fix(remote_env): 批量pip失败时降级逐包安装(部分成功优于全败)

pipeline_service/remote_env_init.py

@@ -312,14 +312,55 @@ async def install_remote_deps(env: dict, specs: list, pip_index: str,
         installed = list(specs)
         failed = []
     else:
-        installed = [s for s in specs if all(
-            not s.lower().startswith(re.split(r"[\[<>=!~\s]", f, 1)[0].lower())
-            for f in failed)]
+        # 批量失败 → 逐包降级：一次 SSH 会话内逐包安装（pip 幂等，已装的秒过），
+        # 部分成功优于全败。
+        return await _install_per_package(env, specs, pip_index, rel, timeout)
     return {"ok": rc == 0, "installed": installed, "failed": failed,
             "venv": "~/" + rel + "/.venv-skills",
             "tail": out[-800:]}
 
 
+async def _install_per_package(env: dict, specs: list, pip_index: str,
+                               rel: str, timeout: int) -> dict:
+    """逐包安装（批量失败后的降级路径）。返回格式与 install_remote_deps 一致。"""
+    payload = base64.b64encode(json.dumps(specs).encode()).decode()
+    index_arg = ('-i ' + shlex.quote(pip_index) + ' ') if pip_index else ''
+    script = (
+        'RD="$HOME/' + rel + '"; VENV="$RD/.venv-skills"; '
+        '[ -x "$VENV/bin/pip" ] || { echo "VENV_MISSING"; exit 4; }; '
+        'set -f; '
+        'echo "' + payload + '" | base64 -d | python3 -c '
+        '"import json,sys;print(chr(10).join(json.load(sys.stdin)))" | '
+        'while IFS= read -r pkg; do '
+        '  [ -n "$pkg" ] || continue; '
+        '  if "$VENV/bin/pip" install --no-input --disable-pip-version-check '
+        + index_arg + '"$pkg" >/dev/null 2>&1; then echo "OK:$pkg"; '
+        '  else echo "FAILED:$pkg"; fi; '
+        'done'
+    )
+    ssh_cmd = ["ssh"] + _ssh_common_args(env) + [_ssh_target(env), script]
+    try:
+        proc = await asyncio.create_subprocess_exec(
+            *ssh_cmd, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE)
+        stdout, _stderr = await asyncio.wait_for(proc.communicate(), timeout=timeout + 30)
+    except asyncio.TimeoutError:
+        return {"ok": False, "error": f"逐包安装超时（{timeout}s），可重跑"}
+    except Exception as e:
+        return {"ok": False, "error": "SSH 执行失败: " + str(e)[:300]}
+
+    out = stdout.decode("utf-8", "replace")
+    installed, failed = [], []
+    for line in out.splitlines():
+        if line.startswith("OK:"):
+            installed.append(line[3:])
+        elif line.startswith("FAILED:"):
+            failed.append(line[7:])
+    return {"ok": len(failed) == 0 and len(installed) > 0,
+            "installed": installed, "failed": failed,
+            "venv": "~/" + rel + "/.venv-skills",
+            "mode": "per-package", "tail": out[-500:]}
+
+
 # ── 总编排 ────────────────────────────────────────────────────
 
 async def init_remote_env(env: dict, org_id: str = "", sor=None,