codebee 0.1.62 → 0.1.64

package/app/core/store.py

@@ -18,7 +18,7 @@ import threading
 import time
 from pathlib import Path
 
-from . import paths, runner, volumes
+from . import artifacts, paths, runner, volumes
 
 LOCK = threading.RLock()
 _TASKS = {}
@@ -746,7 +746,7 @@ def task_side(task_id):
     total_steps = sum(len(r.get("steps") or []) for r in runs)
     # 成品口径闸：任务一步都没跑出来过（如历次都在检出前失败）→ 工作目录里的
     # 文件变动全是并行活动的噪音，不算这个任务的成品
-    wd, arts = run_artifacts(latest["id"], limit=50) if (latest and total_steps > 0) else ("", [])
+    wd, arts = run_artifacts(latest["id"], limit=50, cache=True) if (latest and total_steps > 0) else ("", [])
     return {
         "task": {k: task.get(k) for k in ("id", "title", "status", "workdir", "git_state",
                                           "git_rev")},
@@ -966,41 +966,6 @@ def run_workdir(run_id):
     return str(task.get("workdir") or "") if task else ""
 
 
-_SKIP_DIRS = {".git", "node_modules", "__pycache__", ".venv", "venv", ".idea", ".vscode"}
-
-# 构建产物/依赖缓存目录：里面的文件是工具链再生成的，不是任务成品
-_BUILD_DIRS = {
-    "target", "build", "dist", "out", "bin", "obj",           # 通用构建输出
-    "surefire", "failsafe-reports", "test-output", "reports",  # 测试/报告输出
-    ".next", ".nuxt", ".output", ".gradle", ".gradle-home",    # 前端/Gradle
-    "__MACOSX",
-}
-
-# CLI 自身的历史/评审中间件不属于用户成果。它们通常落在工作目录根部，
-# 只按隐藏目录过滤会漏掉 Aider 的 .aider.* 与流程生成的 *_review.json。
-_PROCESS_ARTIFACT_NAMES = {
-    ".aider.chat.history.md", ".aider.input.history", ".aider.input.history.md",
-    ".aider.tags.cache.v4", ".aider.tags.cache.v3", ".aider.conf.yml",
-}
-_PROCESS_ARTIFACT_SUFFIXES = ("_review.json", ".review.json")
-_PROCESS_ARTIFACT_PREFIXES = ("tutti_prompt_", "_tutti_prompt_")
-
-
-def _is_process_artifact(rel_name):
-    """判断工作目录文件是否为 CLI/评审过程产物，而非可交付成果。"""
-    name = str(rel_name or "").replace("\\", "/")
-    base = name.rsplit("/", 1)[-1].lower()
-    if base in _PROCESS_ARTIFACT_NAMES:
-        return True
-    if base.startswith(".") and base.startswith(".aider"):
-        return True
-    if base.endswith(_PROCESS_ARTIFACT_SUFFIXES):
-        return True
-    if base.startswith(_PROCESS_ARTIFACT_PREFIXES):
-        return True
-    return False
-
-
 def task_first_start(task_id, fallback=""):
     """该任务最早一次运行的开始时间（含回退：任务创建时间 → 指定回退值）。"""
     stamps = []
@@ -1019,13 +984,15 @@ def task_first_start(task_id, fallback=""):
     return best
 
 
-def run_artifacts(run_id, limit=200):
+def run_artifacts(run_id, limit=200, cache=False):
     """列一次运行的「成品文件」：工作目录里自该任务首次运行以来新产生/修改的文件。
 
     断点续跑会拆成多条 run，只按本 run 过滤会漏掉早期章节；这里以「任务首跑」
     为起点。返回 (workdir, files)；files 按 mtime 新→旧，name 为工作目录内
-    相对路径，已跳过 .git / 隐藏目录 / node_modules / target 等构建产物目录，
-    最多 limit 个。
+    相对路径，已跳过 .git / 隐藏目录 / node_modules / target 等构建产物目录。
+    cache=True 供 UI 轮询热路径（详情/侧栏/状态快照）复用短 TTL 缓存与在飞
+    合并；发布/预览/知识库等正确性敏感方保持默认，永远现扫。
+    剪枝遍历与过滤实现在 artifacts.py（几万文件目录从秒级降到百毫秒级）。
     """
     wd = run_workdir(run_id)
     if not wd:
@@ -1040,36 +1007,12 @@ def run_artifacts(run_id, limit=200):
     root = Path(wd)
     if not root.is_dir():
         return wd, []
-    files = []
-    try:
-        for p in root.rglob("*"):
-            if not p.is_file():
-                continue
-            if any(part in _SKIP_DIRS for part in p.parts):
-                continue
-            if any(part in _BUILD_DIRS for part in p.parts[:-1]):
-                continue
-            # 隐藏目录一律是工具过程文件（.mimocode/.zcode/.claude/.codex…），
-            # 不是成品；按前缀通排，免得每来一个新 agent CLI 就补一次白名单
