bump version to 0.2.0

1.  重构bumpversion模块，支持自动识别pyproject.toml和__init__.py文件的版本号格式
2.  提取版本计算、替换字符串构建逻辑，提升代码可维护性
3.  在pymake.py中新增bumpmi命令用于执行次版本号更新
4.  全面升级测试用例，适配新的版本匹配逻辑，修正测试文件类型
5.  保留原始引号和格式，不破坏文件原有排版

pyproject.toml

@@ -17,11 +17,11 @@ license = { text = "MIT" }
 name = "pyflowx"
 readme = "README.md"
 requires-python = ">=3.8"
-version = "0.1.8"
+version = "0.2.0"
 
 [project.scripts]
 autofmt     = "pyflowx.cli.autofmt:main"
-bumpver  = "pyflowx.cli.bumpversion:main"
+bumpversion = "pyflowx.cli.bumpversion:main"
 clr         = "pyflowx.cli.clearscreen:main"
 emlman      = "pyflowx.cli.emlmanager:main"
 envpy       = "pyflowx.cli.envpy:main"

src/pyflowx/__init__.py

@@ -84,7 +84,7 @@ from .runner import CliExitCode, CliRunner
 from .storage import JSONBackend, MemoryBackend, StateBackend
 from .task import TaskCmd, TaskEvent, TaskResult, TaskSpec, TaskStatus
 
-__version__ = "0.1.8"
+__version__ = "0.2.0"
 
 __all__ = [
     "IS_LINUX",

src/pyflowx/cli/bumpversion.py

@@ -5,97 +5,257 @@
 
 from __future__ import annotations
 
-import subprocess
+import argparse
+import re
+from pathlib import Path
+from typing import Literal, get_args
 
 import pyflowx as px
 
-# ============================================================================
-# 辅助函数
-# ============================================================================
+BumpVersionType = Literal["patch", "minor", "major"]
+
+# 针对不同文件类型的版本号匹配模式
+# pyproject.toml: version = "X.Y.Z" 或 version = 'X.Y.Z'
+_PYPROJECT_VERSION_PATTERN = re.compile(
+    r'(?:^|\n)\s*version\s*=\s*["\']'
+    r"(?P<major>0|[1-9]\d*)\.(?P<minor>0|[1-9]\d*)\.(?P<patch>0|[1-9]\d*)"
+    r"(?:-(?P<prerelease>(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*)(?:\.(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*))*))?"
+    r"(?:\+(?P<buildmetadata>[0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?"
+    r'["\']',
+    re.MULTILINE,
+)
+
+# __init__.py: __version__ = "X.Y.Z" 或 __version__ = 'X.Y.Z'
+_INIT_VERSION_PATTERN = re.compile(
+    r'(?:^|\n)\s*__version__\s*=\s*["\']'
+    r"(?P<major>0|[1-9]\d*)\.(?P<minor>0|[1-9]\d*)\.(?P<patch>0|[1-9]\d*)"
+    r"(?:-(?P<prerelease>(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*)(?:\.(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*))*))?"
+    r"(?:\+(?P<buildmetadata>[0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?"
+    r'["\']',
+    re.MULTILINE,
+)
 
 
-def bump_version(part: str = "patch", tag: bool = False, commit: bool = False) -> None:
-    """递增版本号.
+def _get_pattern_for_file(file_name: str) -> re.Pattern[str] | None:
+    """根据文件类型获取对应的正则表达式.
 
     Parameters
     ----------
-    part : str
+    file_name : str
+        文件名
+
+    Returns
+    -------
+    re.Pattern[str] | None
+        对应的正则表达式，如果无法确定则返回 None
+    """
+    if file_name == "pyproject.toml":
+        return _PYPROJECT_VERSION_PATTERN
+    if file_name == "__init__.py":
+        return _INIT_VERSION_PATTERN
+    return None
+
+
+def _calculate_new_version(major: int, minor: int, patch: int, part: BumpVersionType) -> str:
+    """计算新版本号.
+
+    Parameters
+    ----------
+    major : int
+        当前主版本号
+    minor : int
+        当前次版本号
+    patch : int
+        当前补丁版本号
+    part : BumpVersionType
+        要更新的部分
+
+    Returns
+    -------
+    str
+        新版本号
+    """
+    if part == "major":
+        return f"{major + 1}.0.0"
+    if part == "minor":
+        return f"{major}.{minor + 1}.0"
+    return f"{major}.{minor}.{patch + 1}"
+
+
+def _build_replacement_string(original_match: str, new_version: str, file_name: str) -> str:
+    """构建替换字符串，保留原始格式.
+
+    Parameters
+    ----------
+    original_match : str
+        原始匹配的字符串
+    new_version : str
+        新版本号
+    file_name : str
+        文件名
+
+    Returns
+    -------
+    str
+        替换字符串
+    """
+    quote_char = '"' if '"' in original_match else "'"
+
+    if file_name == "pyproject.toml":
+        prefix_match = re.match(r'(\s*version\s*=\s*)["\']', original_match)
+        prefix = prefix_match.group(1) if prefix_match else "version = "
+        return f"{prefix}{quote_char}{new_version}{quote_char}"
+
+    if file_name == "__init__.py":
+        prefix_match = re.match(r'(\s*__version__\s*=\s*)["\']', original_match)
+        prefix = prefix_match.group(1) if prefix_match else "__version__ = "
+        return f"{prefix}{quote_char}{new_version}{quote_char}"
+
+    return new_version
+
+
+def bump_file_version(file_path: Path, part: BumpVersionType = "patch") -> str | None:
+    """更新文件中的版本号.
+
+    Parameters
+    ----------
+    file_path : Path
+        要更新的文件路径
+    part : BumpVersionType
         版本部分: patch, minor, major
-    tag : bool
-        是否创建 Git 标签
-    commit : bool
-        是否提交更改
+
+    Returns
+    -------
+    str | None
+        更新后的新版本号，如果文件中未找到版本号则返回 None
     """
     try:
-        subprocess.run(["bumpversion", part], check=True)
-        if commit:
-            subprocess.run(["git", "add", "."], check=True)
-            subprocess.run(["git", "commit", "-m", f"bump version {part}"], check=True)
-        if tag:
-            # 获取当前版本号
-            result = subprocess.run(
-                ["git", "describe", "--tags", "--abbrev=0"],
-                check=True,
-                capture_output=True,
-                text=True,
-            )
-            version = result.stdout.strip() if result.returncode == 0 else f"v{part}"
-            subprocess.run(
-                ["git", "tag", "-a", version, "-m", f"version {part}"],
-                check=True,
-            )
-    except FileNotFoundError:
-        print("未找到 bumpversion 工具，请先安装: pip install bumpversion")
+        content = file_path.read_text(encoding="utf-8")
+    except Exception as e:
+        print(f"读取文件 {file_path} 时出错: {e}")
         raise
 
+    # 获取文件对应的正则表达式
+    pattern = _get_pattern_for_file(file_path.name)
+
+    # 对于未知文件类型，尝试两种模式
+    if pattern:
+        match = pattern.search(content)
+    else:
+        match = _PYPROJECT_VERSION_PATTERN.search(content) or _INIT_VERSION_PATTERN.search(content)
+
+    if not match:
+        print(f"文件 {file_path} 中未找到版本号模式")
+        return None
+
+    # 提取当前版本号
+    major = int(match.group("major"))
+    minor = int(match.group("minor"))
+    patch = int(match.group("patch"))
+
+    # 计算新版本号
+    new_version = _calculate_new_version(major, minor, patch, part)
+
+    # 构建替换字符串
+    original_match = match.group(0)
+    replacement = _build_replacement_string(original_match, new_version, file_path.name)
+
+    # 更新文件内容
+    content = content.replace(original_match, replacement)
 
-def bump_version_alpha(part: str = "patch") -> None:
-    """递增版本号并添加 alpha 预发布标识."""
     try:
-        subprocess.run(["bumpversion", part, "--new-version", f"{part}-alpha"], check=True)
-    except FileNotFoundError:
-        print("未找到 bumpversion 工具，请先安装: pip install bumpversion")
+        file_path.write_text(content, encoding="utf-8")
+    except Exception as e:
+        print(f"更新文件 {file_path} 版本号时出错: {e}")
         raise
 
-
-# ============================================================================
-# TaskSpec 定义
-# ============================================================================
-
-bump_patch: px.TaskSpec = px.TaskSpec("bump_patch", fn=lambda: bump_version("patch"))
-bump_minor: px.TaskSpec = px.TaskSpec("bump_minor", fn=lambda: bump_version("minor"))
-bump_major: px.TaskSpec = px.TaskSpec("bump_major", fn=lambda: bump_version("major"))
-bump_patch_tag: px.TaskSpec = px.TaskSpec("bump_patch_tag", fn=lambda: bump_version("patch", tag=True))
-bump_minor_tag: px.TaskSpec = px.TaskSpec("bump_minor_tag", fn=lambda: bump_version("minor", tag=True))
-bump_major_tag: px.TaskSpec = px.TaskSpec("bump_major_tag", fn=lambda: bump_version("major", tag=True))
-bump_patch_alpha: px.TaskSpec = px.TaskSpec("bump_patch_alpha", fn=lambda: bump_version_alpha("patch"))
-
-
-# ============================================================================
-# CLI Runner
-# ============================================================================
+    return new_version
 
 
 def main() -> None:
     """版本号管理工具主函数."""
-    runner = px.CliRunner(
-        strategy="thread",
-        description="BumpVersion - 版本号自动管理工具",
-        graphs={
-            # 递增补丁号 (1.0.0 -> 1.0.1)
-            "p": px.Graph.from_specs([bump_patch]),
-            # 递增次版本号 (1.0.0 -> 1.1.0)
-            "m": px.Graph.from_specs([bump_minor]),
-            # 递增主版本号 (1.0.0 -> 2.0.0)
-            "M": px.Graph.from_specs([bump_major]),
-            # 递增补丁号并创建标签
-            "pt": px.Graph.from_specs([bump_patch_tag]),
-            # 递增次版本号并创建标签
-            "mt": px.Graph.from_specs([bump_minor_tag]),
-            # 递增主版本号并创建标签
-            "Mt": px.Graph.from_specs([bump_major_tag]),
-            # 递增补丁号并添加 alpha 预发布标识
-            "pa": px.Graph.from_specs([bump_patch_alpha]),
-        },
+    parser = argparse.ArgumentParser(description="BumpVersion - 版本号自动管理工具")
+    parser.add_argument(
+        "part",
+        type=str,
+        nargs="?",
+        default="patch",
+        choices=get_args(BumpVersionType),
+        help=f"版本部分: {get_args(BumpVersionType)}",
     )
-    runner.run_cli()
+    parser.add_argument(
+        "--no-tag",
+        action="store_true",
+        help="提交后不创建 git tag",
+    )
+
+    args = parser.parse_args()
+    part = args.part
+
+    # 搜索文件，排除常见的虚拟环境和缓存目录
+    ignore_dirs = {".venv", "venv", ".git", "__pycache__", ".tox", "node_modules", "build", "dist", ".eggs"}
+    all_files = set()
+
+    for pattern in ["__init__.py", "pyproject.toml"]:
+        for file in Path.cwd().rglob(pattern):
+            # 检查路径中是否包含需要忽略的目录
+            if not any(ignore_dir in file.parts for ignore_dir in ignore_dirs):
+                all_files.add(file)
+
+    if not all_files:
+        print("未找到包含版本号的文件")
+        return
+
+    print(f"找到 {len(all_files)} 个文件需要更新版本号")
+    for file in sorted(all_files):
+        print(f"  - {file.relative_to(Path.cwd())}")
+
+    # 更新所有文件的版本号（使用顺序执行避免竞争条件）
+    # 使用相对于 cwd 的路径作为任务名，确保唯一性
+    graph = px.Graph.from_specs(
+        [
+            px.TaskSpec(
+                f"bump_{file.relative_to(Path.cwd())}".replace("\\", "_").replace("/", "_").replace(".", "_"),
+                fn=bump_file_version,
+                args=(file, part),
+            )
+            for file in all_files
+        ]
+    )
+    report = px.run(graph, strategy="sequential")
+
+    # 收集新版本号（取第一个成功的结果）
+    new_version = None
+    for task_name in report:
+        result = report[task_name]
+        if result is not None:
+            new_version = result
+            break
+
+    if not new_version:
+        print("未能获取新版本号")
+        return
+
+    print(f"版本号已更新为: {new_version}")
+
+    # 提交修改
+    graph = px.Graph.from_specs(
+        [
+            px.TaskSpec("git_add", cmd=["git", "add", "."]),
+            px.TaskSpec(
+                "git_commit", cmd=["git", "commit", "-m", f"bump version to {new_version}"], depends_on=["git_add"]
+            ),
+        ]
+    )
+    px.run(graph, strategy="sequential")
+
+    # 创建 git tag
+    if not args.no_tag:
+        tag_name = f"v{new_version}"
+        graph = px.Graph.from_specs(
+            [
+                px.TaskSpec("git_tag", cmd=["git", "tag", "-a", tag_name, "-m", f"Release {tag_name}"]),
+            ]
+        )
+        px.run(graph, strategy="sequential")
+        print(f"已创建标签: {tag_name}")

src/pyflowx/cli/gittool.py

@@ -23,6 +23,7 @@ EXCLUDE_DIRS = [
     ".tox",
     ".pytest_cache",
     "node_modules",
+    ".ruff_cache",
 ]
 EXCLUDE_CMDS = [arg for d in EXCLUDE_DIRS for arg in ["-e", d]]
 

src/pyflowx/cli/pymake.py

@@ -20,13 +20,15 @@ def maturin_build_cmd() -> list[str]:
     """
     command = ["maturin", "build", "-r"].copy()
     if Constants.IS_WINDOWS:
-        command.extend([
+        command.extend(
+            [
                 "--target",
                 "x86_64-win7-windows-msvc",
                 "-Zbuild-std",
                 "-i",
                 "python3.8",
-        ])
+            ]
+        )
     return command
 
