support analyze image
This commit is contained in:
parent
3bca7eca5c
commit
8c5bb34e5a
Binary file not shown.
|
|
@ -41,9 +41,17 @@ class TruthSocialRetriever:
|
|||
os.makedirs(self.save_path, exist_ok=True)
|
||||
|
||||
self.ali_api_key = ALI_API_KEY
|
||||
instruction_file = r"./instructions/media_article_instructions.json"
|
||||
with open(instruction_file, "r", encoding="utf-8") as f:
|
||||
self.instruction = json.load(f)
|
||||
text_instruction_file = r"./instructions/media_article_instructions.json"
|
||||
with open(text_instruction_file, "r", encoding="utf-8") as f:
|
||||
self.text_instruction = json.load(f)
|
||||
|
||||
image_instruction_file = r"./instructions/media_image_instructions.json"
|
||||
with open(image_instruction_file, "r", encoding="utf-8") as f:
|
||||
self.image_instruction = json.load(f)
|
||||
|
||||
image_post_instruction_file = r"./instructions/media_image_post_instructions.json"
|
||||
with open(image_post_instruction_file, "r", encoding="utf-8") as f:
|
||||
self.image_post_instruction = json.load(f)
|
||||
|
||||
def get_user_id_from_page(self, handle="realDonaldTrump"):
|
||||
url = f"https://truthsocial.com/@{handle}"
|
||||
|
|
@ -135,14 +143,6 @@ class TruthSocialRetriever:
|
|||
print("获取帖子失败，请检查 API 密钥或网络。")
|
||||
|
||||
if len(results) > 0:
|
||||
# user_path = os.path.join(self.save_path, user_name)
|
||||
# os.makedirs(user_path, exist_ok=True)
|
||||
# now_date_time = datetime.now().strftime("%Y%m%d%H%M%S")
|
||||
# json_file_name = os.path.join(user_path, f"{user_name}_{now_date_time}.json")
|
||||
# # 将results内容写入json_file_name文件中
|
||||
# with open(json_file_name, 'w', encoding='utf-8') as f:
|
||||
# json.dump(results, f, ensure_ascii=False, indent=2)
|
||||
# logger.info(f"已将{len(results)}条数据保存到: {json_file_name}")
|
||||
result_df = pd.DataFrame(results)
|
||||
result_df = self.remove_duplicate_posts(result_df)
|
||||
|
||||
|
|
@ -214,7 +214,27 @@ class TruthSocialRetriever:
|
|||
text = row["text"]
|
||||
media_thumbnail = row["media_thumbnail"]
|
||||
if media_thumbnail and len(media_thumbnail) > 0:
|
||||
self.wechat.send_image(media_thumbnail)
|
||||
response, image_path, base64_str, md5_str = self.wechat.send_image(media_thumbnail)
|
||||
image_format = "jpg"
|
||||
if image_path is not None and len(image_path) > 0:
|
||||
image_format = image_path.split(".")[-1]
|
||||
if image_format == "jpeg":
|
||||
image_format = "jpg"
|
||||
analysis_result, analysis_token = self.analyze_truth_social_content(
|
||||
text=None,
|
||||
image_stream=base64_str,
|
||||
image_format=image_format,
|
||||
media_type="image"
|
||||
)
|
||||
if analysis_result is not None and len(analysis_result) > 0:
|
||||
result_df.at[index, "analysis_result"] = analysis_result
|
||||
result_df.at[index, "analysis_token"] = analysis_token
|
||||
else:
|
||||
result_df.at[index, "analysis_result"] = ""
|
||||
result_df.at[index, "analysis_token"] = 0
|
||||
analysis_text = f"\n\n## 上述图片分析结果\n\n{analysis_result}"
|
||||
analysis_text += f"\n\n## 上述图片分析token\n\n{analysis_token}"
|
||||
self.wechat.send_markdown(analysis_text)
|
||||
else:
|
||||
contents = []
|
||||
contents.append(f"## 川普推文")
|
||||
|
|
@ -223,7 +243,10 @@ class TruthSocialRetriever:
|
|||
contents.append(date_time)
|
||||
mark_down_text = "\n\n".join(contents)
|
||||
analysis_result, analysis_token = self.analyze_truth_social_content(
|
||||
text
|
||||
text=text,
|
||||
image_stream=None,
|
||||
image_format=None,
|
||||
media_type="text"
|
||||
)
|
||||
result_df.at[index, "analysis_result"] = analysis_result
|
||||
result_df.at[index, "analysis_token"] = analysis_token
|
||||
|
|
@ -253,12 +276,52 @@ class TruthSocialRetriever:
|
|||
def calculate_bytes(self, text: str):
|
||||
return len(text.encode("utf-8"))
|
||||
|
||||
def analyze_truth_social_content(self, text: str):
|
||||
def analyze_truth_social_content(self, text: str, image_stream: str, image_format: str, media_type: str):
|
||||
try:
|
||||
token = 0
|
||||
if media_type == "image":
|
||||
if image_stream is None or len(image_stream) == 0:
|
||||
return "", 0
|
||||
instructions = self.image_instruction.get("Instructions", "")
|
||||
output = self.image_instruction.get("Output", "")
|
||||
prompt = f"# Instructions\n\n{instructions}\n\n# Output\n\n{output}"
|
||||
messages_local = [
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{"image": f"data:image/{image_format};base64,{image_stream}"}, # base64 字符串
|
||||
{"text": prompt} # 你的 prompt
|
||||
]
|
||||
}
|
||||
]
|
||||
response = dashscope.MultiModalConversation.call(
|
||||
api_key=self.ali_api_key,
|
||||
model='qwen-vl-plus',
|
||||
messages=messages_local,
|
||||
)
|
||||
if response.status_code == 200:
|
||||
text = (
|
||||
response.get("output", {})
|
||||
.get("choices", [])[0]
|
||||
.get("message", {})
|
||||
.get("content", "")
|
||||
)
|
||||
token = response.get("usage", {}).get("total_tokens", 0)
|
||||
else:
|
||||
text = f"{response.code} {response.message} 无法分析图片"
|
||||
token = 0
|
||||
|
||||
if text is None or len(text) == 0:
|
||||
return "", 0
|
||||
context = text
|
||||
instructions = self.instruction.get("Instructions", "")
|
||||
output = self.instruction.get("Output", "")
|
||||
prompt = f"# Context\n\n{context}\n\n# Instructions\n\n{instructions}\n\n# Output\n\n{output}"
|
||||
if media_type == "text":
|
||||
instructions = self.text_instruction.get("Instructions", "")
|
||||
output = self.text_instruction.get("Output", "")
|
||||
prompt = f"# Context\n\n{context}\n\n# Instructions\n\n{instructions}\n\n# Output\n\n{output}"
|
||||
else:
|
||||
instructions = self.image_post_instruction.get("Instructions", "")
|
||||
output = self.image_post_instruction.get("Output", "")
|
||||
prompt = f"# Context\n\n{context}\n\n# Instructions\n\n{instructions}\n\n# Output\n\n{output}"
|
||||
response = dashscope.Generation.call(
|
||||
api_key=self.ali_api_key,
|
||||
model="qwen-plus",
|
||||
|
|
@ -276,7 +339,7 @@ class TruthSocialRetriever:
|
|||
.get("message", {})
|
||||
.get("content", "")
|
||||
)
|
||||
token = response.get("usage", {}).get("total_tokens", 0)
|
||||
token += response.get("usage", {}).get("total_tokens", 0)
|
||||
else:
|
||||
response_contents = f"{response.code} {response.message}"
|
||||
token = 0
|
||||
|
|
|
|||
|
|
@ -9,6 +9,8 @@ import requests
|
|||
import base64
|
||||
import hashlib
|
||||
import json
|
||||
import os
|
||||
import time
|
||||
logger = logging.logger
|
||||
|
||||
class Wechat:
|
||||
|
|
@ -17,6 +19,8 @@ class Wechat:
|
|||
# 只要启动代码文件在根目录，config就能找到
|
||||
self.key = key
|
||||
self.url = f"https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key={self.key}"
|
||||
self.image_path = r"./output/wechat/image"
|
||||
os.makedirs(self.image_path, exist_ok=True)
|
||||
|
||||
def send_text(self, text: str):
|
||||
"""
|
||||
|
|
@ -46,12 +50,6 @@ class Wechat:
|
|||
"""
|
||||
发送图片消息
|
||||
"""
|
||||
# data = {
|
||||
# "msgtype": "image",
|
||||
# "image": {"url": image_url}
|
||||
# }
|
||||
# response = requests.post(self.url, json=data)
|
||||
# return response.json()
|
||||
image_bytes = self.download_image(image_url)
|
||||
base64_str, md5_str = self.get_base64_and_md5(image_bytes)
|
||||
data = {
|
||||
|
|
@ -61,9 +59,17 @@ class Wechat:
|
|||
"md5": md5_str,
|
||||
}
|
||||
}
|
||||
# 获取url中图片的名称
|
||||
image_name = image_url.split("/")[-1].split(".")[0]
|
||||
# 获取当前时间，格式为YYYYMMDDHHMMSS
|
||||
now_time = time.strftime("%Y%m%d%H%M%S", time.localtime())
|
||||
image_name = f"{image_name}_{now_time}.jpg"
|
||||
image_path = os.path.join(self.image_path, image_name)
|
||||
with open(image_path, "wb") as f:
|
||||
f.write(image_bytes)
|
||||
response = requests.post(self.url, json=data)
|
||||
response.raise_for_status()
|
||||
return response.json()
|
||||
return response.json(), image_path, base64_str, md5_str
|
||||
|
||||
def download_image(self, image_url):
|
||||
"""下载图片并返回 bytes"""
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
{
|
||||
"Context": "{0}\n\n",
|
||||
"Instructions": "你是一个专业的时政与金融分析师，你的任务是分析推文，结合推文时间（北京时间），联网搜索，并给出分析结果。\n\nContext中的文章，就是特朗普在社交媒体发布的文章，不要怀疑这一点。\n并基于此文章内容进行分析。\n\n要求：\n1. 翻译推文为中文，要求符合中文表达习惯；\n2. 分析推文内容，给出推文的核心观点；\n3. 人物分析：分析推文涉及人物以及人物简介；\n4. 区域分析：包括国家与地区；\n5. 行业以及影响分析；\n6. 经济与金融分析：分析涉及经济与金融影响，包括美股、虚拟货币以及中国A股，并列出最有可能被影响的股票品种或虚拟货币的名称与代码；\n\n",
|
||||
"Instructions": "您是一位资深的国际时事与军事政治评论员与经济、金融分析师，你的任务是分析推文，结合推文时间（北京时间），联网搜索，并给出分析结果。\n\nContext中的文章，就是特朗普在社交媒体发布的文章，不要怀疑这一点。\n并基于此文章内容进行分析。\n\n要求：\n1. 翻译推文为中文，要求符合中文表达习惯；\n2. 分析推文内容，给出推文的核心观点；\n3. 人物分析：分析推文涉及人物以及人物简介；\n4. 区域分析：包括国家与地区；\n5. 行业以及影响分析；\n6. 经济与金融分析：分析涉及经济与金融影响，包括美股、虚拟货币以及中国A股，并列出最有可能被影响的股票品种或虚拟货币的名称与代码；\n\n",
|
||||
"Output": "## 输出要求\n\n除了翻译之外，核心观点+人物分析+区域分析+行业及影响分析+经济与金融分析，不超过1000汉字。\n要求对人名、区域、行业、金融产品、股票代码等专属名词，进行粗体处理。\n\n## 输出格式\n\n### 翻译\n\n### 人物分析\n\n### 区域分析\n\n### 行业及影响分析\n\n### 经济与金融分析\n\n"
|
||||
}
|
||||
|
|
@ -0,0 +1,4 @@
|
|||
{
|
||||
"Instructions": "您是一位资深的国际时事与军事政治评论员与经济、金融分析师，请阅读这张图片，并给出分析结果，分析内容包括：图片文字翻译为中文，将非文字部分的图像，做出图像场景描述。\n\n要求：\n1. 如果图片只有图像信息，请根据图像做出场景描述。\n2. 如果图片既有文字又有图像内容，请提取其中的文字，并将文字翻译为中文，并对非文字部分的图像，做出图像场景描述。\n3. 如果图片中只有文字，请提取其中的文字，并将文字翻译为中文。",
|
||||
"Output": "## 输出要求\n\n图片中的文字原文、文字中文翻译、图片场景描述。\n\n## 输出格式如下：\n\n### 图中文字原文\n\n### 图中文字中文翻译\n\n### 图片场景描述"
|
||||
}
|
||||
|
|
@ -0,0 +1,4 @@
|
|||
{
|
||||
"Instructions": "您是一位资深的国际时事与军事政治评论员与经济、金融分析师，Context的内容是通过图片分析到的信息，你的任务是分析其中的信息，进行联网搜索，并给出分析结果。\n\n该信息，就是特朗普在社交媒体发布的，不要怀疑这一点。\n并基于此文章内容进行分析。\n\n要求：\n1. 分析图片中的文字与图像场景描述，给出推文的核心观点；\n2. 人物分析：分析推文涉及人物以及人物简介；\n3. 区域分析：包括国家与地区；\n4. 行业以及影响分析；\n5. 经济与金融分析：分析涉及经济与金融影响，包括美股、虚拟货币以及中国A股，并列出最有可能被影响的股票品种或虚拟货币的名称与代码；\n\n",
|
||||
"Output": "## 输出要求\n\n要求将Context中的文字原文，中文翻译与图片场景描述，进行原文输出，之外的核心观点+人物分析+区域分析+行业及影响分析+经济与金融分析，不超过1000汉字。\n要求对人名、区域、行业、金融产品、股票代码等专属名词，进行粗体处理。\n\n## 输出格式\n\n### 图中文字原文\n\n### 图中文字中文翻译\n\n### 图片场景描述\n\n### 人物分析\n\n### 区域分析\n\n### 行业及影响分析\n\n### 经济与金融分析\n\n"
|
||||
}
|
||||
Loading…
Reference in New Issue