-            if any(part.startswith(".") for part in p.parts[:-1]):
-                continue
-            try:
-                st = p.stat()
-            except OSError:
-                continue
-            if st.st_mtime < t0:
-                continue
-            rel_name = str(p.relative_to(root)).replace("\\", "/")
-            if _is_process_artifact(rel_name):
-                continue
-            files.append({"name": rel_name,
-                          "size": st.st_size, "mtime": int(st.st_mtime)})
-            if len(files) >= 800:  # 防超大目录拖垮接口；截断后再排序取最新
-                break
-    except OSError:
-        pass
-    files.sort(key=lambda f: -f["mtime"])
-    return wd, files[:limit]
+    running = False
+    if cache:
+        task = get_task(run.get("task_id") or "") if run.get("task_id") else None
+        running = bool(task) and task.get("status") in ("queued", "running")
+    return wd, artifacts.scan(str(root), t0, limit=limit,
+                              running=running, cache=cache)
 
 
 def task_step_count(task_id):

package/app/core/task_compile.py

@@ -1,199 +1,199 @@
-# -*- coding: utf-8 -*-
-"""任务编译器：把用户任务和预置流程编译成统一、可审计的运行规格。"""
-from __future__ import annotations
-
-from collections.abc import Mapping
-
-from . import dispatch, flows
-
-SCHEMA_VERSION = 1
-DIFFICULTIES = ("easy", "default", "hard")
-
-
-def _difficulty(task, flow):
-    explicit = str((task or {}).get("difficulty") or "").strip().lower()
-    if explicit in DIFFICULTIES:
-        return explicit, "用户指定"
-    try:
-        threshold = float((task or {}).get("threshold")
-                          if (task or {}).get("threshold") is not None
-                          else (flow or {}).get("threshold") or 7.0)
-    except (TypeError, ValueError):
-        threshold = 7.0
-    if threshold >= 8.5:
-        return "hard", "高质量门槛"
-    if threshold <= 6.0:
-        return "easy", "快速交付门槛"
-    if str((task or {}).get("type") or "").lower() in (
-            "research", "tech_proposal", "rank_scan"):
-        return "hard", "研究/方案类默认需要深度"
-    return "default", "标准难度"
-
-
-def _capabilities(task, dimension, engine):
-    caps = [dimension]
-    if engine == "code" or (task or {}).get("verify_command"):
-        caps.extend(["filesystem", "verification"])
-    if (task or {}).get("serial"):
-        caps.extend(["long_context", "continuity"])
-    if (task or {}).get("attachments"):
-        caps.append("attachments")
-    return list(dict.fromkeys(caps))
-
-
-def compile_task(task):
-    """返回统一任务规格；输入缺失或字段异常时保持可编排的安全兜底。"""
-    raw = dict(task) if isinstance(task, Mapping) else {}
-    ttype = str(raw.get("type") or "direct").strip().lower()
-    flow = flows.get_flow(ttype) or {}
-    engine = str(raw.get("engine") or flow.get("engine") or
-                 ("code" if ttype == "code" else "review")).strip()
-    if engine not in flows.ENGINES:
-        engine = flow.get("engine") or ("code" if ttype == "code" else "review")
-    difficulty, difficulty_reason = _difficulty(raw, flow)
-    dimension = dispatch.task_dimension(ttype, raw.get("role") or "")
-    rubric = raw.get("rubric") or flow.get("rubric") or []
-    if isinstance(rubric, str):
-        rubric = [x.strip() for x in rubric.replace("，", ",").split(",") if x.strip()]
-    elif not isinstance(rubric, (list, tuple)):
-        rubric = flow.get("rubric") or []
-    rubric = [str(x).strip() for x in rubric if str(x).strip()][:8]
-    deliverable = str(raw.get("manuscript") or flow.get("manuscript") or "").strip()
