fix(ocr+bot): image download race, SenseNova context, log path, encoding

Root causes of the screenshot 404 incident:
1. RACE: client uploads image AND sends message concurrently; bot received
   the message before the upload finished writing, so its GET hit a 404
   error page (<100B treated as failure). FIX: _download_image now retries
   3x with 2s backoff.
2. Zhiwei mentioned tesseract/小果 because the failure text never told her
   the pipeline IS SenseNova. FIX: failure messages now name SenseNova
   explicitly and ask for resend.
3. log_xmpp never worked for the bot: sys.path used relative '../..' from
   a symlinked __file__ which resolved to '/' instead of MoFin root. This
   is why the '最近对话' panel never had bot chat data (only cron script
   entries). FIX: absolute path per red line #7. Verified: test message
   now lands in xmpp_messages.jsonl.
4. My PowerShell -replace corrupted the file encoding (UnicodeDecodeError
   crash loop on restart). Restored from git HEAD and re-applied edits with
   the edit tool. Lesson: never use PowerShell string replace on UTF-8
   source files with Chinese content.
5. functional_health: new sense_ocr module (OCR config presence +
   SenseNova API TCP reachability), no token cost.
This commit is contained in:
hmo
2026-07-20 21:42:59 +08:00
parent 60dbb64f92
commit 299ddc1796
5 changed files with 107 additions and 20 deletions
+38 -20
View File
@@ -73,21 +73,32 @@ def _load_ocr_config():
return None, None, None
def _download_image(url, timeout=30):
"""下载图片字节。失败返回 None。"""
def _download_image(url, timeout=30, retries=3):
"""下载图片字节,带重试。失败返回 None。
重试原因:客户端上传和发消息是并发的,bot 收到消息时上传可能还没写完,
立即 GET 会 404<100 字节错误页)。等 2 秒再试即可。
"""
import urllib.request
import ssl
try:
ctx = ssl.create_default_context()
ctx.check_hostname = False
ctx.verify_mode = ssl.CERT_NONE
req = urllib.request.Request(url, headers={'User-Agent': 'curl/8.5.0'})
with urllib.request.urlopen(req, timeout=timeout, context=ctx) as resp:
data = resp.read()
return data if len(data) > 100 else None
except Exception as e:
log.error(f"图片下载失败 {url[:80]}: {e}")
return None
import time as _t
ctx = ssl.create_default_context()
ctx.check_hostname = False
ctx.verify_mode = ssl.CERT_NONE
for attempt in range(1, retries + 1):
try:
req = urllib.request.Request(url, headers={'User-Agent': 'curl/8.5.0'})
with urllib.request.urlopen(req, timeout=timeout, context=ctx) as resp:
data = resp.read()
if len(data) > 100:
return data
log.warning(f"图片下载第{attempt}次: 响应过小({len(data)}B, 可能上传未完成),重试...")
except Exception as e:
log.warning(f"图片下载第{attempt}次失败: {str(e)[:100]}")
if attempt < retries:
_t.sleep(2)
log.error(f"图片下载最终失败({retries}次尝试): {url}")
return None
def _ocr_image(img_data, prompt=None):
@@ -132,11 +143,15 @@ def _process_image_message(url):
"""下载+OCR 一张截图,返回注入 LLM 的上下文文本。"""
img = _download_image(url)
if not img:
return "[老爸发来一张截图,但下载失败,无法识别]"
return (f"[老爸发来一张截图,bot 下载失败(HTTP错误),无法识别"
f"图片URL: {url}\n"
f"注:bot 的识图管道是 SenseNova(商汤云OCR),不是小果也不是 tesseract。"
f"下载失败可能是 URL 被截断或 ejabberd 上传过期,请老爸重发一次试试。")
ok, text = _ocr_image(img)
if not ok:
return f"[老爸发来一张截图,OCR识别失败: {text}]"
return (f"[老爸发来一张截图,OCR识别内容如下]\n{text}\n"
return (f"[老爸发来一张截图,SenseNova OCR 识别失败: {text}"
f"注:识图管道是 SenseNova,如持续失败请检查 OCR key 配额。")
return (f"[老爸发来一张截图,SenseNova OCR 识别内容如下]\n{text}\n"
f"[截图内容结束] 请基于截图内容回应老爸。")
# ── 全局队列 ──
@@ -149,11 +164,14 @@ RECENT_SENT_MAX = 50
# ── XMPP 消息日志(dashboard 健康Tab「最近对话」数据源)──
try:
sys.path.insert(0, os.path.join(os.path.dirname(os.path.abspath(__file__)), '..', '..'))