 
@@ -47,6 +49,7 @@ test_coverage: px.TaskSpec = px.TaskSpec(
 ruff_lint: px.TaskSpec = px.TaskSpec("lint", cmd=["ruff", "check", "--fix", "--unsafe-fixes"])
 ruff_format: px.TaskSpec = px.TaskSpec("format", cmd=["ruff", "format", "."], depends_on=("lint",))
 typecheck: px.TaskSpec = px.TaskSpec("pyrefly_check", cmd=["pyrefly", "check", "."])
+git_add_all: px.TaskSpec = px.TaskSpec("git_add_all", cmd=["git", "add", "-A"])
 bump: px.TaskSpec = px.TaskSpec("bumpversion", cmd=["bumpversion", "-t"])
 doc: px.TaskSpec = px.TaskSpec("doc", cmd=["sphinx-build", "-b", "html", "docs", "docs/_build"])
 git_push: px.TaskSpec = px.TaskSpec("git_push", cmd=["git", "push"])
@@ -84,7 +87,10 @@ def main():
     📦 发布命令:
       pymake pb   - 发布到 PyPI (twine + hatch)
 
-    💡 常用工作流:
+    � 版本管理:
+      pymake bump  - 自动升级版本号并提交修改 (清理 + 检查 + 格式化 + git add + bumpversion)
+
+    �💡 常用工作流:
       1. 日常开发: pymake lint && pymake t
       2. 构建发布包: pymake ba
       3. 多版本兼容性测试: pymake tox
@@ -99,26 +105,27 @@ def main():
       pymake type        # 类型检查
     """
     runner = px.CliRunner(
-        strategy="thread",
+        strategy="sequential",
         description="PyMake - Python 构建工具",
         graphs={
             # 构建命令
             "b": px.Graph.from_specs([uv_build]),
             "bc": px.Graph.from_specs([maturin_build]),
-            "ba": px.Graph.from_specs([uv_build, maturin_build]),
+            "ba": px.Graph.from_specs(["b", "bc"]),
             # 安装命令
             "sync": px.Graph.from_specs([uv_sync]),
             # 清理命令
             "c": px.Graph.from_specs([git_clean]),
             # 开发工具
-            "bump": px.Graph.from_specs([git_clean, typecheck, ruff_lint, ruff_format, bump]),
+            "bump": px.Graph.from_specs(["c", "tc", git_add_all, bump]),
+            "bumpmi": px.Graph.from_specs([px.TaskSpec("bumpversion_minor", cmd=["bumpversion", "minor"])]),
             "cov": px.Graph.from_specs([git_clean, test_coverage]),
             "doc": px.Graph.from_specs([doc]),
             "lint": px.Graph.from_specs([ruff_lint, ruff_format]),
             "pb": px.Graph.from_specs([twine_publish, hatch_publish]),
             "t": px.Graph.from_specs([test]),
             "tf": px.Graph.from_specs([test_fast]),
-            "tc": px.Graph.from_specs([typecheck, ruff_lint, ruff_format]),
+            "tc": px.Graph.from_specs([typecheck, "lint"]),
             "tox": px.Graph.from_specs([tox]),
             # 发布命令
             "p": px.Graph.from_specs([git_clean, git_push, git_push_tags]),

src/pyflowx/graph.py

@@ -57,18 +57,59 @@ class Graph:
         return self
 
     @classmethod
-    def from_specs(cls, specs: Iterable[TaskSpec[Any]]) -> Graph:
-        """从可迭代的 task spec 构建图。
+    def from_specs(cls, specs: Iterable[TaskSpec[Any] | str]) -> Graph:
+        """从可迭代的 task spec 构建图.
 
         先收集所有 spec，再统一校验。这意味着任务可以引用*后出现*的
         依赖——顺序无关，就像声明式配置文件的读取方式。
+
+        支持字符串引用，允许引用其他命令图中的任务。
+        字符串引用将在CliRunner中解析展开。
+
+        Parameters
+        ----------
+        specs : Iterable[TaskSpec[Any] | str]
+            TaskSpec对象或字符串引用的列表
+
+        Returns
+        -------
+        Graph
+            构建完成的图
+
+        Note
+        -----
+        字符串引用格式：
+        - "command_name" - 引用整个命令图
+        - "command_name.task_name" - 引用特定任务
+
+        Examples
+        --------
+        >>> graph = Graph.from_specs([
+        ...     TaskSpec("build", cmd=["uv", "build"]),
+        ...     "test",  # 引用test命令图
+        ... ])
         """
         graph = cls()
+        pending_refs: list[str] = []
+
         for spec in specs:
+            if isinstance(spec, str):
+                # 字符串引用，稍后解析
+                pending_refs.append(spec)
+            elif isinstance(spec, TaskSpec):
                 if spec.name in graph.specs:
                     raise DuplicateTaskError(spec.name)
                 graph.specs[spec.name] = spec
                 graph.deps[spec.name] = spec.depends_on
+            else:
+                raise TypeError(f"from_specs只接受TaskSpec或str，收到: {type(spec)}")
+
+        # 存储待解析的引用
+        if pending_refs:
+            # 使用特殊属性存储引用，稍后在CliRunner中解析
+            # 由于Graph是frozen dataclass，我们需要特殊处理
+            object.__setattr__(graph, "_pending_refs", pending_refs)
+
         graph._validate_references()
         graph.validate()
         return graph

src/pyflowx/runner.py

@@ -114,6 +114,156 @@ class CliRunner:
         if not self.graphs:
             raise ValueError("CliRunner 至少需要一个命令 (通过关键字参数提供)")
 
+        # 解析并展开字符串引用
+        self._resolve_graph_refs()
+
+    def _resolve_graph_refs(self) -> None:
+        """解析并展开图中的字符串引用.
+
+        支持两种引用格式：
+        1. "command_name" - 引用整个命令图
+        2. "command_name.task_name" - 引用特定任务
+
+        递归解析所有引用，直到所有图都只包含TaskSpec对象。
+        """
+        resolved_graphs: dict[str, Graph] = {}
+
+        for cmd_name, graph in self.graphs.items():
+            resolved_graph = self._expand_refs(graph, cmd_name)
+            resolved_graphs[cmd_name] = resolved_graph
+
+        # 更新graphs字典
+        object.__setattr__(self, "graphs", resolved_graphs)
+
+    def _expand_refs(self, graph: Graph, current_cmd: str) -> Graph:
+        """展开图中的字符串引用.
+
+        Parameters
+        ----------
+        graph : Graph
+            包含可能的字符串引用的图
+        current_cmd : str
+            当前命令名（用于避免循环引用）
+
+        Returns
+        -------
+        Graph
+            展开后的图，只包含TaskSpec对象
+
+        Note
+        -----
+        引用按顺序展开，后续引用的任务依赖于前面引用的任务完成。
+        例如：["c", "tc", bump] 会展开为：
+        - c的所有任务（无依赖）
+        - tc的所有任务（依赖于c的最后一个任务）
+        - bump任务（依赖于tc的最后一个任务）
+        """
+        # 检查是否有待解析的引用
+        pending_refs = getattr(graph, "_pending_refs", None)
+        if not pending_refs:
+            return graph
+
+        # 收集所有TaskSpec（按正确顺序：先引用，后原始TaskSpec）
+        all_specs: list[TaskSpec[Any]] = []
+
+        # 记录每个引用展开后的所有任务名，用于建立依赖链
+        previous_ref_last_task: str | None = None
+
+        # 先解析每个引用，并建立依赖关系
+        for ref in pending_refs:
+            expanded_specs = self._parse_ref(ref, current_cmd)
+
+            # 如果有前面的引用，让当前引用的所有任务依赖于前面引用的最后一个任务
+            if previous_ref_last_task and expanded_specs:
+                # 为当前引用的每个任务添加依赖
+                for i, task in enumerate(expanded_specs):
+                    # 只为没有依赖的任务添加依赖，或者为第一个任务添加依赖
+                    if i == 0 or not task.depends_on:
+                        updated_task = replace(task, depends_on=tuple({*task.depends_on, previous_ref_last_task}))
+                        expanded_specs[i] = updated_task
+
+            # 记录当前引用的最后一个任务名
+            if expanded_specs:
+                previous_ref_last_task = expanded_specs[-1].name
+
+            all_specs.extend(expanded_specs)
+
+        # 然后添加原始图中的TaskSpec，并让它们按顺序执行
+        original_specs = list(graph.all_specs().values())
+        if original_specs:
+            # 第一个原始TaskSpec依赖于最后一个引用的任务
+            if previous_ref_last_task:
+                first_original = original_specs[0]
+                updated_first = replace(
+                    first_original, depends_on=tuple({*first_original.depends_on, previous_ref_last_task})
+                )
+                all_specs.append(updated_first)
+            else:
+                # 如果没有引用，直接添加第一个原始TaskSpec
+                all_specs.append(original_specs[0])
+
+            # 后续的原始TaskSpec依赖于前一个原始TaskSpec
+            for i in range(1, len(original_specs)):
+                current_task = original_specs[i]
+                previous_task_name = original_specs[i - 1].name
+                # 更新依赖，确保顺序执行
+                updated_task = replace(current_task, depends_on=tuple({*current_task.depends_on, previous_task_name}))
+                all_specs.append(updated_task)
+
+        # 创建新的图（不包含引用）
+        return Graph.from_specs(all_specs)
+
+    def _parse_ref(self, ref: str, current_cmd: str) -> list[TaskSpec[Any]]:
+        """解析单个字符串引用.
+
+        Parameters
+        ----------
+        ref : str
+            引用字符串（如"tc"或"tc.lint"）
+        current_cmd : str
+            当前命令名（用于避免循环引用）
+
+        Returns
+        -------
+        list[TaskSpec[Any]]
+            解析后的TaskSpec列表
+
+        Raises
+        ------
+        ValueError
+            如果引用无效或存在循环引用
+        """
+        # 避免循环引用
+        if ref == current_cmd:
+            raise ValueError(f"循环引用: 命令 '{current_cmd}' 引用了自己")
+
+        # 解析引用格式
+        if "." in ref:
+            # 特定任务引用: "command_name.task_name"
+            cmd_name, task_name = ref.split(".", 1)
+            if cmd_name not in self.graphs:
+                raise ValueError(f"引用的命令 '{cmd_name}' 不存在")
+
+            # 获取特定任务
+            ref_graph = self.graphs[cmd_name]
+            if task_name not in ref_graph.all_specs():
+                raise ValueError(f"任务 '{task_name}' 不存在于命令 '{cmd_name}' 中")
+
+            return [ref_graph.all_specs()[task_name]]
+        else:
+            # 整个命令图引用: "command_name"
+            cmd_name = ref
+            if cmd_name not in self.graphs:
+                raise ValueError(f"引用的命令 '{cmd_name}' 不存在")
+
+            # 获取整个图的所有任务
+            ref_graph = self.graphs[cmd_name]
+
+            # 递归展开引用（如果引用的图也有引用）
+            ref_graph = self._expand_refs(ref_graph, cmd_name)
+
+            return list(ref_graph.all_specs().values())
+
     # ------------------------------------------------------------------ #
     # 内省
     # ------------------------------------------------------------------ #

tests/cli/test_bumpversion.py

@@ -2,105 +2,261 @@
 
 from __future__ import annotations
 
-from unittest.mock import MagicMock, patch
+from pathlib import Path
 
 import pytest
 
-import pyflowx as px
 from pyflowx.cli import bumpversion
 
 
 # ---------------------------------------------------------------------- #
-# bump_version
+# bump_file_version
 # ---------------------------------------------------------------------- #
-class TestBumpVersion:
-    """Test bump_version function."""
+class TestBumpFileVersion:
+    """Test bump_file_version function."""
 
-    def test_bump_version_patch(self) -> None:
-        """Should bump patch version."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0)
-            bumpversion.bump_version("patch")
-            assert mock_run.called
+    def test_bump_patch_version(self, tmp_path: Path) -> None:
+        """Should bump patch version correctly."""
+        test_file = tmp_path / "pyproject.toml"
+        test_file.write_text('version = "1.2.3"', encoding="utf-8")
 
-    def test_bump_version_minor(self) -> None:
-        """Should bump minor version."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0)
-            bumpversion.bump_version("minor")
-            assert mock_run.called
+        result = bumpversion.bump_file_version(test_file, "patch")
 
-    def test_bump_version_major(self) -> None:
-        """Should bump major version."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0)
-            bumpversion.bump_version("major")
-            assert mock_run.called
+        assert result == "1.2.4"
+        assert test_file.read_text(encoding="utf-8") == 'version = "1.2.4"'
 
-    def test_bump_version_with_tag(self) -> None:
-        """Should bump version with tag."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0, stdout="v1.0.0")
-            bumpversion.bump_version("patch", tag=True)
-            assert mock_run.called
+    def test_bump_minor_version(self, tmp_path: Path) -> None:
+        """Should bump minor version correctly."""
+        test_file = tmp_path / "pyproject.toml"
+        test_file.write_text('version = "1.2.3"', encoding="utf-8")
 
-    def test_bump_version_with_commit(self) -> None:
-        """Should bump version with commit."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0)
-            bumpversion.bump_version("patch", commit=True)
-            assert mock_run.called
+        result = bumpversion.bump_file_version(test_file, "minor")
 
-    def test_bump_version_file_not_found(self) -> None:
-        """Should handle FileNotFoundError."""
-        with patch("subprocess.run", side_effect=FileNotFoundError), pytest.raises(FileNotFoundError):
-            bumpversion.bump_version("patch")
+        assert result == "1.3.0"
+        assert test_file.read_text(encoding="utf-8") == 'version = "1.3.0"'
+
+    def test_bump_major_version(self, tmp_path: Path) -> None:
+        """Should bump major version correctly."""
+        test_file = tmp_path / "pyproject.toml"
+        test_file.write_text('version = "1.2.3"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "major")
+
+        assert result == "2.0.0"
+        assert test_file.read_text(encoding="utf-8") == 'version = "2.0.0"'
+
+    def test_version_pattern_with_prerelease(self, tmp_path: Path) -> None:
+        """Should handle version with prerelease suffix."""
+        test_file = tmp_path / "pyproject.toml"
+        test_file.write_text('version = "1.2.3-alpha.1"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "1.2.4"
+        # 预发布版本应该被清除
+        content = test_file.read_text(encoding="utf-8")
+        assert "alpha" not in content
+
+    def test_version_pattern_with_build_metadata(self, tmp_path: Path) -> None:
+        """Should handle version with build metadata."""
+        test_file = tmp_path / "pyproject.toml"
+        test_file.write_text('version = "1.2.3+build.123"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "1.2.4"
+        # 构建元数据应该被清除
+        content = test_file.read_text(encoding="utf-8")
+        assert "build" not in content
+
+    def test_no_version_found(self, tmp_path: Path, capsys) -> None:
+        """Should return None when no version pattern found."""
+        test_file = tmp_path / "test.txt"
+        test_file.write_text("no version here", encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result is None
+        captured = capsys.readouterr()
+        assert "未找到版本号模式" in captured.out
+
+    def test_utf8_encoding(self, tmp_path: Path) -> None:
+        """Should handle UTF-8 encoded files correctly."""
+        test_file = tmp_path / "__init__.py"
+        test_file.write_text('__version__ = "1.2.3"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "1.2.4"
+        assert test_file.read_text(encoding="utf-8") == '__version__ = "1.2.4"'
+
+    def test_pyproject_toml_format(self, tmp_path: Path) -> None:
+        """Should handle pyproject.toml format correctly."""
+        test_file = tmp_path / "pyproject.toml"
+        content = """
+[project]
+name = "test"
+version = "0.1.0"
+description = "Test project"
+"""
+        test_file.write_text(content, encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "minor")
+
+        assert result == "0.2.0"
+        updated = test_file.read_text(encoding="utf-8")
+        assert 'version = "0.2.0"' in updated
+        assert 'name = "test"' in updated
+
+    def test_init_py_format(self, tmp_path: Path) -> None:
+        """Should handle __init__.py format correctly."""
+        test_file = tmp_path / "__init__.py"
+        content = '''"""Package info."""
+
+__version__ = "1.0.0"
+'''
+        test_file.write_text(content, encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "major")
+
+        assert result == "2.0.0"
+        updated = test_file.read_text(encoding="utf-8")
+        assert '__version__ = "2.0.0"' in updated
+
+    def test_multiple_versions_in_file(self, tmp_path: Path) -> None:
+        """Should only bump the project version, not dependencies."""
+        test_file = tmp_path / "pyproject.toml"
+        content = """
+[project]
+version = "1.0.0"
+dependencies = ["lib >= 2.0.0", "other >= 3.0.0"]
+"""
+        test_file.write_text(content, encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "1.0.1"
+        updated = test_file.read_text(encoding="utf-8")
+        assert 'version = "1.0.1"' in updated
+        # 确保 dependencies 中的版本没有被更新
+        assert "lib >= 2.0.0" in updated
+        assert "other >= 3.0.0" in updated
+
+    def test_file_read_error(self, tmp_path: Path, capsys) -> None:
+        """Should handle file read errors."""
+        # 创建一个目录而不是文件
+        test_file = tmp_path / "test_dir"
+        test_file.mkdir()
+
+        with pytest.raises(Exception):  # noqa: B017
+            bumpversion.bump_file_version(test_file, "patch")
+
+    def test_file_write_error(self, tmp_path: Path, capsys) -> None:
+        """Should handle file write errors."""
+        # 在只读目录中创建文件（这个测试在某些系统上可能不适用）
+        test_file = tmp_path / "readonly.toml"
+        test_file.write_text('version = "1.0.0"', encoding="utf-8")
+        # 设置为只读
+        test_file.chmod(0o444)
+
+        try:
+            with pytest.raises(Exception):  # noqa: B017
+                bumpversion.bump_file_version(test_file, "patch")
+        finally:
+            # 恢复权限以便清理
+            test_file.chmod(0o644)
 
 
 # ---------------------------------------------------------------------- #
-# bump_version_alpha
+# Version pattern tests
 # ---------------------------------------------------------------------- #
-class TestBumpVersionAlpha:
-    """Test bump_version_alpha function."""
+class TestVersionPattern:
+    """Test version pattern matching."""
 
