codebee 0.1.95 → 0.1.97

package/app/core/dispatch_log.py

@@ -1,128 +1,128 @@
-# -*- coding: utf-8 -*-
-"""调度事件台账：记录脱敏的候选、决策和结果，支持按运行回放。"""
-from __future__ import annotations
-
-import json
-import hashlib
-import threading
-import time
-
-from . import paths
-from .redact import scrub_text
-
-LOCK = threading.RLock()
-
-
-def _dispatch_dir():
-    # 读取动态 DATA_DIR，测试和多实例运行可在启动后重定向数据目录。
-    return paths.DATA_DIR / "dispatch"
-
-
-def _month_file(day):
-    return _dispatch_dir() / ("dispatch-%s.jsonl" % day[:7].replace("-", ""))
-
-
-def _text(value, limit=240):
-    value = str(value or "").replace("\r", " ").replace("\n", " ")
-    return scrub_text(value, limit=limit)
-
-
-def _candidate(row):
-    if not isinstance(row, dict):
-        return {}
-    return {"agent_id": _text(row.get("agent_id"), 64),
-            "label": _text(row.get("label"), 80),
-            "kind": _text(row.get("kind"), 32),
-            "score": row.get("score", 0),
-            "reason": _text(row.get("reason"), 240),
-            "order": row.get("order", 0)}
-
-
-def record_event(run_id="", task_id="", task_type="", difficulty="", role="",
-                 phase="selected", selected="", participants=(), candidates=(),
-                 fallback=(), selection_reason="", result="", verify_pass=None,
-                 review_pass=None, trace_id="", span_id="", parent_event_id=""):
-    """追加一条事件；仅保存路由元数据，不保存正文、提示词、密钥或文件内容。"""
-    try:
-        day = time.strftime("%Y-%m-%d")
-        event = {
-            "ts": time.strftime("%Y-%m-%d %H:%M:%S"),
-            "run_id": _text(run_id, 64), "task_id": _text(task_id, 64),
-            "task_type": _text(task_type, 32), "difficulty": _text(difficulty, 16),
-            "role": _text(role, 40), "phase": _text(phase, 16),
-            "selected": _text(selected, 64),
-            "participants": [_text(x.get("id") if isinstance(x, dict) else x, 64)
-                             for x in (participants or ())],
-            "candidates": [_candidate(x) for x in (candidates or ()) if isinstance(x, dict)],
-            "fallback": [_text(x, 64) for x in (fallback or ())],
-            "selection_reason": _text(selection_reason, 400),
-            "result": _text(result, 32),
-        }
-        # Trace identifiers are routing metadata only.  They deliberately do
-        # not carry prompt/output contents and remain optional for old callers.
-        if run_id:
-            from . import tracing
-            event["trace_id"] = _text(trace_id or tracing.trace_id(run_id), 64)
-        if span_id:
-            event["span_id"] = _text(span_id, 32)
-        if parent_event_id:
-            event["parent_event_id"] = _text(parent_event_id, 64)
-        event["event_id"] = hashlib.sha256(
-            ("%s|%s|%s|%s|%s|%s" % (
-                time.time_ns(), run_id, phase, role, selected, result
-            )).encode("utf-8")
-        ).hexdigest()[:24]
-        if verify_pass is not None:
-            event["verify_pass"] = bool(verify_pass)
-        if review_pass is not None:
-            event["review_pass"] = bool(review_pass)
-        with LOCK:
-            _dispatch_dir().mkdir(parents=True, exist_ok=True)
-            with open(_month_file(day), "a", encoding="utf-8") as fh:
-                fh.write(json.dumps(event, ensure_ascii=False) + "\n")
-        return event
-    except Exception:
-        return None
-
-
-def _iter_events(run_id="", task_type="", limit=100):
-    """从新到旧读取，满足 limit 即停，避免回放请求解析全部历史。"""
-    try:
-        directory = _dispatch_dir()
-        files = sorted(directory.glob("dispatch-*.jsonl"), reverse=True) \
-            if directory.is_dir() else []
-    except Exception:
-        return []
-    events = []