-    serial = raw.get("serial") if isinstance(raw.get("serial"), dict) else None
-    return {
-        "schema_version": SCHEMA_VERSION,
-        "type": ttype,
-        "title": str(raw.get("title") or "")[:120],
-        "goal": str(raw.get("goal") or "")[:2000],
-        "engine": engine,
-        "dimension": dimension,
-        "difficulty": difficulty,
-        "difficulty_reason": difficulty_reason,
-        "thinking": (str(raw.get("thinking") or "standard").lower()
-                     if str(raw.get("thinking") or "standard").lower()
-                     in ("auto", "low", "standard", "high") else "auto"),
-        "capabilities": _capabilities(raw, dimension, engine),
-        "deliverable": deliverable,
-        "quality_dimensions": rubric,
-        "serial": serial,
-        "explicit_agent": str(raw.get("implementer") or "").strip(),
-        "has_verification": bool(raw.get("verify_command")),
-    }
-
-
-def summary(spec):
-    """给运行日志/UI 的短说明，不包含用户正文。"""
-    spec = spec or {}
-    return "%s/%s/%s（能力：%s）" % (
-        spec.get("type") or "direct", spec.get("engine") or "direct",
-        spec.get("difficulty") or "default",
-        "、".join(spec.get("capabilities") or []) or "通用")
-
-
-def code_workflow(task, difficulty, plan=None, mode="auto", planned=True):
-    """把难度、验收能力和规划建议收敛成代码任务的实际步骤策略。
-
-    自动模式允许低风险任务走短链；手动模式与高风险任务保留完整质量门禁。
-    规划器只能在安全边界内减少步骤，不能让无确定性验证的任务跳过评审。
-    """
-    task = task if isinstance(task, Mapping) else {}
-    rec = (plan or {}).get("workflow") if isinstance(plan, Mapping) else None
-    rec = rec if isinstance(rec, Mapping) else {}
-    has_verify = bool(str(task.get("verify_command") or "").strip())
-    risk_text = " ".join(str(task.get(k) or "") for k in ("title", "goal", "context")).lower()
-    high_risk = bool(any(k in risk_text for k in (
-        "安全", "权限", "鉴权", "认证", "加密", "密钥", "注入", "迁移", "数据库",
-        "并发", "竞态", "支付", "架构", "公共接口", "public api", "公共 api",
-        "breaking change", "security", "permission", "auth", "migration",
-        "concurrency", "architecture")))
-    fast_mode = mode == "fast"
-    expert_mode = mode == "expert"
-    easy_auto = (mode == "auto" and difficulty == "easy") or fast_mode
-
-    review_required = True
-    max_repairs = 2
-    allow_switch = True
-    reason = "复杂或手动任务使用完整质量门禁"
-    if easy_auto:
-        review_required = high_risk or not has_verify
-        max_repairs = 1
-        allow_switch = False
-        reason = ("低风险任务以确定性验证代替模型评审"
-                  if has_verify else "低风险任务无验证命令，保留一次模型评审")
-        if isinstance(rec.get("review_required"), bool):
-            review_required = high_risk or rec["review_required"] or not has_verify
-        try:
-            if "max_repair_rounds" in rec:
-                max_repairs = max(1 if not has_verify else 0,
-                                  min(1, int(rec["max_repair_rounds"])))
-        except (TypeError, ValueError):
-            pass
-    elif expert_mode or (mode == "auto" and difficulty == "hard"):
-        review_required = True
-        max_repairs = 2
-        allow_switch = True
-        try:
-            if "max_repair_rounds" in rec:
-                max_repairs = max(1, min(2, int(rec["max_repair_rounds"])))
-        except (TypeError, ValueError):
-            pass
-        if isinstance(rec.get("allow_switch"), bool):
-            allow_switch = rec["allow_switch"]
-        reason = "复杂任务保留规划、验证与模型评审"
-
-    return {
-        "planning": "llm" if planned else "inline",
-        "implementation_steps": len((plan or {}).get("steps") or []) or 1,
-        "verification": has_verify,
-        "review_required": review_required,
-        "reviewers": 1 if review_required else 0,
-        "max_repair_rounds": max_repairs,
-        "allow_switch": allow_switch,
-        "reason": reason,
-    }
-
-
-def content_workflow(task, difficulty, plan=None, mode="auto"):
-    """为非连载内容任务生成动态大纲/评审/修订策略。"""
-    task = task if isinstance(task, Mapping) else {}
-    ttype = str(task.get("type") or "doc")
-    requested_rounds = max(1, min(5, int(task.get("rounds") or 2)))
-    rec = (plan or {}).get("workflow") if isinstance(plan, Mapping) else None
-    rec = rec if isinstance(rec, Mapping) else {}
-    light_types = {"email", "weekly_report", "translation"}
-    deep_types = {"novel", "research", "tech_proposal"}
-
-    if mode == "fast":
-        return {"outline": False, "reviewers": 1, "review_rounds": 1,
-                "reason": "快速模式省略独立大纲，仅保留一次快速评审"}
-    if mode == "expert":
-        return {"outline": True, "reviewers": 2,
-                "review_rounds": min(requested_rounds, 3),
-                "reason": "专家模式启用完整大纲、多评审与修订门禁"}
-    if mode != "auto":
-        return {"outline": True, "reviewers": 0,
-                "review_rounds": requested_rounds,
-                "reason": "手动模式尊重用户指定的评审组与轮数"}
-
-    if ttype in light_types and difficulty != "hard":
-        return {"outline": False, "reviewers": 1, "review_rounds": 1,
-                "reason": "轻量内容省略独立大纲，只做一次快速评审"}
-
-    outline = ttype in deep_types or difficulty == "hard"
-    reviewers = 2 if outline else 1
-    review_rounds = min(requested_rounds, 2 if outline else 1)
-    try:
-        if "reviewers" in rec:
-            reviewers = max(1, min(2, int(rec["reviewers"])))
-        if "review_rounds" in rec:
-            review_rounds = max(1, min(requested_rounds, 3,
-                                       int(rec["review_rounds"])))
-    except (TypeError, ValueError):
-        pass
-    # 深度类型至少保留一轮评审，但允许规划器把双评审降成一名；高门槛强制双评审。
-    if float(task.get("threshold") or 7.0) >= 8.5:
-        reviewers = 2
-    return {"outline": outline, "reviewers": reviewers,
-            "review_rounds": review_rounds,
-            "reason": ("深度内容保留大纲和多轮质量门禁" if outline
-                       else "标准内容使用单评审短链")}
+# -*- coding: utf-8 -*-
+"""任务编译器：把用户任务和预置流程编译成统一、可审计的运行规格。"""
+from __future__ import annotations
+
+from collections.abc import Mapping
+
+from . import dispatch, flows
+
+SCHEMA_VERSION = 1
+DIFFICULTIES = ("easy", "default", "hard")
+
+
+def _difficulty(task, flow):
+    explicit = str((task or {}).get("difficulty") or "").strip().lower()
+    if explicit in DIFFICULTIES:
+        return explicit, "用户指定"
+    try:
+        threshold = float((task or {}).get("threshold")
+                          if (task or {}).get("threshold") is not None
+                          else (flow or {}).get("threshold") or 7.0)
+    except (TypeError, ValueError):
+        threshold = 7.0
+    if threshold >= 8.5:
+        return "hard", "高质量门槛"
+    if threshold <= 6.0:
+        return "easy", "快速交付门槛"
+    if str((task or {}).get("type") or "").lower() in (
+            "research", "tech_proposal", "rank_scan"):
+        return "hard", "研究/方案类默认需要深度"
+    return "default", "标准难度"
+
+
+def _capabilities(task, dimension, engine):
+    caps = [dimension]
+    if engine == "code" or (task or {}).get("verify_command"):
+        caps.extend(["filesystem", "verification"])
+    if (task or {}).get("serial"):