-    def test_bump_version_alpha_patch(self) -> None:
-        """Should bump alpha patch version."""
-        with patch("subprocess.run") as mock_run:
-            mock_run.return_value = MagicMock(returncode=0)
-            bumpversion.bump_version_alpha("patch")
-            assert mock_run.called
+    def test_simple_version(self, tmp_path: Path) -> None:
+        """Should match simple version."""
+        test_file = tmp_path / "__init__.py"
+        test_file.write_text('__version__ = "1.0.0"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "1.0.1"
+
+    def test_version_with_zeros(self, tmp_path: Path) -> None:
+        """Should handle versions with zeros correctly."""
+        test_file = tmp_path / "__init__.py"
+        test_file.write_text('__version__ = "0.0.0"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        assert result == "0.0.1"
+
+    def test_large_version_numbers(self, tmp_path: Path) -> None:
+        """Should handle large version numbers."""
+        test_file = tmp_path / "__init__.py"
+        test_file.write_text('__version__ = "10.20.30"', encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "minor")
+
+        assert result == "10.21.0"
+
+    def test_version_in_url(self, tmp_path: Path) -> None:
+        """Should not match version in URL or other contexts."""
+        test_file = tmp_path / "test.txt"
+        test_file.write_text("https://example.com/v1.2.3/download", encoding="utf-8")
+
+        result = bumpversion.bump_file_version(test_file, "patch")
+
+        # 不应该匹配 URL 中的版本号
+        assert result is None
 
 
 # ---------------------------------------------------------------------- #
-# TaskSpec definitions
+# Edge cases
 # ---------------------------------------------------------------------- #
-class TestTaskSpecDefinitions:
-    """Test that all TaskSpec definitions are valid."""
+class TestEdgeCases:
+    """Test edge cases and error handling."""
 
-    def test_bump_patch_spec(self) -> None:
-        """bump_patch spec should be properly defined."""
-        assert bumpversion.bump_patch.name == "bump_patch"
-        assert bumpversion.bump_patch.fn is not None
+    def test_empty_file(self, tmp_path: Path, capsys) -> None:
+        """Should handle empty file."""
+        test_file = tmp_path / "empty.txt"
+        test_file.write_text("", encoding="utf-8")
 
-    def test_bump_minor_spec(self) -> None:
-        """bump_minor spec should be properly defined."""
-        assert bumpversion.bump_minor.name == "bump_minor"
-        assert bumpversion.bump_minor.fn is not None
+        result = bumpversion.bump_file_version(test_file, "patch")
 
-    def test_bump_major_spec(self) -> None:
-        """bump_major spec should be properly defined."""
-        assert bumpversion.bump_major.name == "bump_major"
-        assert bumpversion.bump_major.fn is not None
+        assert result is None
+        captured = capsys.readouterr()
+        assert "未找到版本号模式" in captured.out
 
+    def test_file_with_special_chars(self, tmp_path: Path) -> None:
+        """Should handle file with special characters."""
+        test_file = tmp_path / "__init__.py"
+        content = '# 中文注释\n__version__ = "1.0.0"\n# 特殊字符: @#$%'
+        test_file.write_text(content, encoding="utf-8")
 
-# ---------------------------------------------------------------------- #
-# main function
-# ---------------------------------------------------------------------- #
-class TestMain:
-    """Test main function."""
+        result = bumpversion.bump_file_version(test_file, "patch")
 
-    def test_main_calls_run_cli(self) -> None:
-        """main() should create a CliRunner and call run_cli()."""
-        with patch.object(px.CliRunner, "run_cli") as mock_run_cli:
-            bumpversion.main()
-            assert mock_run_cli.called
+        assert result == "1.0.1"
+        updated = test_file.read_text(encoding="utf-8")
+        assert "# 中文注释" in updated
+        assert "# 特殊字符: @#$%" in updated
+
+    def test_consecutive_bumps(self, tmp_path: Path) -> None:
+        """Should handle consecutive version bumps correctly."""
+        test_file = tmp_path / "__init__.py"
+        test_file.write_text('__version__ = "1.0.0"', encoding="utf-8")
+
+        # 第一次 bump
+        result1 = bumpversion.bump_file_version(test_file, "patch")
+        assert result1 == "1.0.1"
+
+        # 第二次 bump
+        result2 = bumpversion.bump_file_version(test_file, "minor")
+        assert result2 == "1.1.0"
+
+        # 第三次 bump
+        result3 = bumpversion.bump_file_version(test_file, "major")
+        assert result3 == "2.0.0"
+
+        # 验证最终结果
+        assert test_file.read_text(encoding="utf-8") == '__version__ = "2.0.0"'

tests/cli/test_packtool.py

@@ -90,16 +90,14 @@ class TestInstallEmbedPython:
         output_dir = tmp_path / "python"
 
         # Create a mock cache file that doesn't exist (force download)
-        with patch("urllib.request.urlretrieve") as mock_urlretrieve, \
-             patch("zipfile.ZipFile") as mock_zipfile:
-
+        with patch("urllib.request.urlretrieve") as mock_urlretrieve, patch("zipfile.ZipFile") as mock_zipfile:
             # Mock successful download
             mock_urlretrieve.return_value = None
             mock_zip_instance = MagicMock()
             mock_zipfile.return_value.__enter__.return_value = mock_zip_instance
 
             # Ensure cache doesn't exist by using tmp_path as cache dir
-            with patch.object(packtool, 'DEFAULT_CACHE_DIR', str(tmp_path / ".cache")):
+            with patch.object(packtool, "DEFAULT_CACHE_DIR", str(tmp_path / ".cache")):
                 packtool.install_embed_python("3.10", output_dir)
 