-    for path in files:
-        try:
-            lines = path.read_text(encoding="utf-8", errors="replace").splitlines()
-            for line in reversed(lines):
-                if not line.strip().startswith("{"):
-                    continue
-                try:
-                    item = json.loads(line)
-                except Exception:
-                    continue
-                if not isinstance(item, dict):
-                    continue
-                if run_id and item.get("run_id") != run_id:
-                    continue
-                if task_type and item.get("task_type") != task_type:
-                    continue
-                events.append(item)
-                if len(events) >= limit:
-                    return events
-        except Exception:
-            continue
-    return events
-
-
-def replay(run_id="", task_type="", limit=100):
-    """返回最新的脱敏调度事件，支持按 run_id/task_type 筛选。"""
-    try:
-        limit = max(1, min(1000, int(limit)))
-    except (TypeError, ValueError):
-        limit = 100
-    rid, ttype = _text(run_id, 64), _text(task_type, 32)
-    return _iter_events(rid, ttype, limit)
+# -*- coding: utf-8 -*-
+"""调度事件台账：记录脱敏的候选、决策和结果，支持按运行回放。"""
+from __future__ import annotations
+
+import json
+import hashlib
+import threading
+import time
+
+from . import paths
+from .redact import scrub_text
+
+LOCK = threading.RLock()
+
+
+def _dispatch_dir():
+    # 读取动态 DATA_DIR，测试和多实例运行可在启动后重定向数据目录。
+    return paths.DATA_DIR / "dispatch"
+
+
+def _month_file(day):
+    return _dispatch_dir() / ("dispatch-%s.jsonl" % day[:7].replace("-", ""))
+
+
+def _text(value, limit=240):
+    value = str(value or "").replace("\r", " ").replace("\n", " ")
+    return scrub_text(value, limit=limit)
+
+
+def _candidate(row):
+    if not isinstance(row, dict):
+        return {}
+    return {"agent_id": _text(row.get("agent_id"), 64),
+            "label": _text(row.get("label"), 80),
+            "kind": _text(row.get("kind"), 32),
+            "score": row.get("score", 0),
+            "reason": _text(row.get("reason"), 240),
+            "order": row.get("order", 0)}
+
+
+def record_event(run_id="", task_id="", task_type="", difficulty="", role="",
+                 phase="selected", selected="", participants=(), candidates=(),
+                 fallback=(), selection_reason="", result="", verify_pass=None,
+                 review_pass=None, trace_id="", span_id="", parent_event_id=""):
+    """追加一条事件；仅保存路由元数据，不保存正文、提示词、密钥或文件内容。"""
+    try:
+        day = time.strftime("%Y-%m-%d")
+        event = {
+            "ts": time.strftime("%Y-%m-%d %H:%M:%S"),
+            "run_id": _text(run_id, 64), "task_id": _text(task_id, 64),
+            "task_type": _text(task_type, 32), "difficulty": _text(difficulty, 16),
+            "role": _text(role, 40), "phase": _text(phase, 16),
+            "selected": _text(selected, 64),
+            "participants": [_text(x.get("id") if isinstance(x, dict) else x, 64)
+                             for x in (participants or ())],
+            "candidates": [_candidate(x) for x in (candidates or ()) if isinstance(x, dict)],
+            "fallback": [_text(x, 64) for x in (fallback or ())],
+            "selection_reason": _text(selection_reason, 400),
+            "result": _text(result, 32),
+        }
+        # Trace identifiers are routing metadata only.  They deliberately do
+        # not carry prompt/output contents and remain optional for old callers.
+        if run_id:
+            from . import tracing
+            event["trace_id"] = _text(trace_id or tracing.trace_id(run_id), 64)
+        if span_id:
+            event["span_id"] = _text(span_id, 32)
+        if parent_event_id:
+            event["parent_event_id"] = _text(parent_event_id, 64)
+        event["event_id"] = hashlib.sha256(
+            ("%s|%s|%s|%s|%s|%s" % (
+                time.time_ns(), run_id, phase, role, selected, result
+            )).encode("utf-8")
+        ).hexdigest()[:24]
+        if verify_pass is not None:
+            event["verify_pass"] = bool(verify_pass)
+        if review_pass is not None:
+            event["review_pass"] = bool(review_pass)
+        with LOCK:
+            _dispatch_dir().mkdir(parents=True, exist_ok=True)
+            with open(_month_file(day), "a", encoding="utf-8") as fh:
+                fh.write(json.dumps(event, ensure_ascii=False) + "\n")
+        return event
+    except Exception:
+        return None
+
+
+def _iter_events(run_id="", task_type="", limit=100):
+    """从新到旧读取，满足 limit 即停，避免回放请求解析全部历史。"""
+    try:
+        directory = _dispatch_dir()
+        files = sorted(directory.glob("dispatch-*.jsonl"), reverse=True) \
+            if directory.is_dir() else []
+    except Exception:
+        return []
+    events = []
+    for path in files:
+        try:
+            lines = path.read_text(encoding="utf-8", errors="replace").splitlines()
+            for line in reversed(lines):
+                if not line.strip().startswith("{"):
+                    continue
+                try:
+                    item = json.loads(line)
+                except Exception:
+                    continue
+                if not isinstance(item, dict):
+                    continue
+                if run_id and item.get("run_id") != run_id:
+                    continue
+                if task_type and item.get("task_type") != task_type:
+                    continue
+                events.append(item)
+                if len(events) >= limit:
+                    return events
+        except Exception:
+            continue
+    return events
+
+
+def replay(run_id="", task_type="", limit=100):
+    """返回最新的脱敏调度事件，支持按 run_id/task_type 筛选。"""
+    try:
+        limit = max(1, min(1000, int(limit)))
+    except (TypeError, ValueError):
+        limit = 100
+    rid, ttype = _text(run_id, 64), _text(task_type, 32)
+    return _iter_events(rid, ttype, limit)

package/app/core/env_scrub.py

@@ -1,86 +1,86 @@
-# -*- coding: utf-8 -*-
-"""环境变量净化：剥离含 KEY/SECRET/TOKEN/PASSWORD 等密钥名的变量。
-
-设计稿：docs/migration/01-defense-patterns.md §5A。
-参考 dsh：docs/defensive-patterns.zh.md 第 31 行（"绝不将环境变量或可预测路径暴露给不可信输出"）。
-"""
-from __future__ import annotations
-
-import logging
-import re
-
-log = logging.getLogger(__name__)
-
-# 黑名单：环境变量名含这些子串（不区分大小写）即剥离。
-_DENY_RE = re.compile(r"(KEY|SECRET|TOKEN|PASSWORD|PASSWD|CREDENTIAL)", re.IGNORECASE)
-
-# 白名单：精确匹配（不分大小写），用于覆盖黑名单。
-_ALLOW_EXACT = frozenset({
-    # 系统/环境（TEMP/TMP/PATHEXT 等是 Windows 名，TMPDIR 是 macOS 每用户
-    # 临时目录——三边各留各的，多出的名字在别的平台上不存在、无副作用）
-    "PATH", "PATHEXT", "SYSTEMROOT", "TEMP", "TMP", "TMPDIR",
-    "USERPROFILE", "USERDNSDOMAIN", "HOMEPATH", "HOMEDRIVE",
-    "LANG", "LC_ALL", "LC_CTYPE",
-    "HOME", "SHELL", "TERM", "PYTHONIOENCODING",
-    "PYTHONPATH", "PYTHONUNBUFFERED",
-    "COMSPEC", "WINDIR",
-    # Tutti / DSH 内部
-    "DSH_HOME", "TUTTI_HOME", "ALLINONE_HOME",
-    # Claude CLI 自有（不是密钥本身，是配置）
-    "CLAUDE_CODE_ENTRYPOINT", "CLAUDE_CODE_SSE_PORT",
-    # PowerShell / Windows
-    "PSMODULEPATH", "PROMPT",
-})
-
-# 白名单前缀（区分大小写以匹配 Windows 习惯）；含前缀视为放行。
-_ALLOW_PREFIX = (
-    "CODEX_", "CLAUDE_CODE_", "OPENAI_", "ANTHROPIC_",
-    "QWEN_", "AIDER_", "OPENCODE_", "GEMINI_",
-    "TUTTI_", "ALLINONE_", "DSH_",
-    "LANG", "LC_",  # locale 系列
-    "PYTHON",       # PYTHONPATH / PYTHONIOENCODING 等
-    "XDG_",         # XDG_CONFIG_HOME 等（macOS/Linux CLI 认它找配置/缓存落点）
-    "CHROME", "EDGE", "PLAYWRIGHT",  # 本机测试工具
-)
-
-
-def _should_keep(k: str) -> bool:
-    if k in _ALLOW_EXACT:
-        return True
-    for p in _ALLOW_PREFIX:
-        if k.startswith(p):
-            return True
-    return False
-
-
-def scrub_env(env: dict, *, mode: str = "drop") -> dict:
-    """返回净化后的 env 副本（原 dict 不修改）。
-
-    Args:
-        env: 原始环境变量字典
-        mode:
-          "drop" - 直接删除命中黑名单的 key（默认）
-          "stub" - 把值替换为 "__REDACTED__"（子进程若检查值类型可能崩）
-          "off"  - 完全不过滤（用于测试 / 排查）
-    """
-    if mode == "off":
-        return dict(env)
-    out: dict = {}
-    dropped: list = []
-    stubbed: list = []
-    for k, v in env.items():
-        if _should_keep(k):
-            out[k] = v
-        elif _DENY_RE.search(k):
-            if mode == "stub":
-                out[k] = "__REDACTED__"
-                stubbed.append(k)
-            else:
-                dropped.append(k)
-        else:
-            out[k] = v
-    if dropped:
-        log.info("env-scrub dropped %d keys: %s", len(dropped), dropped[:20])
-    if stubbed:
-        log.info("env-scrub stubbed %d keys: %s", len(stubbed), stubbed[:20])
+# -*- coding: utf-8 -*-
+"""环境变量净化：剥离含 KEY/SECRET/TOKEN/PASSWORD 等密钥名的变量。
+
+设计稿：docs/migration/01-defense-patterns.md §5A。
+参考 dsh：docs/defensive-patterns.zh.md 第 31 行（"绝不将环境变量或可预测路径暴露给不可信输出"）。
+"""
+from __future__ import annotations
+
+import logging
+import re
+
+log = logging.getLogger(__name__)
+
+# 黑名单：环境变量名含这些子串（不区分大小写）即剥离。
+_DENY_RE = re.compile(r"(KEY|SECRET|TOKEN|PASSWORD|PASSWD|CREDENTIAL)", re.IGNORECASE)
+
+# 白名单：精确匹配（不分大小写），用于覆盖黑名单。
+_ALLOW_EXACT = frozenset({
+    # 系统/环境（TEMP/TMP/PATHEXT 等是 Windows 名，TMPDIR 是 macOS 每用户
+    # 临时目录——三边各留各的，多出的名字在别的平台上不存在、无副作用）
+    "PATH", "PATHEXT", "SYSTEMROOT", "TEMP", "TMP", "TMPDIR",
+    "USERPROFILE", "USERDNSDOMAIN", "HOMEPATH", "HOMEDRIVE",
+    "LANG", "LC_ALL", "LC_CTYPE",
+    "HOME", "SHELL", "TERM", "PYTHONIOENCODING",
+    "PYTHONPATH", "PYTHONUNBUFFERED",
+    "COMSPEC", "WINDIR",
+    # Tutti / DSH 内部
+    "DSH_HOME", "TUTTI_HOME", "ALLINONE_HOME",
+    # Claude CLI 自有（不是密钥本身，是配置）
+    "CLAUDE_CODE_ENTRYPOINT", "CLAUDE_CODE_SSE_PORT",
+    # PowerShell / Windows
+    "PSMODULEPATH", "PROMPT",
+})
+
+# 白名单前缀（区分大小写以匹配 Windows 习惯）；含前缀视为放行。
+_ALLOW_PREFIX = (
+    "CODEX_", "CLAUDE_CODE_", "OPENAI_", "ANTHROPIC_",
+    "QWEN_", "AIDER_", "OPENCODE_", "GEMINI_",
+    "TUTTI_", "ALLINONE_", "DSH_",
+    "LANG", "LC_",  # locale 系列
+    "PYTHON",       # PYTHONPATH / PYTHONIOENCODING 等
+    "XDG_",         # XDG_CONFIG_HOME 等（macOS/Linux CLI 认它找配置/缓存落点）
+    "CHROME", "EDGE", "PLAYWRIGHT",  # 本机测试工具
+)
+
+
+def _should_keep(k: str) -> bool:
+    if k in _ALLOW_EXACT:
+        return True
+    for p in _ALLOW_PREFIX:
+        if k.startswith(p):
+            return True
+    return False
+
+
+def scrub_env(env: dict, *, mode: str = "drop") -> dict:
+    """返回净化后的 env 副本（原 dict 不修改）。
+
+    Args:
+        env: 原始环境变量字典
+        mode:
+          "drop" - 直接删除命中黑名单的 key（默认）
+          "stub" - 把值替换为 "__REDACTED__"（子进程若检查值类型可能崩）
+          "off"  - 完全不过滤（用于测试 / 排查）
+    """
+    if mode == "off":
+        return dict(env)
+    out: dict = {}
+    dropped: list = []
+    stubbed: list = []
+    for k, v in env.items():
+        if _should_keep(k):
+            out[k] = v
+        elif _DENY_RE.search(k):
+            if mode == "stub":
+                out[k] = "__REDACTED__"
+                stubbed.append(k)
+            else:
+                dropped.append(k)
+        else:
+            out[k] = v
+    if dropped:
+        log.info("env-scrub dropped %d keys: %s", len(dropped), dropped[:20])
+    if stubbed:
+        log.info("env-scrub stubbed %d keys: %s", len(stubbed), stubbed[:20])
     return out