diff options
Diffstat (limited to 'main.py')
| -rw-r--r-- | main.py | 113 |
1 files changed, 33 insertions, 80 deletions
@@ -1,94 +1,47 @@ -import os +import concurrent.futures import time -from openai import OpenAI import miniflux from markdownify import markdownify as md -import concurrent.futures - -miniflux_base_url = os.getenv('miniflux_base_url') -miniflux_api_key = os.getenv('miniflux_api_key') -llm_base_url = os.getenv('llm_base_url') -llm_api_key = os.getenv('llm_api_key') -llm_model = os.getenv('llm_model') -feed_blacklist = os.getenv('feed_blacklist').split(',') +from openai import OpenAI +from yaml import load, Loader -miniflux_client = miniflux.Client(miniflux_base_url, api_key=miniflux_api_key) -llm_client = OpenAI(base_url=llm_base_url, api_key=llm_api_key) +config = load(open('config.yml', encoding='utf8'), Loader=Loader) +miniflux_client = miniflux.Client(config['miniflux']['base_url'], api_key=config['miniflux']['api_key']) +llm_client = OpenAI(base_url=config['llm']['base_url'], api_key=config['llm']['api_key']) def process_entry(entry): - if (not entry['content'].startswith('摘要')) & (entry['feed']['site_url'] not in feed_blacklist): - completion = llm_client.chat.completions.create( - model=llm_model, - messages=[ - { - "role": "system", - "content": ( - "You are a highly skilled AI assistant capable of understanding and summarizing complex content from various " - "Your task is to read the provided content, understand the main points, and produce a concise summary in Chinese." - "Limit the summary to 50 words and 2 sentences. Do not add any additional text." - ) - }, - { - "role": "user", - "content": ( - "Summarize the following content in Chinese: 'The latest advancements in AI chip technology have enabled " - "faster processing speeds and lower energy consumption. These innovations are paving the way for more efficient " - "machine learning models, and companies are rapidly adopting these technologies to stay competitive.'" - ) - }, - { - "role": "assistant", - "content": ( - "最新的AI芯片技术取得了突破，使处理速度更快、能耗更低。这些创新为更高效的机器学习模型铺平了道路，企业纷纷采用这些技术以保持竞争力。" - ) - }, - { - "role": "user", - "content": ( - "Summarize the following content in Chinese: 'The government has announced new policies aimed at reducing " - "carbon emissions by 2030. These measures include investing in renewable energy, imposing stricter regulations " - "on industries, and promoting electric vehicles. Experts believe these policies will significantly reduce the " - "country's carbon footprint.'" - ) - }, - { - "role": "assistant", - "content": ( - "政府宣布了到2030年减少碳排放的新政策，包括投资可再生能源、加强行业监管和推广电动汽车。专家认为这些政策将显著减少国家的碳足迹。" - ) - }, - { - "role": "user", - "content": ( - "Summarize the following content in Chinese: 'Participants are debating the pros and cons of remote work. " - "Some argue that it increases productivity and work-life balance, while others believe it leads to isolation and " - "decreased collaboration. Overall, the consensus is that remote work is beneficial if managed properly.'" - ) - }, - { - "role": "assistant", - "content": ( - "论坛讨论了远程工作的利弊。有人认为它提高了生产力和平衡了工作与生活，有人则认为它导致孤立和减少了协作。总体而言，大家认为远程工作在管理得当的情况下是有益的。" - ) - }, - { - "role": "user", - "content": ( - "Summarize the following content in Chinese: '" + md(entry['content']) + "'" - ) - } - ] - ) - llm_result = completion.choices[0].message.content - print(llm_result) - miniflux_client.update_entry(entry['id'], content='摘要：' + llm_result + '<hr><br />' + entry['content']) - return None + llm_result = '' + start_with_list = [name[1]['title'] for name in config['agents'].items()] + style_block = [name[1]['style_block'] for name in config['agents'].items()] + [start_with_list.append('<pre') for i in style_block if i] + + for agent in config['agents'].items(): + messages = [{"role": "system", "content": agent[1]['prompt']}, {"role": "user", "content": "The following is the input content:\n---\n " + md(entry['content'])}] + # filter, if AI is not generating, and in whitelist, or not in blacklist + if ((not entry['content'].startswith(tuple(start_with_list))) and + (((agent[1]['whitelist'] is not None) and (entry['feed']['site_url'] in agent[1]['whitelist'])) or + (agent[1]['blacklist'] is not None and entry['feed']['site_url'] not in agent[1]['blacklist']) or + (agent[1]['whitelist'] is None and agent[1]['blacklist'] is None))): + completion = llm_client.chat.completions.create( model=config['llm']['model'], messages= messages ) + response_content = completion.choices[0].message.content + print(time.strftime("%Y/%m/%d %H:%M:%S", time.localtime()), agent[0], entry['feed']['feed_url'], response_content) + + if agent[1]['style_block']: + llm_result = llm_result + f"<pre style=\"white-space: pre-wrap;\"><code>\n{agent[1]['title']}：{response_content}</code></pre><hr><br />" + else: + llm_result = llm_result + f"{agent[1]['title']}：{response_content}<hr><br />" + + if len(llm_result) > 0: + miniflux_client.update_entry(entry['id'], content= llm_result + entry['content']) while True: # Fetch entries with status unread entries = miniflux_client.get_entries(status=['unread'], limit=10000) - + print(time.strftime("%Y/%m/%d %H:%M:%S", time.localtime()), 'Fetched new entries: ' + str(len(entries['entries']))) if len(entries['entries']) > 0 else print(time.strftime("%Y/%m/%d %H:%M:%S", time.localtime()), 'No new entries') with concurrent.futures.ThreadPoolExecutor() as executor: futures = [executor.submit(process_entry, i) for i in entries['entries']] + + if len(entries['entries']) > 0: + print(time.strftime("%Y/%m/%d %H:%M:%S", time.localtime()), 'Done') time.sleep(60)
\ No newline at end of file |