                 # Verify download was called

tests/test_command_refs.py

@@ -0,0 +1,499 @@
+"""Tests for command reference feature in CliRunner."""
+
+from __future__ import annotations
+
+import pytest
+
+import pyflowx as px
+
+
+class TestCommandReferences:
+    """Test string references in Graph.from_specs."""
+
+    def test_simple_command_reference(self) -> None:
+        """Should expand simple command reference."""
+        build_task = px.TaskSpec("build", cmd=["echo", "building"])
+        test_task = px.TaskSpec("test", cmd=["echo", "testing"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "build": px.Graph.from_specs([build_task]),
+                "test": px.Graph.from_specs([test_task]),
+                "all": px.Graph.from_specs([build_task, "test"]),
+            },
+        )
+
+        # Check that 'all' command has both tasks
+        all_tasks = list(runner.graphs["all"].all_specs().keys())
+        assert "build" in all_tasks
+        assert "test" in all_tasks
+        assert len(all_tasks) == 2
+
+    def test_multiple_command_references(self) -> None:
+        """Should expand multiple command references."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs([task2]),
+                "cmd3": px.Graph.from_specs([task3]),
+                "all": px.Graph.from_specs(["cmd1", "cmd2", "cmd3"]),
+            },
+        )
+
+        # Check that 'all' command has all tasks
+        all_tasks = list(runner.graphs["all"].all_specs().keys())
+        assert set(all_tasks) == {"task1", "task2", "task3"}
+
+    def test_specific_task_reference(self) -> None:
+        """Should expand specific task reference."""
+        lint_task = px.TaskSpec("lint", cmd=["echo", "linting"])
+        format_task = px.TaskSpec("format", cmd=["echo", "formatting"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "lint": px.Graph.from_specs([lint_task, format_task]),
+                "quick": px.Graph.from_specs(["lint.lint"]),
+            },
+        )
+
+        # Check that 'quick' command only has lint task
+        quick_tasks = list(runner.graphs["quick"].all_specs().keys())
+        assert quick_tasks == ["lint"]
+
+    def test_nested_command_reference(self) -> None:
+        """Should expand nested command references."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs(["cmd1", task2]),
+                "cmd3": px.Graph.from_specs(["cmd2", task3]),
+            },
+        )
+
+        # Check that 'cmd3' has all tasks
+        cmd3_tasks = list(runner.graphs["cmd3"].all_specs().keys())
+        assert set(cmd3_tasks) == {"task1", "task2", "task3"}
+
+    def test_circular_reference_error(self) -> None:
+        """Should raise error for circular references."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+
+        with pytest.raises(ValueError, match="循环引用"):
+            px.CliRunner(
+                strategy="sequential",
+                graphs={
+                    "cmd1": px.Graph.from_specs(["cmd1", task1]),
+                },
+            )
+
+    def test_invalid_command_reference_error(self) -> None:
+        """Should raise error for invalid command reference."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+
+        with pytest.raises(ValueError, match="引用的命令 'invalid' 不存在"):
+            px.CliRunner(
+                strategy="sequential",
+                graphs={
+                    "cmd1": px.Graph.from_specs(["invalid", task1]),
+                },
+            )
+
+    def test_invalid_task_reference_error(self) -> None:
+        """Should raise error for invalid task reference."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+
+        with pytest.raises(ValueError, match="任务 'invalid' 不存在于命令 'cmd1' 中"):
+            px.CliRunner(
+                strategy="sequential",
+                graphs={
+                    "cmd1": px.Graph.from_specs([task1]),
+                    "cmd2": px.Graph.from_specs(["cmd1.invalid"]),
+                },
+            )
+
+    def test_reference_preserves_dependencies(self) -> None:
+        """Should preserve dependencies when expanding references."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"], depends_on=("task1",))
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1, task2]),
+                "cmd2": px.Graph.from_specs(["cmd1"]),
+            },
+        )
+
+        # Check that dependencies are preserved
+        cmd2_deps = runner.graphs["cmd2"].deps
+        assert cmd2_deps["task2"] == ("task1",)
+
+    def test_mixed_references_and_tasks(self) -> None:
+        """Should handle mixed references and direct tasks."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1, task2]),
+                "cmd2": px.Graph.from_specs(["cmd1", task3]),
+            },
+        )
+
+        # Check that 'cmd2' has all tasks
+        cmd2_tasks = list(runner.graphs["cmd2"].all_specs().keys())
+        assert set(cmd2_tasks) == {"task1", "task2", "task3"}
+
+    def test_execution_order_with_references(self) -> None:
+        """Should execute references in correct order."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "step1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "step2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "step3"])
+        task4 = px.TaskSpec("task4", cmd=["echo", "step4"])
+        task5 = px.TaskSpec("task5", cmd=["echo", "step5"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs([task2, task3]),
+                "cmd3": px.Graph.from_specs([task4]),
+                "ordered": px.Graph.from_specs(["cmd1", "cmd2", "cmd3", task5]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["ordered"].layers()
+
+        # Layer 1 should have task1 (cmd1)
+        assert "task1" in layers[0]
+
+        # Layer 2 should have task2 and task3 (cmd2)
+        assert "task2" in layers[1]
+        assert "task3" in layers[1]
+
+        # Layer 3 should have task4 (cmd3)
+        assert "task4" in layers[2]
+
+        # Layer 4 should have task5 (original task)
+        assert "task5" in layers[3]
+
+        # Verify total layers
+        assert len(layers) == 4
+
+    def test_execution_order_multiple_original_tasks(self) -> None:
+        """Should execute multiple original TaskSpecs in correct order."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+        task4 = px.TaskSpec("task4", cmd=["echo", "4"])
+        task5 = px.TaskSpec("task5", cmd=["echo", "5"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs([task2]),
+                "all": px.Graph.from_specs(["cmd1", "cmd2", task3, task4, task5]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # Layer 1: task1 (cmd1)
+        assert "task1" in layers[0]
+
+        # Layer 2: task2 (cmd2)
+        assert "task2" in layers[1]
+
+        # Layer 3: task3 (first original TaskSpec)
+        assert "task3" in layers[2]
+
+        # Layer 4: task4 (second original TaskSpec)
+        assert "task4" in layers[3]
+
+        # Layer 5: task5 (third original TaskSpec)
+        assert "task5" in layers[4]
+
+        # Verify total layers
+        assert len(layers) == 5
+
+    def test_execution_order_with_internal_dependencies(self) -> None:
+        """Should preserve internal dependencies within referenced commands."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"], depends_on=("task1",))
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+        task4 = px.TaskSpec("task4", cmd=["echo", "4"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1, task2]),
+                "cmd2": px.Graph.from_specs([task3]),
+                "all": px.Graph.from_specs(["cmd1", "cmd2", task4]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # Layer 1: task1
+        assert "task1" in layers[0]
+
+        # Layer 2: task2 (depends on task1)
