【GitHub探索】代码开发AI辅助工具trae-agent

作为国内表现上比较agentic的开发工具之一，trae项目开源了自家trae-agent开发工具，整体也有一段时间了。借着日常闲暇的机会，笔者计划研究下trae-agent的代码实现，看下这个本地AI开发辅助工具是如何运作的。

从开源项目的角度，trae-agent给的文档信息是比较充足的，但代码实现上只能说是MVP版本，只有一些基础功能。当然这个也可以理解，相信实际生产用的trae-agent会远比这个实现更加复杂，以及不可能所有人力都投入到开源项目的建设中。所以今天就简单看看这个项目的主流程。

主流程的话在agent.py当中，可以参考最深处BaseAgent中execute_task和_run_llm_step的实现，此处展示下基础流程部分：

class BaseAgent(ABC):async def execute_task(self) -> AgentExecution:try:messages = self._initial_messagesstep_number = 1execution.agent_state = AgentState.RUNNINGwhile step_number <= self._max_steps:step = AgentStep(step_number=step_number, state=AgentStepState.THINKING)try: # 调用_run_llm_step和llm做chatmessages = await self._run_llm_step(step, messages, execution)await self._finalize_step( # 记录tracestep, messages, execution)  # record trajectory for this step and update the CLI consoleif execution.agent_state == AgentState.COMPLETED:breakstep_number += 1except Exception as error:execution.agent_state = AgentState.ERRORstep.state = AgentStepState.ERRORstep.error = str(error)await self._finalize_step(step, messages, execution)breakif step_number > self._max_steps and not execution.success:execution.final_result = "Task execution exceeded maximum steps without completion."execution.agent_state = AgentState.ERRORreturn executionasync def _run_llm_step(self, step: "AgentStep", messages: list["LLMMessage"], execution: "AgentExecution") -> list["LLMMessage"]:# Display thinking statestep.state = AgentStepState.THINKINGself._update_cli_console(step, execution)# Get LLM responsellm_response = self._llm_client.chat(messages, self._model_config, self._tools)step.llm_response = llm_response# Display step with LLM responseself._update_cli_console(step, execution)# Update token usageself._update_llm_usage(llm_response, execution)if self.llm_indicates_task_completed(llm_response):if self._is_task_completed(llm_response):execution.agent_state = AgentState.COMPLETEDexecution.final_result = llm_response.contentexecution.success = Truereturn messageselse:execution.agent_state = AgentState.RUNNINGreturn [LLMMessage(role="user", content=self.task_incomplete_message())]else:tool_calls = llm_response.tool_callsreturn await self._tool_call_handler(tool_calls, step)# 支默认持的tools
tools_registry: dict[str, type[Tool]] = {"bash": BashTool,"str_replace_based_edit_tool": TextEditorTool,"json_edit_tool": JSONEditTool,"sequentialthinking": SequentialThinkingTool,"task_done": TaskDoneTool,"ckg": CKGTool,
}

在设置里头我们可以设置每个问题的最大step次数，execute_task中就通过loop去要求LLM不断迭代问题解决方案。_run_llm_step中则传递历史消息和tools给到LLM，_llm_client支持了多个provider，给回结果后，通过匹配硬编码的关键字判断是否完成task，否则继续迭代，当然如果识别出需要tool-call的话，就由代码逻辑主动调用对应tool。

针对messages，在execute_task的caller的new_task过程中，已经注入了硬编码的system-prompt，以及带模板的user-prompt，可以在agent_prompt.py中看到。

TRAE_AGENT_SYSTEM_PROMPT = """You are an expert AI software engineering agent.File Path Rule: All tools that take a `file_path` as an argument require an **absolute path**. You MUST construct the full, absolute path by combining the `[Project root path]` provided in the user's message with the file's path inside the project.For example, if the project root is `/home/user/my_project` and you need to edit `src/main.py`, the correct `file_path` argument is `/home/user/my_project/src/main.py`. Do NOT use relative paths like `src/main.py`.Your primary goal is to resolve a given GitHub issue by navigating the provided codebase, identifying the root cause of the bug, implementing a robust fix, and ensuring your changes are safe and well-tested.Follow these steps methodically:"""
# 下面忽略class TraeAgent(BaseAgent):@overridedef new_task(self,task: str,extra_args: dict[str, str] | None = None,tool_names: list[str] | None = None,):# fixed system promptself._initial_messages: list[LLMMessage] = []self._initial_messages.append(LLMMessage(role="system", content=self.get_system_prompt()))# user message with issue to solveuser_message = ""if not extra_args:raise AgentError("Project path and issue information are required.")if "project_path" not in extra_args:raise AgentError("Project path is required")self.project_path = extra_args.get("project_path", "")if self.docker_config:user_message += r"[Project root path]:\workspace\n\n"else:user_message += f"[Project root path]:\n{self.project_path}\n\n"if "issue" in extra_args:user_message += f"[Problem statement]: We're currently solving the following issue within our repository. Here's the issue text:\n{extra_args['issue']}\n"optional_attrs_to_set = ["base_commit", "must_patch", "patch_path"]for attr in optional_attrs_to_set:if attr in extra_args:setattr(self, attr, extra_args[attr])self._initial_messages.append(LLMMessage(role="user", content=user_message))

由于每次message只传递当次问答的message，所以针对单个问题，每个LLMClient需要实现对历史问答的记录。比如OllamaClient的实现：

class OllamaClient(BaseLLMClient):@overridedef chat(self,messages: list[LLMMessage],model_config: ModelConfig,tools: list[Tool] | None = None,reuse_history: bool = True,) -> LLMResponse:msgs: ResponseInputParam = self.parse_messages(messages)if reuse_history:self.message_history = self.message_history + msgselse:self.message_history = msgs

主流程大概是这些，代码架构上来看实现也不算太复杂，用专有Agent框架复刻也是可以的，然后硬编码注入的内容比较多，所以整体上来看trae-agent开源版离实际投产的标准还有一些距离。如果自己的项目要复用这套框架开发内容，最好保证以下事情，一是有一个足够可靠确定性高的模型，二是需要单独一套prompt管理模块把硬编码注入的prompt等文本做统一管理，三是有一套执行稳定性跟评测机制让整个agent过程能够针对自己项目的问题给出解决方案，至少得有面子上过得去的output才行。