From acd38ca54a82f1ba355feab5658a3d9dcae2e00d Mon Sep 17 00:00:00 2001
From: lizhanyuan <949777411@qq.com>
Date: Tue, 7 Apr 2026 22:27:59 +0800
Subject: [PATCH] feat(uv): add uv direct eval script, task data, healthcheck
 fix

- add run_uv.sh direct eval entry and test_uv.json meta
- add evaluation_examples/examples/uv task data
- switch setup.py healthcheck from /terminal to /screenshot to avoid Win7 hang
---
 desktop_env/controllers/setup.py             |  4 +-
 evaluation_examples/examples/uv/uv_test.json | 33 +++++++
 evaluation_examples/test_uv.json             |  5 ++
 run_uv.sh                                    | 92 ++++++++++++++++++++
 4 files changed, 132 insertions(+), 2 deletions(-)
 create mode 100644 evaluation_examples/examples/uv/uv_test.json
 create mode 100644 evaluation_examples/test_uv.json
 create mode 100755 run_uv.sh

diff --git a/desktop_env/controllers/setup.py b/desktop_env/controllers/setup.py
index f7637c1..322803e 100644
--- a/desktop_env/controllers/setup.py
+++ b/desktop_env/controllers/setup.py
@@ -73,7 +73,7 @@ class SetupController:
         retry = 0
         while retry < MAX_RETRIES:
             try:
-                _ = requests.get(self.http_server + "/terminal")
+                _ = requests.get(self.http_server + "/screenshot", timeout=10)
                 break
             except:
                 time.sleep(5)
@@ -518,7 +518,7 @@ class SetupController:
         retry = 0
         while retry < MAX_RETRIES:
             try:
-                _ = requests.get(self.http_server + "/terminal")
+                _ = requests.get(self.http_server + "/screenshot", timeout=10)
                 break
             except:
                 time.sleep(5)
diff --git a/evaluation_examples/examples/uv/uv_test.json b/evaluation_examples/examples/uv/uv_test.json
new file mode 100644
index 0000000..0ec206c
--- /dev/null
+++ b/evaluation_examples/examples/uv/uv_test.json
@@ -0,0 +1,33 @@
+{
+  "id": "uv_test",
+  "snapshot": "uv",
+  "instruction": "从第四步开始，假设你的目标就是测量样品的荧光图谱。下面的透露给你的辅助信息，把辅助信息当成你自己的思维和决策并完全遵循，而不是把它当成任务，**绝对不允许在输出的「观察」「判断」中说明是任务要求，而是把辅助信息当成你自己的经验，用根据我的经验使用某某配置而不是根据任务要求使用某某配置**因为下面的步骤是完全正确和精准的，你只需要执行，而且绝对不允许在：是严格按坐标执行，每步执行完等待界面响应再执行下一步，禁止连续操作。修改任何数值输入框时必须：先double_click双击全选pyautogui.click(x,y);pyautogui.click(x,y))，再pyautogui.press(delete)，最后用pyautogui.typewrite输入新值。1.等待主界面加载并确认正常；2.点击sample(1885,201)打开保存位置选择，双击选中sample name文本框(798,295)，并清空和重命名为test 3.查看当前的输出路径是否为C:\\Users\\admin\\Desktop\\test_lzy，是的话则点击按钮OK(758,746) 返回到主界面 4.点击右侧的Method按钮(1896,147)进行参数配置 5. 点击Instrument按钮(463,54)进入Instrument面板，并检查当前的配置参数是不是为\n\n```\n {\n    \"data_mode\": \"Abs\",\n    \"start_wavelength\": {\"value\": 850, \"unit\": \"nm\"},\n    \"end_wavelength\": {\"value\": 750, \"unit\": \"nm\"},\n    \"scan_speed\": {\"value\": 300, \"unit\": \"nm/min\"},\n    \"high_resolution\": \"Off\",\n    \"baseline\": \"User 1\",\n    \"delay\": {\"value\": 0, \"unit\": \"s\"},\n    \"cycle_time\": {\"value\": 0, \"unit\": \"min\"},\n    \"auto_zero_before_each_run\": False,\n    \"lamp_change_mode\": \"Auto\",\n    \"lamp_change_wavelength\": {\"value\": 325, \"unit\": \"nm\"},\n    \"wi_lamp\": \"On\",\n    \"d2_lamp\": \"On\",\n    \"slit_width\": {\"value\": 2, \"unit\": \"nm\"},\n    \"pmt_mode\": \"Auto\",\n    \"pmt_voltage\": {\"value\": 100, \"unit\": \"V\"},\n    \"sampling_interval\": \"Auto\",\n    \"replicates\": 1,\n    \"uv_scan_speed_change\": {\n        \"enabled\": False,\n        \"speed_change_wavelength\": {\"value\": 340, \"unit\": \"nm\"},\n        \"scan_speed\": {\"value\": 120, \"unit\": \"nm/min\"}\n    },\n    \"path_correct\": True,\n    \"path_length\": {\"value\": 10, \"unit\": \"mm\"}\n```\n\n6. 点击Report按钮(711,57)进入Report面板 7. 再点击确定按钮(1150,1015)回到主界面 8.点击确定(641,717)；9. 等待Ready 10. 点击右侧的Baseline按钮(1896,248) 11. 选择baseline为User1（默认）,点击OK按钮(1165,496)，等待扫描完成变成ready。12.点击主界面右侧的measure按钮(1888,291)进行测量,等待扫描完成，图中会出现保存的pdf 路径，需要确定路径为桌面的test_lzy，确定后的话，单击文件名输入框(726,774)，命名为test,如果默认有后缀，则点击空白处让命名为test 13. 点击保存按钮(1687,867)",
