):
-                    for item in msg["content"]:
-                        if "text" not in item:
-                            logger.warning(
-                                f"Non-text content in message content detected: {item}. Ignoring."
-                            )
-                            ignore = True
-                            break
+                extracted_tool_call_id = ""
+                for item_idx, item in enumerate(msg["content"]):
+                    if isinstance(item, dict) and item.get("type") == "tool_result":
+                        is_tool_result_msg = True  # 标记为 tool_result 消息
+                        # Extract tool_call_id from the tool_result block
+                        if item.get("id"):
+                            extracted_tool_call_id = item.get("id", "")
+                        output = item.get("output", "")
+                        if isinstance(output, str):
+                            str_content += output
+                        elif isinstance(output, list):
+                            # output can be List[TextBlock | ImageBlock | AudioBlock]
+                            for out_item in output:
+                                if isinstance(out_item, str):
+                                    str_content += out_item
+                                elif isinstance(out_item, dict) and "text" in out_item:
+                                    str_content += str(out_item["text"])
+                        else:
+                            str_content += str(output)
+                    elif isinstance(item, dict) and "text" in item:
                         if isinstance(item["text"], str):
                             str_content += str(item["text"])
                         else:
                             str_content = ""
-                        msg["content"] = str_content
-                else:
-                    raise ValueError(
-                        f"Unsupported non-str message content type: {type(msg['content'])}, Message:\n {msg}"
-                    )
+                    else:
+                        logger.warning(
+                            f"Non-text content in message content detected: {item}. Ignoring."
+                        )
+                        ignore = True
+                        break
+                msg["content"] = str_content
+                msg["tool_call_id"] = extracted_tool_call_id  # Store extracted tool_call_id
+                
+                # ★ 关键修复：如果是 tool_result 消息，将 role 恢复为 "tool"（OpenAI 格式）
+                if is_tool_result_msg and extracted_tool_call_id:
+                    msg["role"] = "tool"
+                
 
                 if ignore:
                     continue
@@ -143,6 +167,7 @@ def step_spawn_timeline(self, messages: List[dict], tools: List = [], disable_to
                     tokenizer=self.tokenizer,
                     tools=tools,
                     tool_calls=(msg["tool_calls"] if "tool_calls" in msg else []),
+                    tool_call_id=(msg["tool_call_id"] if "tool_call_id" in msg else ""),
                     token_generator="auto",
                     first_message=(i == 0),
                 )
@@ -580,3 +605,25 @@ def check_context_token_num_safe(
         else:
             ret = (False, token_overflow, "token_overflow")
         return ret
+
+    def get_grouped_steps_openai_format(self) -> List[List[Dict[str, Any]]]:
+        """
+        将 grouped_steps 转换为 OpenAI 格式并返回。
+        
+        Returns:
+            OpenAI 格式的轨迹数据 (List of List of dict)
+            每条消息格式如：
+            - {"role": "assistant", "content": "...", "tool_calls": [...]}
+            - {"role": "tool", "content": "...", "tool_call_id": "call_xxx"}
+            - {"role": "user/system", "content": "..."}
+        """
+        return convert_grouped_steps_to_openai_format(self.grouped_steps)
+
+    def get_full_context_openai_format(self) -> List[Dict[str, Any]]:
+        """
+        将当前 full_context 转换为 OpenAI 格式并返回。
+        
+        Returns:
+            OpenAI 格式的消息列表 (List of dict)
+        """
+        return [convert_ext_msg_to_openai_format(msg) for msg in self.full_context]

ajet/schema/extended_msg.py

@@ -72,6 +72,7 @@ def __init__(
         build_from_uuid="",
         tools=[],
         tool_calls=[],
+        tool_call_id="",
         token_logprob_arr=[],
         first_message=False,
     ):
@@ -88,6 +89,7 @@ def __init__(
         self.clip = clip
         self.tools = tools
         self.tool_calls = tool_calls
+        self.tool_call_id = tool_call_id
         self.uuid = uuid.uuid4().hex
         self.build_from_uuid = build_from_uuid
         self.first_message = first_message
@@ -143,6 +145,8 @@ def auto_tokenize_non_first_message(self, tokenizer, tools):
             }
             if self.tool_calls:
                 auto_tokenize_target.update({"tool_calls": self.tool_calls})
+            if self.tool_call_id:
+                auto_tokenize_target.update({"tool_call_id": self.tool_call_id})
             text_frag_to = ajet_apply_chat_template(
                 tokenizer=tokenizer,
                 conversation=DUMMY_MSG + [auto_tokenize_target],