+        caps.extend(["long_context", "continuity"])
+    if (task or {}).get("attachments"):
+        caps.append("attachments")
+    return list(dict.fromkeys(caps))
+
+
+def compile_task(task):
+    """返回统一任务规格；输入缺失或字段异常时保持可编排的安全兜底。"""
+    raw = dict(task) if isinstance(task, Mapping) else {}
+    ttype = str(raw.get("type") or "direct").strip().lower()
+    flow = flows.get_flow(ttype) or {}
+    engine = str(raw.get("engine") or flow.get("engine") or
+                 ("code" if ttype == "code" else "review")).strip()
+    if engine not in flows.ENGINES:
+        engine = flow.get("engine") or ("code" if ttype == "code" else "review")
+    difficulty, difficulty_reason = _difficulty(raw, flow)
+    dimension = dispatch.task_dimension(ttype, raw.get("role") or "")
+    rubric = raw.get("rubric") or flow.get("rubric") or []
+    if isinstance(rubric, str):
+        rubric = [x.strip() for x in rubric.replace("，", ",").split(",") if x.strip()]
+    elif not isinstance(rubric, (list, tuple)):
+        rubric = flow.get("rubric") or []
+    rubric = [str(x).strip() for x in rubric if str(x).strip()][:8]
+    deliverable = str(raw.get("manuscript") or flow.get("manuscript") or "").strip()
+    serial = raw.get("serial") if isinstance(raw.get("serial"), dict) else None
+    return {
+        "schema_version": SCHEMA_VERSION,
+        "type": ttype,
+        "title": str(raw.get("title") or "")[:120],
+        "goal": str(raw.get("goal") or "")[:2000],
+        "engine": engine,
+        "dimension": dimension,
+        "difficulty": difficulty,
+        "difficulty_reason": difficulty_reason,
+        "thinking": (str(raw.get("thinking") or "standard").lower()
+                     if str(raw.get("thinking") or "standard").lower()
+                     in ("auto", "low", "standard", "high") else "auto"),
+        "capabilities": _capabilities(raw, dimension, engine),
+        "deliverable": deliverable,
+        "quality_dimensions": rubric,
+        "serial": serial,
+        "explicit_agent": str(raw.get("implementer") or "").strip(),
+        "has_verification": bool(raw.get("verify_command")),
+    }
+
+
+def summary(spec):
+    """给运行日志/UI 的短说明，不包含用户正文。"""
+    spec = spec or {}
+    return "%s/%s/%s（能力：%s）" % (
+        spec.get("type") or "direct", spec.get("engine") or "direct",
+        spec.get("difficulty") or "default",
+        "、".join(spec.get("capabilities") or []) or "通用")
+
+
+def code_workflow(task, difficulty, plan=None, mode="auto", planned=True):
+    """把难度、验收能力和规划建议收敛成代码任务的实际步骤策略。
+
+    自动模式允许低风险任务走短链；手动模式与高风险任务保留完整质量门禁。
+    规划器只能在安全边界内减少步骤，不能让无确定性验证的任务跳过评审。
+    """
+    task = task if isinstance(task, Mapping) else {}
+    rec = (plan or {}).get("workflow") if isinstance(plan, Mapping) else None
+    rec = rec if isinstance(rec, Mapping) else {}
+    has_verify = bool(str(task.get("verify_command") or "").strip())
+    risk_text = " ".join(str(task.get(k) or "") for k in ("title", "goal", "context")).lower()
+    high_risk = bool(any(k in risk_text for k in (
+        "安全", "权限", "鉴权", "认证", "加密", "密钥", "注入", "迁移", "数据库",
+        "并发", "竞态", "支付", "架构", "公共接口", "public api", "公共 api",
+        "breaking change", "security", "permission", "auth", "migration",
+        "concurrency", "architecture")))
+    fast_mode = mode == "fast"
+    expert_mode = mode == "expert"
+    easy_auto = (mode == "auto" and difficulty == "easy") or fast_mode
+
+    review_required = True
+    max_repairs = 2
+    allow_switch = True
+    reason = "复杂或手动任务使用完整质量门禁"