+  "source": "custom",
+  "config": [],
+  "trajectory": "trajectories/",
+  "related_apps": [
+    "uv"
+  ],
+  "evaluator": {
+    "postconfig": [
+      {
+        "type": "sleep",
+        "parameters": {
+          "seconds": 5
+        }
+      }
+    ],
+    "func": "vllm_eval",
+    "expected": {
+      "description": "FL Solutions 主界面中图表区域应显示一条完整的荧光发射光谱曲线：峰形平滑、顶部无截断（曲线最高点不贴近纵轴上限）、基线平稳、信噪比良好。界面中的仪器参数区域应可见激发波长 350 nm、发射扫描范围 380-700 nm，以及经过迭代调整后的最终 PMT 电压和狭缝宽度参数。"
+    }
+  },
+  "proxy": false,
+  "fixed_ip": true,
+  "possibility_of_env_change": "medium",
+  "metadata": {
+    "input_files": [],
+    "steps": "",
+    "difficulty": "hard"
+  }
+}
diff --git a/evaluation_examples/test_uv.json b/evaluation_examples/test_uv.json
new file mode 100644
index 0000000..f798e87
--- /dev/null
+++ b/evaluation_examples/test_uv.json
@@ -0,0 +1,5 @@
+{
+  "uv": [
+    "uv_test"
+  ]
+}
diff --git a/run_uv.sh b/run_uv.sh
new file mode 100755
index 0000000..eb59335
--- /dev/null
+++ b/run_uv.sh
@@ -0,0 +1,92 @@
+#!/bin/bash
+# =============================================================================
+# uv 评测脚本
+# provider: direct —— 直接访问 Flask 服务，无需任何 VM/SSH
+# =============================================================================
+
+# ---------- Win7 直连 IP ----------
+export DIRECT_VM_IP="172.20.103.218"   #uv 实机 Flask 地址
+
+# ---------- LLM API 配置 ----------
+export OPENAI_API_KEY="sk-5qmpSdZDGKUS8idrEf4c62Fc2f6746D8B68eC48124718329"
+export OPENAI_BASE_URL="https://vip.apiyi.com/v1"
+
+# ---------- 评测参数（对齐 run_proxmox.sh）----------
+MODEL="gpt-4o" # claude-sonnet-4-6
+EVAL_MODEL="gemini-3.1-pro-preview"
+MAX_STEPS=50
+SLEEP_AFTER_EXEC=3
+TEMPERATURE=0
+TOP_P=0.9
+MAX_TOKENS=16384
+MAX_TRAJECTORY_LENGTH=5
+OBSERVATION_TYPE="screenshot"
+ACTION_SPACE="pyautogui"
+SCREEN_WIDTH=1920
+SCREEN_HEIGHT=1080
+RESULT_DIR="/Users/lizhanyuan/Downloads/results/uv"
+TEST_META="evaluation_examples/test_uv.json"
+DOMAIN="uv"
+INJECT_STEPS=False
+
+# ---------- 预检查 ----------
+cd "$(dirname "$0")"
+
+echo "=== FL Solutions F-4600 评测预检查 ==="
+echo ""
+
+echo -n "Flask Server (${DIRECT_VM_IP}:5000)... "
+HTTP_CODE=$(curl -s --connect-timeout 5 "http://${DIRECT_VM_IP}:5000/screenshot" \
+    -o /dev/null -w "%{http_code}" 2>/dev/null)
+if [ "$HTTP_CODE" = "200" ]; then
+    echo "OK"
+else
+    echo "FAIL (HTTP ${HTTP_CODE})"
+    echo "[ERROR] Win7 Flask Server 不可达，请先在 Win7 运行: python D:\python_server\main.py"
+    exit 1
+fi
+
+mkdir -p "${RESULT_DIR}" logs
+
+echo ""
+echo "=== 开始评测 ==="
+echo "  Provider:   direct (无 VM 管理，直连 Flask)"
+echo "  Win7 IP:    ${DIRECT_VM_IP}"
+echo "  Model:      ${MODEL}"
+echo "  Eval:       ${EVAL_MODEL}"
+echo "  Task:       flsol_task4_measure"
+echo "  Obs Type:   ${OBSERVATION_TYPE}  (screenshot only, Win7 a11y unstable)"
+echo "  Max Steps:  ${MAX_STEPS}"
+echo "  Max Tokens: ${MAX_TOKENS}"
+echo "  Results:    ${RESULT_DIR}"
+echo ""
+
+if [ "${INJECT_STEPS}" = true ]; then
+    INJECT_FLAG="--inject_steps"
+else
+    INJECT_FLAG="--no_inject_steps"
+fi
+
+python3 run.py \
+    --provider_name "direct" \
+    --path_to_vm "ignored" \
+    --observation_type "${OBSERVATION_TYPE}" \
+    --action_space "${ACTION_SPACE}" \
+    --model "${MODEL}" \
+    --eval_model "${EVAL_MODEL}" \
+    --temperature "${TEMPERATURE}" \
+    --top_p "${TOP_P}" \
+    --max_tokens "${MAX_TOKENS}" \
+    --max_trajectory_length "${MAX_TRAJECTORY_LENGTH}" \
+    --screen_width "${SCREEN_WIDTH}" \
+    --screen_height "${SCREEN_HEIGHT}" \
+    --sleep_after_execution "${SLEEP_AFTER_EXEC}" \
+    --max_steps "${MAX_STEPS}" \
+    --result_dir "${RESULT_DIR}" \
+    --test_all_meta_path "${TEST_META}" \
+    --domain "${DOMAIN}" \
+    ${INJECT_FLAG}
+
+echo ""
+echo "=== 评测完成 ==="
+echo "结果保存在: ${RESULT_DIR}"