File size: 6,429 Bytes
49654fc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
#!/usr/bin/env python3
"""图片与原始 query → scene(content、intent、note);使用原始 Claude CLI prompt。"""

import argparse
import json
import os
import subprocess
import threading
import time
from concurrent.futures import ThreadPoolExecutor, as_completed

STAGE1_PROMPT = """请观察图片并结合用户的问题,以 JSON 格式回答以下三点:

1. "content":图中是什么东西或什么场景?如果物品的具体类型会影响解决方案,请说清楚(例如螺丝是内六角还是十字的,因为这决定需要什么工具)。若某个关键特征看不清或不确定,明确写出"疑似"或"无法确定",不要臆断。
2. "intent":结合图片和提问,用户想解决什么问题?尽量往搜索商品上靠(如"寻找清洗白色运动鞋的用品");但如果靠不上(比如用户只是想了解、评价、比较、询问某属性),就别硬靠,直接输出原始意图,如"想了解这款衣服的评价如何"。
3. "note":根据识别出的物品/场景和用户意图,指出一句解决该问题时的注意事项,重点是会影响"该选什么商品"的地方。如果没有特别需要注意的,就留空 ""。

示例:
- 菜刀图 + "生锈了怎么处理" →
  {"content": "菜刀,刀面有锈斑", "intent": "寻找给菜刀除锈的工具或用品", "note": "菜刀直接接触食物,除锈处理需满足食品安全要求,不能用工业除锈剂"}
- 内六角螺丝图 + "怎么拧下来" →
  {"content": "内六角螺丝", "intent": "寻找拧下内六角螺丝的工具", "note": "内六角螺丝需要特定规格的工具,规格不匹配无法使用"}
- 免打孔挂杆图 + "厨房墙上挂东西用啥" →
  {"content": "厨房墙面收纳挂杆", "intent": "寻找厨房墙面免打孔收纳挂杆", "note": "需免打孔安装;厨房环境需考虑承重和防油污"}
- 针织衫图 + "这种容易起球吗" →
  {"content": "针织上衣", "intent": "想了解这款衣服是否容易起球", "note": ""}

请直接输出 JSON 对象,不要输出任何其他内容。"""

SYSTEM_PROMPT = "你是一个多模态场景分析助手。用户会提供一张图片的文件路径和一个搜索查询。请先用 Read 工具读取图片,然后根据图片内容和查询回答。只输出 JSON 对象,不要输出任何解释或其他内容。"


def load_input_jsonl(path):
    with open(path, encoding="utf-8") as f:
        rows = [json.loads(line) for line in f if line.strip()]
    return [dict(row, idx=str(row["idx"])) for row in rows]


def load_done_jsonl(path):
    done = set()
    if os.path.exists(path):
        with open(path, encoding="utf-8") as f:
            for line in f:
                line = line.strip()
                if not line:
                    continue
                try:
                    done.add(str(json.loads(line)["idx"]))
                except (json.JSONDecodeError, KeyError):
                    continue
    return done


_write_lock = threading.Lock()


def append_jsonl(path, obj):
    with _write_lock:
        with open(path, "a", encoding="utf-8") as f:
            f.write(json.dumps(obj, ensure_ascii=False) + "\n")
            f.flush()


def extract_json_obj(text):
    import re
    try:
        match = re.search(r"\{.*\}", text, re.DOTALL)
        if match:
            parsed = json.loads(match.group())
            if isinstance(parsed, dict):
                return parsed
    except (json.JSONDecodeError, ValueError):
        pass
    return {"content": "未知", "intent": "未知", "note": "未知", "_raw": text}


def call_claude(image_path, query, model, timeout_seconds):
    user_prompt = f"请先用 Read 工具读取这张图片: {image_path}\n\n用户搜索查询:{query}\n\n{STAGE1_PROMPT}"

    result = subprocess.run(
        [
            "claude", "-p", user_prompt,
            "--system-prompt", SYSTEM_PROMPT,
            "--model", model,
            "--output-format", "text",
            "--allowedTools", "Read",
            "--no-session-persistence",
            "--bare",
        ],
        stdin=subprocess.DEVNULL,
        capture_output=True,
        text=True,
        timeout=timeout_seconds,
    )
    if result.returncode != 0:
        stderr = (result.stderr or "").strip()
        raise RuntimeError(f"claude -p exit {result.returncode}: {stderr}")

    return result.stdout.strip()


def call_claude_with_retry(image_path, query, model, max_retries, timeout_seconds):
    for attempt in range(1, max_retries + 1):
        try:
            raw = call_claude(image_path, query, model, timeout_seconds)
            return raw, None
        except Exception as exc:
            if attempt == max_retries:
                return None, str(exc)
            time.sleep(min(2 ** attempt, 10))
    return None, "max retries exceeded"


def run_stage1(args):
    rows = load_input_jsonl(args.input)
    done = load_done_jsonl(args.output)
    os.makedirs(os.path.dirname(os.path.abspath(args.output)), exist_ok=True)

    def process_one(row):
        image_path = os.path.abspath(os.path.join(args.image_dir, row["image_url"]))
        if not os.path.isfile(image_path):
            raise FileNotFoundError(image_path)
        raw, error = call_claude_with_retry(
            image_path, row["query"], args.model, args.max_retries, args.timeout)
        if error:
            raise RuntimeError(f"idx={row['idx']}: {error}")
        append_jsonl(args.output, {
            **row, "scene": extract_json_obj(raw), "raw_response_s1": raw,
        })

    with ThreadPoolExecutor(max_workers=args.concurrency) as pool:
        futures = [pool.submit(process_one, row) for row in rows if row["idx"] not in done]
        for future in as_completed(futures):
            future.result()
    print(f"意图分析完成:{args.output}")


def main():
    p = argparse.ArgumentParser(description=__doc__)
    p.add_argument("--input", required=True, help="JSONL: idx, query, image_url(本地图片文件名)")
    p.add_argument("--output", required=True)
    p.add_argument("--image-dir", required=True)
    p.add_argument("--model", default="claude-sonnet-4-6")
    p.add_argument("--concurrency", type=int, default=8)
    p.add_argument("--max-retries", type=int, default=3)
    p.add_argument("--timeout", type=int, default=120)
    run_stage1(p.parse_args())


if __name__ == "__main__":
    main()