sys.path.insert(0, '/home/hmo/MoFin') # 绝对路径(红线#7),相对解析会指到 /
from xmpp_logger import log_xmpp as _log_xmpp
except Exception:
_LOG_XMPP_OK = True
except Exception as _e:
_LOG_XMPP_OK = False
def _log_xmpp(*a, **kw):
pass
print(f"[WARN] xmpp_logger import 失败,消息日志降级为 no-op: {_e}", file=sys.stderr)
def _rs(p):
@@ -289,12 +307,12 @@ class XmppAgent(slixmpp.ClientXMPP):
oob_url = ''
if oob_url and _is_image_url(oob_url):
body = f"[IMAGE] {oob_url}"
log.info(f"📩 收到图片(OOB): from={msg['from']} url={oob_url[:80]}")
log.info(f"📩 收到图片(OOB): from={msg['from']} url={oob_url}")
else:
return
# body 本身是上传 URL(部分客户端把 URL 直接放 body)
elif body.startswith('http') and _is_image_url(body):
log.info(f"📩 收到图片(URL body): from={msg['from']} url={body[:80]}")
log.info(f"📩 收到图片(URL body): from={msg['from']} url={body}")
body = f"[IMAGE] {body}"
log.info(f"📩 收到: from={msg['from']} type={msg['type']} body={body[:60]}")
_log_xmpp('in', sender_str := str(msg['from']), f"{AGENT_NAME}@yoin.fun", body)
@@ -50,6 +50,9 @@ REGISTRY = [
{"module": "gateway_llm", "function": "LLM调用链可用(gateway agent.log)",
"check": {"type": "agent_log", "max_age_min": 30, "when": "always"},
"repair": {"action": "llm_diagnose"}},
{"module": "sense_ocr", "function": "识图服务(SenseNova OCR配置+API可达)",
"check": {"type": "ocr_health", "when": "always"},
"repair": {"action": "llm_diagnose"}},
{"module": "xmpp_bot", "function": "XMPP消息收发(bot journal)",
"check": {"type": "bot_activity", "when": "always"},
"repair": {"action": "llm_diagnose"}},
@@ -127,6 +130,31 @@ def check_bot_activity(chk, now):
return "fail", f"检查失败: {e}"
def check_ocr_health(chk, now):
"""识图服务健康:OCR 配置存在 + SenseNova API 可达(TCP 443)。
不发真实 OCR 请求(省钱),真实端到端验证走 K 测试。"""
import json as _json, socket
# 1. OCR 配置存在且含 key
cfg_path = '/home/hmo/.config/mofin/ocr_config.json'
if not os.path.exists(cfg_path):
return "fail", "OCR 配置文件不存在: /home/hmo/.config/mofin/ocr_config.json"
try:
cfg = _json.load(open(cfg_path))
if not cfg.get('key') or not cfg.get('base_url'):
return "fail", "OCR 配置缺 key 或 base_url"
except Exception as e:
return "fail", f"OCR 配置解析失败: {str(e)[:60]}"
# 2. SenseNova API TCP 可达
try:
host = cfg['base_url'].split('//')[1].split('/')[0].split(':')[0]
port = int(cfg['base_url'].split('//')[1].split('/')[0].split(':')[1]) if ':' in cfg['base_url'].split('//')[1].split('/')[0] else 443
s = socket.create_connection((host, port), timeout=5)
s.close()
except Exception as e:
return "fail", f"SenseNova API 不可达 ({host}): {str(e)[:60]}"
return "ok", f"配置 OK + {host}:{port} 可达"
def check_cron_engine(chk, now):
"""cron 引擎:最近 max_age_min 内是否有任何 job 运行过"""
try:
@@ -180,6 +208,8 @@ def main():
status, reason = check_agent_log(chk, now)
elif t == "bot_activity":
status, reason = check_bot_activity(chk, now)
elif t == "ocr_health":
status, reason = check_ocr_health(chk, now)
elif t == "cron_engine":
status, reason = check_cron_engine(chk, now)
else: