From bcb2128b02be32699461e0f1ef4fc4da464f5553 Mon Sep 17 00:00:00 2001
From: sora <2075279110@qq.com>
Date: Thu, 17 Sep 2026 08:56:48 +0000
Subject: [PATCH] tau2 env: auto-install the engine (local checkout -> GitHub),
 manual fallback

Same UX as the docker image chain: try local sources first, then the
GitHub URL, and on total failure print the exact manual commands --
with a warning that PyPI's 'tau2' is an unrelated physics package.

Co-Authored-By: Claude <noreply@anthropic.com>
---
 evalharness/agent/envs/tau2_official.py | 50 +++++++++++++++++++++++++
 1 file changed, 50 insertions(+)

diff --git a/evalharness/agent/envs/tau2_official.py b/evalharness/agent/envs/tau2_official.py
index f91d686..2f4413c 100644
--- a/evalharness/agent/envs/tau2_official.py
+++ b/evalharness/agent/envs/tau2_official.py
@@ -11,6 +11,7 @@ The env consumes a Sample whose metadata carries the official Task json
 
 import asyncio
 import json
+import os
 from typing import Any, Dict, List, Optional
 
 from ...data.sample import ChatMessage, Sample
@@ -19,6 +20,54 @@ from ..loop import Environment, register_env
 _PATCHED = False
 
 
+_TAU2_GIT = 'git+https://github.com/sierra-research/tau2-bench'
+# 本机常见的 tau2-bench checkout 位置（es 仓库 tools/ 下），按序探测
+_TAU2_LOCAL_CANDIDATES = [
+    '/data1/sora/evalscope/tools/tau2-bench',
+    os.path.expanduser('~/tau2-bench',
+                       ) if hasattr(os.path, 'expanduser') else '',
+]
+
+
+def _ensure_tau2_engine():
+    """Auto-install the REAL tau2-bench engine when missing.
+
+    PyPI's 'tau2' is an unrelated physics package -- never 'pip install
+    tau2'. We try: (1) already importable, (2) a local checkout
+    (editable), (3) the GitHub URL, then print the manual fallback."""
+    try:
+        import tau2.data_model.message  # noqa: F401
+        return
+    except ImportError:
+        pass
+    import importlib
+    import subprocess
+    import sys
+
+    cands = [p for p in _TAU2_LOCAL_CANDIDATES if p and os.path.isdir(p)]
+    attempts = [('local checkout', ['-e', p]) for p in cands] + \
+               [('github', [_TAU2_GIT])]
+    for src, arg in attempts:
+        print(f'· tau2 engine missing -- installing from {src} ...', flush=True)
+        r = subprocess.run([sys.executable, '-m', 'pip', 'install', '-q', *arg],
+                           capture_output=True, text=True, timeout=600)
+        if r.returncode == 0:
+            try:
+                importlib.invalidate_caches()
+                import tau2.data_model.message  # noqa: F401
+                print(f'✓ tau2 engine installed from {src}', flush=True)
+                return
+            except ImportError:
+                continue
+    raise RuntimeError(
+        'tau2-bench 引擎不可用且自动安装失败。手动安装（任选一）：\n'
+        f'  A. pip install git+{_TAU2_GIT}\n'
+        '  B. 本机源码: pip install -e /path/to/tau2-bench\n'
+        '  C. 内网无出口时：在有网的机器 '
+        '`pip download tau2-bench --no-deps -d pkg/`？注意 PyPI 的 tau2 是'
+        '无关物理库，必须用 GitHub 源或本机源码！')
+
+
 def _patch_tau2_generate(adapter, user_adapter=None, gen_kwargs=None) -> None:
     """user_adapter: separate model for the USER simulator (es production
     parity: strong user model like DeepSeek while the agent under test stays
@@ -28,6 +77,7 @@ def _patch_tau2_generate(adapter, user_adapter=None, gen_kwargs=None) -> None:
     if _PATCHED:
         return
 
+    _ensure_tau2_engine()
     import tau2.utils.llm_utils as llm_utils
 
     original = llm_utils.generate