+        assert "task2" in layers[1]
+
+        # Layer 3: task3 (cmd2, depends on task2)
+        assert "task3" in layers[2]
+
+        # Layer 4: task4 (original TaskSpec, depends on task3)
+        assert "task4" in layers[3]
+
+        # Verify total layers
+        assert len(layers) == 4
+
+    def test_execution_order_pymake_bump_scenario(self) -> None:
+        """Should execute pymake bump command in correct order."""
+        # Simulate pymake bump scenario
+        git_clean = px.TaskSpec("git_clean", cmd=["echo", "clean"])
+        typecheck = px.TaskSpec("typecheck", cmd=["echo", "typecheck"])
+        lint = px.TaskSpec("lint", cmd=["echo", "lint"])
+        format_task = px.TaskSpec("format", cmd=["echo", "format"], depends_on=("lint",))
+        git_add_all = px.TaskSpec("git_add_all", cmd=["echo", "git add -A"])
+        bump = px.TaskSpec("bumpversion", cmd=["echo", "bumpversion -t"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "c": px.Graph.from_specs([git_clean]),
+                "tc": px.Graph.from_specs([typecheck, "lint"]),
+                "lint": px.Graph.from_specs([lint, format_task]),
+                "bump": px.Graph.from_specs(["c", "tc", git_add_all, bump]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["bump"].layers()
+
+        # Layer 1: git_clean (c)
+        assert "git_clean" in layers[0]
+
+        # Layer 2: lint (tc.lint, depends on git_clean)
+        assert "lint" in layers[1]
+
+        # Layer 3: format (tc.lint.format, depends on lint)
+        assert "format" in layers[2]
+
+        # Layer 4: typecheck (tc.typecheck, depends on format)
+        assert "typecheck" in layers[3]
+
+        # Layer 5: git_add_all (original TaskSpec, depends on typecheck)
+        assert "git_add_all" in layers[4]
+
+        # Layer 6: bumpversion (original TaskSpec, depends on git_add_all)
+        assert "bumpversion" in layers[5]
+
+        # Verify total layers
+        assert len(layers) == 6
+
+    def test_execution_order_only_references(self) -> None:
+        """Should execute only references without original TaskSpecs."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs([task2]),
+                "cmd3": px.Graph.from_specs([task3]),
+                "all": px.Graph.from_specs(["cmd1", "cmd2", "cmd3"]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # Layer 1: task1 (cmd1)
+        assert "task1" in layers[0]
+
+        # Layer 2: task2 (cmd2, depends on task1)
+        assert "task2" in layers[1]
+
+        # Layer 3: task3 (cmd3, depends on task2)
+        assert "task3" in layers[2]
+
+        # Verify total layers
+        assert len(layers) == 3
+
+    def test_execution_order_only_original_tasks(self) -> None:
+        """Should execute only original TaskSpecs without references."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "all": px.Graph.from_specs([task1, task2, task3]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # All tasks should be in layer 1 (no dependencies)
+        assert "task1" in layers[0]
+        assert "task2" in layers[0]
+        assert "task3" in layers[0]
+
+        # Verify total layers
+        assert len(layers) == 1
+
+    def test_execution_order_single_reference(self) -> None:
+        """Should execute single reference correctly."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1, task2]),
+                "all": px.Graph.from_specs(["cmd1"]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # Should have the same structure as cmd1
+        assert "task1" in layers[0]
+        assert "task2" in layers[0]
+
+        # Verify total layers
+        assert len(layers) == 1
+
+    def test_execution_order_deep_nesting(self) -> None:
+        """Should execute deeply nested references correctly."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+        task4 = px.TaskSpec("task4", cmd=["echo", "4"])
+        task5 = px.TaskSpec("task5", cmd=["echo", "5"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1]),
+                "cmd2": px.Graph.from_specs(["cmd1", task2]),
+                "cmd3": px.Graph.from_specs(["cmd2", task3]),
+                "cmd4": px.Graph.from_specs(["cmd3", task4]),
+                "cmd5": px.Graph.from_specs(["cmd4", task5]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["cmd5"].layers()
+
+        # Should execute in order: task1 -> task2 -> task3 -> task4 -> task5
+        assert "task1" in layers[0]
+        assert "task2" in layers[1]
+        assert "task3" in layers[2]
+        assert "task4" in layers[3]
+        assert "task5" in layers[4]
+
+        # Verify total layers
+        assert len(layers) == 5
+
+    def test_execution_order_with_parallel_tasks_in_reference(self) -> None:
+        """Should handle parallel tasks within referenced commands."""
+        task1 = px.TaskSpec("task1", cmd=["echo", "1"])
+        task2 = px.TaskSpec("task2", cmd=["echo", "2"])
+        task3 = px.TaskSpec("task3", cmd=["echo", "3"])
+        task4 = px.TaskSpec("task4", cmd=["echo", "4"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "cmd1": px.Graph.from_specs([task1, task2]),  # Parallel tasks
+                "cmd2": px.Graph.from_specs([task3, task4]),  # Parallel tasks
+                "all": px.Graph.from_specs(["cmd1", "cmd2"]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["all"].layers()
+
+        # Layer 1: task1 and task2 (cmd1, parallel)
+        assert "task1" in layers[0]
+        assert "task2" in layers[0]
+
+        # Layer 2: task3 and task4 (cmd2, depends on cmd1's last task)
+        # Note: Both task3 and task4 should depend on the last task of cmd1
+        assert "task3" in layers[1]
+        assert "task4" in layers[1]
+
+        # Verify total layers
+        assert len(layers) == 2
+
+    def test_execution_order_complex_mixed_scenario(self) -> None:
+        """Should handle complex mixed scenario with references and TaskSpecs."""
+        # Create a complex scenario
+        clean = px.TaskSpec("clean", cmd=["echo", "clean"])
+        build1 = px.TaskSpec("build1", cmd=["echo", "build1"])
+        build2 = px.TaskSpec("build2", cmd=["echo", "build2"], depends_on=("build1",))
+        test1 = px.TaskSpec("test1", cmd=["echo", "test1"])
+        test2 = px.TaskSpec("test2", cmd=["echo", "test2"])
+        package = px.TaskSpec("package", cmd=["echo", "package"])
+        deploy = px.TaskSpec("deploy", cmd=["echo", "deploy"])
+
+        runner = px.CliRunner(
+            strategy="sequential",
+            graphs={
+                "clean": px.Graph.from_specs([clean]),
+                "build": px.Graph.from_specs([build1, build2]),
+                "test": px.Graph.from_specs([test1, test2]),
+                "release": px.Graph.from_specs(["clean", "build", "test", package, deploy]),
+            },
+        )
+
+        # Check execution order through layers
+        layers = runner.graphs["release"].layers()
+
+        # Layer 1: clean
+        assert "clean" in layers[0]
+
+        # Layer 2: build1 (depends on clean)
+        assert "build1" in layers[1]
+
+        # Layer 3: build2 (depends on build1)
+        assert "build2" in layers[2]
+
+        # Layer 4: test1 and test2 (depends on build2)
+        assert "test1" in layers[3]
+        assert "test2" in layers[3]
+
+        # Layer 5: package (depends on test1/test2)
+        assert "package" in layers[4]
+
+        # Layer 6: deploy (depends on package)
+        assert "deploy" in layers[5]
+
+        # Verify total layers
+        assert len(layers) == 6

uv.lock

@@ -2184,7 +2184,7 @@ wheels = [
 
 [[package]]
 name = "pyflowx"
-version = "0.1.8"
+version = "0.1.13"
 source = { editable = "." }
 dependencies = [
     { name = "graphlib-backport", marker = "python_full_version < '3.9'" },