+    if easy_auto:
+        review_required = high_risk or not has_verify
+        max_repairs = 1
+        allow_switch = False
+        reason = ("低风险任务以确定性验证代替模型评审"
+                  if has_verify else "低风险任务无验证命令，保留一次模型评审")
+        if isinstance(rec.get("review_required"), bool):
+            review_required = high_risk or rec["review_required"] or not has_verify
+        try:
+            if "max_repair_rounds" in rec:
+                max_repairs = max(1 if not has_verify else 0,
+                                  min(1, int(rec["max_repair_rounds"])))
+        except (TypeError, ValueError):
+            pass
+    elif expert_mode or (mode == "auto" and difficulty == "hard"):
+        review_required = True
+        max_repairs = 2
+        allow_switch = True
+        try:
+            if "max_repair_rounds" in rec:
+                max_repairs = max(1, min(2, int(rec["max_repair_rounds"])))
+        except (TypeError, ValueError):
+            pass
+        if isinstance(rec.get("allow_switch"), bool):
+            allow_switch = rec["allow_switch"]
+        reason = "复杂任务保留规划、验证与模型评审"
+
+    return {
+        "planning": "llm" if planned else "inline",
+        "implementation_steps": len((plan or {}).get("steps") or []) or 1,
+        "verification": has_verify,
+        "review_required": review_required,
+        "reviewers": 1 if review_required else 0,
+        "max_repair_rounds": max_repairs,
+        "allow_switch": allow_switch,
+        "reason": reason,
+    }
+
+
+def content_workflow(task, difficulty, plan=None, mode="auto"):
+    """为非连载内容任务生成动态大纲/评审/修订策略。"""
+    task = task if isinstance(task, Mapping) else {}
+    ttype = str(task.get("type") or "doc")
+    requested_rounds = max(1, min(5, int(task.get("rounds") or 2)))
+    rec = (plan or {}).get("workflow") if isinstance(plan, Mapping) else None
+    rec = rec if isinstance(rec, Mapping) else {}
+    light_types = {"email", "weekly_report", "translation"}
+    deep_types = {"novel", "research", "tech_proposal"}
+
+    if mode == "fast":
+        return {"outline": False, "reviewers": 1, "review_rounds": 1,
+                "reason": "快速模式省略独立大纲，仅保留一次快速评审"}
+    if mode == "expert":
+        return {"outline": True, "reviewers": 2,
+                "review_rounds": min(requested_rounds, 3),
+                "reason": "专家模式启用完整大纲、多评审与修订门禁"}
+    if mode != "auto":
+        return {"outline": True, "reviewers": 0,
+                "review_rounds": requested_rounds,
+                "reason": "手动模式尊重用户指定的评审组与轮数"}
+
+    if ttype in light_types and difficulty != "hard":
+        return {"outline": False, "reviewers": 1, "review_rounds": 1,
+                "reason": "轻量内容省略独立大纲，只做一次快速评审"}
+
+    outline = ttype in deep_types or difficulty == "hard"
+    reviewers = 2 if outline else 1
+    review_rounds = min(requested_rounds, 2 if outline else 1)
+    try:
+        if "reviewers" in rec:
+            reviewers = max(1, min(2, int(rec["reviewers"])))
+        if "review_rounds" in rec:
+            review_rounds = max(1, min(requested_rounds, 3,
+                                       int(rec["review_rounds"])))
+    except (TypeError, ValueError):
+        pass
+    # 深度类型至少保留一轮评审，但允许规划器把双评审降成一名；高门槛强制双评审。
+    if float(task.get("threshold") or 7.0) >= 8.5:
+        reviewers = 2
+    return {"outline": outline, "reviewers": reviewers,
+            "review_rounds": review_rounds,
+            "reason": ("深度内容保留大纲和多轮质量门禁" if outline
+                       else "标准内容使用单评审短链")}