BruceSun
/
MaiBot
mirror of https://github.com/Mai-with-u/MaiBot.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

恢复

pull/1269/head
foxplaying 2025-10-26 15:01:18 +08:00 committed by GitHub
parent d5696c12d4
commit 9662d818a7
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
src/llm_models/model_client/gemini_client.py
View File

@ -250,29 +250,16 @@ def _build_stream_api_resp(
if fr: if fr:
reason = str(fr) reason = str(fr)
break break
if str(reason).endswith("MAX_TOKENS"): if str(reason).endswith("MAX_TOKENS"):
warn_target = "max_tokens"
try:
usage_meta = getattr(last_resp, "usage_metadata", None)
if usage_meta and getattr(usage_meta, "prompt_tokens_details", None):
for detail in usage_meta.prompt_tokens_details:
modality = str(getattr(detail, "modality", "")).upper()
token_count = getattr(detail, "token_count", 0)
if "IMAGE" in modality and token_count > 0:
warn_target = "img_tokens"
break
except Exception as ee:
logger.debug(f"检测 img_tokens 信息失败: {ee}")
has_visible_output = bool(resp.content and resp.content.strip()) has_visible_output = bool(resp.content and resp.content.strip())
if has_visible_output: if has_visible_output:
logger.warning( logger.warning(
f"⚠ Gemini 响应因达到 {warn_target} 限制被部分截断，\n" "⚠ Gemini 响应因达到 max_tokens 限制被部分截断，\n"
f" 可能会对回复内容造成影响，建议修改模型 {warn_target} 配置！" " 可能会对回复内容造成影响，建议修改模型 max_tokens 配置！"
) )
else: else:
logger.warning(f"⚠ Gemini 响应因达到 {warn_target} 限制被截断，\n 请修改模型 {warn_target} 配置！") logger.warning("⚠ Gemini 响应因达到 max_tokens 限制被截断，\n 请修改模型 max_tokens 配置！")
if not resp.content and not resp.tool_calls: if not resp.content and not resp.tool_calls:
if not getattr(resp, "reasoning_content", None): if not getattr(resp, "reasoning_content", None):
@ -393,19 +380,6 @@ def _default_normal_response_parser(
c0 = resp.candidates[0] c0 = resp.candidates[0]
reason = getattr(c0, "finish_reason", None) or getattr(c0, "finishReason", None) reason = getattr(c0, "finish_reason", None) or getattr(c0, "finishReason", None)
if reason and "MAX_TOKENS" in str(reason): if reason and "MAX_TOKENS" in str(reason):
warn_target = "max_tokens"
try:
usage_meta = getattr(resp, "usage_metadata", None)
if usage_meta and getattr(usage_meta, "prompt_tokens_details", None):
for detail in usage_meta.prompt_tokens_details:
modality = str(getattr(detail, "modality", "")).upper()
token_count = getattr(detail, "token_count", 0)
if "IMAGE" in modality and token_count > 0:
warn_target = "img_tokens"
break
except Exception as ee:
logger.debug(f"检测 img_tokens 信息失败: {ee}")
# 检查第二个及之后的 parts 是否有内容 # 检查第二个及之后的 parts 是否有内容
has_real_output = False has_real_output = False
if getattr(c0, "content", None) and getattr(c0.content, "parts", None): if getattr(c0, "content", None) and getattr(c0.content, "parts", None):
@ -419,11 +393,11 @@ def _default_normal_response_parser(
if has_real_output: if has_real_output:
logger.warning( logger.warning(
f"⚠ Gemini 响应因达到 {warn_target} 限制被部分截断，\n" "⚠ Gemini 响应因达到 max_tokens 限制被部分截断，\n"
f" 可能会对回复内容造成影响，建议修改模型 {warn_target} 配置！" " 可能会对回复内容造成影响，建议修改模型 max_tokens 配置！"
) )
else: else:
logger.warning(f"⚠ Gemini 响应因达到 {warn_target} 限制被截断，\n 请修改模型 {warn_target} 配置！") logger.warning("⚠ Gemini 响应因达到 max_tokens 限制被截断，\n 请修改模型 max_tokens 配置！")
return api_response, _usage_record return api_response, _usage_record
except Exception as e: except Exception as e: