comi-logic/modalDeploy
7
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

合并分支

Browse Source 
* Merge branch 'main' into cluster-gemini
* FIX Gemini推理配置修复langfuse来源
* PERF Gemini推理配置适配langfuse来源 FIX 时间转换适配更多时间格式
* PERF prompt接入langfuse
* Merge branch 'main' into cluster-gemini
* FIX 修复pydantic导致的问题 PERF 预先使用prompt传输格式要求，如果失败fallback到调用gemini处理格式错误
* Merge branch 'main' into cluster-gemini
* FIX 修复上传问题
* PERF 规范化gemini调用方式, 使用response_schema规范化输出
* PERF 修改gemini upload调用为规范调用
* PERF 优化prompt词提高识别准确率 FIX 修复升级httpx导致部署失败的问题
* Merge branch 'main' into cluster-gemini
* PERF 优化prompt词提高识别准确率
* FIX 修复prompt时间问题
* Merge branch 'main' into cluster-gemini
* PERF 修复时间最大限制问题
* Merge branch 'main' into cluster-gemini
* FIX 修复缩放分辨率计算问题 ADD Gemini推理改为二阶段 FIX 修复时间合并计算问题
* Merge branch 'main' into cluster-gemini
* ADD gemini数据源使用cloud storage

---------

Merge request URL: https://g-ldyi2063.coding.net/p/dev/d/modalDeploy/git/merge/4861
Co-authored-by: 康宇佳
This commit is contained in:
康宇佳
2025-06-23 16:13:05 +08:00
committed by Coding
parent bffd102622
commit 7666102c27
3 changed files with 155 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
src/BowongModalFunctions/models/web_model.py
View File

@@ -1,3 +1,4 @@
import json
import uuid import uuid
from datetime import datetime from datetime import datetime
from enum import Enum from enum import Enum
@@ -652,3 +653,25 @@ class GeminiSecondStageResponseModel(RootModel):
表示一个包含多个商品及其时间线信息的列表。 表示一个包含多个商品及其时间线信息的列表。
""" """
root: List[ProductTimeline] root: List[ProductTimeline]
class GeminiFirstStagePromptVariables(BaseModel):
product_list: List[str] = Field(description="商品名列表")
@computed_field(description="xml格式排列的商品列表")
@property
def product_list_xml(self) -> str:
xml_items = [f" <product>{product}</product>" for product in self.product_list]
xml_string = "\n".join(xml_items)
return f"<products>\n{xml_string}\n </products>"
class GeminiSecondStagePromptVariables(BaseModel):
product_json_list: List[dict] = Field(description="识别出的商品JSON列表")
@computed_field(description="xml格式排列的识别出的商品JSON列表")
@property
def product_json_list_xml(self) -> str:
xml_items = [f" <product>{json.dumps(product,ensure_ascii=False)}</product>" for product in self.product_json_list]
xml_string = "\n".join(xml_items)
return f"<products>\n{xml_string}\n </products>"

16
src/BowongModalFunctions/utils/TimeUtils.py
View File

@@ -26,19 +26,29 @@ def parse_time(time_str):
if re.match(r"^\d{2}:\d{2}:\d{2}\.\d{3}$",time_str): if re.match(r"^\d{2}:\d{2}:\d{2}\.\d{3}$",time_str):
# 先尝试完整格式 HH:MM:SS.fff # 先尝试完整格式 HH:MM:SS.fff
parsed_time = datetime.strptime(time_str, '%H:%M:%S.%f') parsed_time = datetime.strptime(time_str, '%H:%M:%S.%f')
elif re.match(r"^\d{2}:\d{2}:\d{2}:\d{3}$",time_str):
# 如果失败，尝试 HH:MM:SS:fff 格式
parsed_time = datetime.strptime(time_str, '%H:%M:%S:%f')
elif re.match(r"^\d{2}:\d{2}\.\d{3}$",time_str): elif re.match(r"^\d{2}:\d{2}\.\d{3}$",time_str):
# 如果失败，尝试 MM:SS.fff 格式 # 如果失败，尝试 MM:SS.fff 格式
dt = datetime.strptime(time_str, '%M:%S.%f') dt = datetime.strptime(time_str, '%M:%S.%f')
# 将小时设为0，只保留分钟和秒 # 将小时设为0，只保留分钟和秒
parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0)) parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0))
elif re.match(r"^\d{2}\.\d{2}:\d{3}$",time_str):
# 如果失败，尝试 MM.SS:fff 格式
dt = datetime.strptime(time_str, '%M.%S:%f')
# 将小时设为0，只保留分钟和秒
parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0))
elif re.match(r"^\d{2}\.\d{2}\.\d{3}$",time_str):
# 如果失败，尝试 MM.SS.fff 格式
dt = datetime.strptime(time_str, '%M.%S.%f')
# 将小时设为0，只保留分钟和秒
parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0))
elif re.match(r"^\d{2}:\d{2}:\d{3}$",time_str): elif re.match(r"^\d{2}:\d{2}:\d{3}$",time_str):
# 如果失败，尝试 MM:SS:fff 格式 # 如果失败，尝试 MM:SS:fff 格式
dt = datetime.strptime(time_str, '%M:%S:%f') dt = datetime.strptime(time_str, '%M:%S:%f')
# 将小时设为0，只保留分钟和秒 # 将小时设为0，只保留分钟和秒
parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0)) parsed_time = datetime.combine(dt.date(), dt.time().replace(hour=0))
elif re.match(r"^\d{2}:\d{2}:\d{2}:\d{3}$",time_str):
# 如果失败，尝试 HH:MM:SS:fff 格式
parsed_time = datetime.strptime(time_str, '%H:%M:%S:%f')
else: else:
logger.error(f"转换时间格式失败 {time_str}") logger.error(f"转换时间格式失败 {time_str}")
return parsed_time return parsed_time

239
src/cluster/video_apps/hls_slice_inference.py
View File

@@ -3,6 +3,9 @@ import uuid
import modal import modal
from google.genai import types from google.genai import types
from jinja2 import Template
from langfuse._client.client import Langfuse
from langfuse.model import TextPromptClient
from BowongModalFunctions.utils.HTTPUtils import GoogleAuthUtils, FlatJsonSchemaGenerator from BowongModalFunctions.utils.HTTPUtils import GoogleAuthUtils, FlatJsonSchemaGenerator
from ..video import downloader_image, app, config from ..video import downloader_image, app, config
@@ -16,8 +19,8 @@ with downloader_image.imports():
from starlette import status from starlette import status
from BowongModalFunctions.utils.SentryUtils import SentryUtils from BowongModalFunctions.utils.SentryUtils import SentryUtils
from BowongModalFunctions.models.media_model import MediaSource from BowongModalFunctions.models.media_model import MediaSource
from BowongModalFunctions.models.web_model import SentryTransactionInfo, WebhookNotify, \ from BowongModalFunctions.models.web_model import SentryTransactionInfo, WebhookNotify, GeminiFirstStagePromptVariables, \
GeminiFirstStageResponseModel, GeminiSecondStageResponseModel GeminiSecondStagePromptVariables, GeminiFirstStageResponseModel, GeminiSecondStageResponseModel
from BowongModalFunctions.models.ffmpeg_worker_model import FFMpegSliceSegment from BowongModalFunctions.models.ffmpeg_worker_model import FFMpegSliceSegment
from BowongModalFunctions.utils.TimeUtils import TimeDelta, merge_product_data from BowongModalFunctions.utils.TimeUtils import TimeDelta, merge_product_data
from BowongModalFunctions.utils.VideoUtils import FFMPEGSliceOptions, VideoUtils from BowongModalFunctions.utils.VideoUtils import FFMPEGSliceOptions, VideoUtils
@@ -25,7 +28,6 @@ with downloader_image.imports():
@app.function(cpu=(0.5, 64), timeout=1800, @app.function(cpu=(0.5, 64), timeout=1800,
max_containers=config.video_downloader_concurrency, max_containers=config.video_downloader_concurrency,
secrets=[modal.Secret.from_name("gemini-prompt")],
volumes={ volumes={
config.S3_mount_dir: modal.CloudBucketMount( config.S3_mount_dir: modal.CloudBucketMount(
bucket_name=config.S3_bucket_name, bucket_name=config.S3_bucket_name,
@@ -65,6 +67,11 @@ with downloader_image.imports():
"europe-west8", "europe-west8",
] ]
langfuse = Langfuse(host="https://us.cloud.langfuse.com",
secret_key="sk-lf-dd20cb0b-ef2e-49f6-80f0-b2d9cff1bb11",
public_key="pk-lf-15f9d809-0bf6-4a84-ae1c-18f7a7d927c7",
tracing_enabled=False)
logger.info(f"\nmedia || {media.urn}," logger.info(f"\nmedia || {media.urn},"
f"\nstart_time || {start_time}, " f"\nstart_time || {start_time}, "
f"\nend_time || {end_time}, " f"\nend_time || {end_time}, "
@@ -80,10 +87,20 @@ with downloader_image.imports():
regions=gemini_region, access_token=google_api_key, regions=gemini_region, access_token=google_api_key,
) )
# 动态Prompt def split_prompt_config(c:dict):
if "IMAGE_PRODUCT_IDENTIFICATION_PROMPT" in os.environ: return {
IMAGE_PRODUCT_IDENTIFICATION_PROMPT = os.environ["IMAGE_PRODUCT_IDENTIFICATION_PROMPT"] "temperature": c["temperature"],
"top_p": c["top_p"],
"safety_settings": c["safety_settings"]
},c
# 动态Prompt, langfuse获取失败使用默认值
IMAGE_PRODUCT_IDENTIFICATION_PROMPT = langfuse.get_prompt("Gemini自动切条/商品识别", type="text", label="latest")
if IMAGE_PRODUCT_IDENTIFICATION_PROMPT:
prompt_config = IMAGE_PRODUCT_IDENTIFICATION_PROMPT.config
first_stage_generate_config, first_stage_correct_config = split_prompt_config(prompt_config)
else: else:
logger.warning("🥺获取一阶段Prompt失败，使用内置Prompt")
IMAGE_PRODUCT_IDENTIFICATION_PROMPT = """<prompt> IMAGE_PRODUCT_IDENTIFICATION_PROMPT = """<prompt>
<instruction> <instruction>
你是专业的商品识别专家。我上传了商品图片网格，需要你识别图片中的商品并与商品列表进行匹配。 你是专业的商品识别专家。我上传了商品图片网格，需要你识别图片中的商品并与商品列表进行匹配。
@@ -108,32 +125,66 @@ with downloader_image.imports():
- ✅ 输出商品数量不得超过图片中的黑色边框区域数量 - ✅ 输出商品数量不得超过图片中的黑色边框区域数量
**商品列表-标准商品名称**： **商品列表-标准商品名称**：
<items> {{PRODUCT_LIST}}
{PRODUCT_LIST}
</items>
</instruction> </instruction>
<output_format> <output_format>
请按以下JSON格式输出： 请按以下JSON格式输出：
[ [
{{ {
"image_order": 1, "image_order": 1,
"image_name": "图片上显示的原始文字", "image_name": "图片上显示的原始文字",
"matched_product": "匹配到的标准商品名称(必须与商品列表-标准商品名称一致)或null", "matched_product": "匹配到的标准商品名称(必须与商品列表-标准商品名称一致)或null",
"match_confidence": 95, "match_confidence": 95,
"visual_features": {{ "visual_features": {
"color": "商品详细配色等有辨识度的色彩特征", "color": "商品详细配色等有辨识度的色彩特征",
"pattern": "商品详细图案纹理布料等有辨识度的材质特征，材质可根据商品名以及图片综合进行判断", "pattern": "商品详细图案纹理布料等有辨识度的材质特征，材质可根据商品名以及图片综合进行判断",
"style": "商品详细款式风格等有辨识度的款式特征" "style": "商品详细款式风格等有辨识度的款式特征"
}} }
}} }
] ]
</output_format> </output_format>
</prompt>""" </prompt>"""
first_stage_correct_config = {
"temperature": 0.01,
"top_p": 0.7,
"safety_settings": [
{
"category": "HARM_CATEGORY_HARASSMENT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_CIVIC_INTEGRITY",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"threshold": "BLOCK_NONE"
}
],
"response_mime_type": "application/json",
"response_schema": GeminiFirstStageResponseModel.model_json_schema(schema_generator=FlatJsonSchemaGenerator)
}
first_stage_generate_config = {
"temperature": first_stage_correct_config["temperature"],
"top_p": first_stage_correct_config["top_p"],
"safety_settings": first_stage_correct_config["safety_settings"]
}
if "VIDEO_TIMELINE_ANALYSIS_PROMPT" in os.environ: VIDEO_TIMELINE_ANALYSIS_PROMPT = langfuse.get_prompt("Gemini自动切条/视频时间点识别", type="text", label="latest")
VIDEO_TIMELINE_ANALYSIS_PROMPT = os.environ["VIDEO_TIMELINE_ANALYSIS_PROMPT"] if VIDEO_TIMELINE_ANALYSIS_PROMPT:
prompt_config = VIDEO_TIMELINE_ANALYSIS_PROMPT.config
second_stage_generate_config, second_stage_correct_config = split_prompt_config(prompt_config)
else: else:
logger.warning("🥺获取二阶段Prompt失败，使用内置Prompt")
VIDEO_TIMELINE_ANALYSIS_PROMPT = """<prompt> VIDEO_TIMELINE_ANALYSIS_PROMPT = """<prompt>
<instruction> <instruction>
基于已识别的商品清单，分析视频中每个商品的出现时间段。 基于已识别的商品清单，分析视频中每个商品的出现时间段。
@@ -176,24 +227,55 @@ with downloader_image.imports():
- 常见错误：将05:23.500解释为5分钟23.5秒 - 常见错误：将05:23.500解释为5分钟23.5秒
**已识别商品清单(product-商品名 feature-商品特征 feature.color-商品详细颜色配色等色彩特征 feature.pattern-商品详细图案纹理布料等材质特征 feature.style-商品详细款式风格等有辨识度的款式特征)**： **已识别商品清单(product-商品名 feature-商品特征 feature.color-商品详细颜色配色等色彩特征 feature.pattern-商品详细图案纹理布料等材质特征 feature.style-商品详细款式风格等有辨识度的款式特征)**：
<items> {{IDENTIFIED_PRODUCTS}}
{IDENTIFIED_PRODUCTS}
</items>
</instruction> </instruction>
<output_format> <output_format>
请按以下JSON格式输出，确保输出格式复核JSON格式要求： 请按以下JSON格式输出，确保输出格式复核JSON格式要求：
[ [
{{ {
"product": "标准商品名称", "product": "标准商品名称",
"timeline": [ "timeline": [
"01:15.200 - 02:30.800 (穿着本品+介绍本品)", "01:15.200 - 02:30.800 (穿着本品+介绍本品)",
"05:10.100 - 06:25.600 (穿着本品+他品)" "05:10.100 - 06:25.600 (穿着本品+他品)"
] ]
}} }
] ]
</output_format> </output_format>
</prompt>""" </prompt>"""
second_stage_correct_config = {
"temperature": 0.01,
"top_p": 0.7,
"safety_settings": [
{
"category": "HARM_CATEGORY_HARASSMENT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_CIVIC_INTEGRITY",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"threshold": "BLOCK_NONE"
},
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"threshold": "BLOCK_NONE"
}
],
"response_mime_type": "application/json",
"response_schema": GeminiSecondStageResponseModel.model_json_schema(schema_generator=FlatJsonSchemaGenerator)
}
second_stage_generate_config = {
"temperature": second_stage_correct_config["temperature"],
"top_p": second_stage_correct_config["top_p"],
"safety_settings": second_stage_correct_config["safety_settings"]
}
async def upload(file_path): async def upload(file_path):
@@ -215,7 +297,7 @@ with downloader_image.imports():
return upload_response.urn return upload_response.urn
def convert_json(json_like_str:str, json_model): def convert_json(json_like_str:str, correct_config):
try: try:
resp, resp_code = client.generate_content(model_id="gemini-2.5-flash", resp, resp_code = client.generate_content(model_id="gemini-2.5-flash",
contents=[types.Content(role='user', contents=[types.Content(role='user',
@@ -232,35 +314,7 @@ with downloader_image.imports():
] ]
) )
], ],
config=types.GenerateContentConfig( config=types.GenerateContentConfig.model_validate(correct_config), timeout=900)
temperature=0.01,
top_p=0.7,
safety_settings=[
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HARASSMENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_CIVIC_INTEGRITY,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HATE_SPEECH,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
)
],
response_mime_type="application/json",
response_schema=json_model.model_json_schema(
schema_generator=FlatJsonSchemaGenerator)
), timeout=900)
except Exception as e: except Exception as e:
logger.exception(f"😭格式化json请求失败 {e}") logger.exception(f"😭格式化json请求失败 {e}")
return None return None
@@ -282,9 +336,7 @@ with downloader_image.imports():
logger.warning("🥵请求负载过高, 随机更换地区重试") logger.warning("🥵请求负载过高, 随机更换地区重试")
return None return None
def parse_stage1_result(result_text, correct_config):
def parse_stage1_result(result_text):
"""解析第一阶段结果，提取已识别的商品列表""" """解析第一阶段结果，提取已识别的商品列表"""
try: try:
# 清理结果文本，提取JSON部分 # 清理结果文本，提取JSON部分
@@ -298,7 +350,7 @@ with downloader_image.imports():
try: try:
stage1_data = json.loads(clean_result) stage1_data = json.loads(clean_result)
except json.decoder.JSONDecodeError: except json.decoder.JSONDecodeError:
stage1_data = convert_json(clean_result, GeminiFirstStageResponseModel) stage1_data = convert_json(clean_result, correct_config)
if not stage1_data: if not stage1_data:
raise Exception("json为空") raise Exception("json为空")
identified_products = [] identified_products = []
@@ -321,7 +373,6 @@ with downloader_image.imports():
@SentryUtils.sentry_tracker(sentry_trace.x_trace_id, sentry_trace.x_baggage, op="inference_gemini", @SentryUtils.sentry_tracker(sentry_trace.x_trace_id, sentry_trace.x_baggage, op="inference_gemini",
name="Gemini推理", fn_id=current_function_call_id()) name="Gemini推理", fn_id=current_function_call_id())
async def _handler(media: MediaSource, async def _handler(media: MediaSource,
google_api_key: str,
product_grid_list: List[str], product_grid_list: List[str],
product_list: List[any], product_list: List[any],
start_time: str, start_time: str,
@@ -372,9 +423,6 @@ with downloader_image.imports():
async def inference_api_first_stage(): async def inference_api_first_stage():
try: try:
logger.info("🎈开始一阶段推理") logger.info("🎈开始一阶段推理")
product_list_str = ""
for i, product in enumerate(product_title_list):
product_list_str += f"<item>{i}. {product}</item>\n"
image_parts = [] image_parts = []
for i in product_grid_list: for i in product_grid_list:
image_parts.append( image_parts.append(
@@ -387,7 +435,10 @@ with downloader_image.imports():
) )
image_parts.append( image_parts.append(
types.Part.from_text( types.Part.from_text(
text=IMAGE_PRODUCT_IDENTIFICATION_PROMPT.format(PRODUCT_LIST=product_list_str) text=Template(IMAGE_PRODUCT_IDENTIFICATION_PROMPT.prompt
if isinstance(IMAGE_PRODUCT_IDENTIFICATION_PROMPT, TextPromptClient)
else IMAGE_PRODUCT_IDENTIFICATION_PROMPT).render(
PRODUCT_LIST=GeminiFirstStagePromptVariables(product_list=product_title_list).product_list_xml)
) )
) )
resp, resp_code = client.generate_content(model_id="gemini-2.5-flash", resp, resp_code = client.generate_content(model_id="gemini-2.5-flash",
@@ -395,32 +446,7 @@ with downloader_image.imports():
parts=image_parts parts=image_parts
) )
], ],
config=types.GenerateContentConfig( config=types.GenerateContentConfig.model_validate(first_stage_generate_config), timeout=900)
temperature=0.01,
top_p=0.7,
safety_settings=[
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HARASSMENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_CIVIC_INTEGRITY,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HATE_SPEECH,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
)
],
), timeout=900)
except Exception as e: except Exception as e:
logger.exception(f"😭Gemini一阶段推理请求失败: {e}") logger.exception(f"😭Gemini一阶段推理请求失败: {e}")
raise e raise e
@@ -430,7 +456,7 @@ with downloader_image.imports():
if reason == "STOP": if reason == "STOP":
result_text: str = resp.candidates[0].content.parts[0].text result_text: str = resp.candidates[0].content.parts[0].text
# 解析识别结果 # 解析识别结果
identified_products = parse_stage1_result(result_text) identified_products = parse_stage1_result(result_text, first_stage_correct_config)
return identified_products return identified_products
else: else:
logger.error(f"😭Gemini一阶段推理失败, Reason {reason}") logger.error(f"😭Gemini一阶段推理失败, Reason {reason}")
@@ -444,11 +470,6 @@ with downloader_image.imports():
async def inference_api_second_stage(identified_products): async def inference_api_second_stage(identified_products):
try: try:
logger.info("🎈开始二阶段推理") logger.info("🎈开始二阶段推理")
# 构建已识别商品清单字符串
identified_products_str = ""
for i, product in enumerate(identified_products):
product_info_json = json.dumps(product,ensure_ascii=False)
identified_products_str += f"<item>{i}. {product_info_json}<item>\n"
video_parts = [] video_parts = []
video_parts.append( video_parts.append(
types.Part( types.Part(
@@ -460,7 +481,10 @@ with downloader_image.imports():
) )
video_parts.append( video_parts.append(
types.Part.from_text( types.Part.from_text(
text=VIDEO_TIMELINE_ANALYSIS_PROMPT.format(IDENTIFIED_PRODUCTS=identified_products_str) text=Template(VIDEO_TIMELINE_ANALYSIS_PROMPT.prompt
if isinstance(VIDEO_TIMELINE_ANALYSIS_PROMPT, TextPromptClient)
else VIDEO_TIMELINE_ANALYSIS_PROMPT).render(
IDENTIFIED_PRODUCTS=GeminiSecondStagePromptVariables(product_json_list=identified_products).product_json_list_xml)
) )
) )
resp, resp_code = client.generate_content(model_id="gemini-2.5-flash", resp, resp_code = client.generate_content(model_id="gemini-2.5-flash",
@@ -468,32 +492,7 @@ with downloader_image.imports():
parts=video_parts parts=video_parts
) )
], ],
config=types.GenerateContentConfig( config=types.GenerateContentConfig.model_validate(second_stage_generate_config), timeout=900)
temperature=0.01,
top_p=0.7,
safety_settings=[
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HARASSMENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_CIVIC_INTEGRITY,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
),
types.SafetySetting(
category=types.HarmCategory.HARM_CATEGORY_HATE_SPEECH,
threshold=types.HarmBlockThreshold.BLOCK_NONE,
)
],
), timeout=900)
except Exception as e: except Exception as e:
logger.exception(f"😭Gemini二阶段推理请求失败: {e}") logger.exception(f"😭Gemini二阶段推理请求失败: {e}")
raise e raise e
@@ -512,7 +511,7 @@ with downloader_image.imports():
try: try:
parts = json.loads(parts_text) parts = json.loads(parts_text)
except json.decoder.JSONDecodeError: except json.decoder.JSONDecodeError:
parts = convert_json(parts_text, GeminiSecondStageResponseModel) parts = convert_json(parts_text, second_stage_correct_config)
if not parts: if not parts:
raise Exception("json为空") raise Exception("json为空")
logger.info(f"👌合并前 {json.dumps(parts, indent=4, ensure_ascii=False)}") logger.info(f"👌合并前 {json.dumps(parts, indent=4, ensure_ascii=False)}")
@@ -562,5 +561,5 @@ with downloader_image.imports():
logger.exception(f"🥲推理失败, {e}") logger.exception(f"🥲推理失败, {e}")
raise Exception(f"🥲推理失败, {e}") raise Exception(f"🥲推理失败, {e}")
return await _handler(media, google_api_key, product_grid_list, product_list, start_time, end_time, options, return await _handler(media, product_grid_list, product_list, start_time, end_time, options,
sentry_trace, retry_time, scale) sentry_trace, retry_time, scale)