Spaces:
Paused
Paused
import os | |
import sys | |
import time | |
import gradio as gr | |
import requests | |
from langchain.prompts import ChatPromptTemplate | |
from langchain_community.llms import Ollama | |
import subprocess | |
from datetime import datetime | |
from func_ai import classify_comment, retrieve_from_vdb, VECTOR_API_URL | |
from func_facebook import get_page_id, has_page_replied, get_unanswered_comments, reply_comment, hide_negative_comments | |
def log_message(message): | |
timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S') | |
print(f"[{timestamp}] {message}") | |
# Wait for the server to start | |
time.sleep(10) | |
llm = Ollama(model="llama3.1") | |
log_message("Модель Ollama 'llama3.1' инициализирована.") | |
template = """ | |
You are an assistant answering users' questions using the provided context. Your tasks: | |
1. **Brevity**: Respond concisely, using only relevant information from the context. | |
2. **Politeness**: Start your response with a greeting and maintain a respectful tone. | |
3. **Clarity**: Avoid unnecessary explanations and use simple language. | |
4. **Language of the response**: Detect the language of the user's comment and reply in the same language. | |
5. **Safety**: Do not use phrases like "according to the context" and remove any warnings. | |
6. **Accuracy**: Provide the user with only important and verified purchase links. | |
<context> | |
{context} | |
</context> | |
Question: {input} | |
""" | |
def delete_faiss_index(): | |
log_message("Удаляем FAISS индекс.") | |
response = requests.delete(f"{VECTOR_API_URL}/delete_index/") | |
if response.status_code == 200: | |
log_message("FAISS индекс успешно удален.") | |
return "Faiss успешно удален." | |
else: | |
log_message(f"Ошибка при удалении FAISS индекса: {response.json().get('detail')}") | |
return {"status": "error", "message": response.json().get("detail", "Ошибка при удалении FAISS индекса.")} | |
def upload_file_vdb(file): | |
log_message("Загружаем файл") | |
API_URL = f"{VECTOR_API_URL}/upload/" | |
file_path = file | |
file_name = os.path.basename(file_path) | |
# Открываем файл в бинарном режиме | |
with open(file_path, 'rb') as f: | |
files = {'file': (file_name, f)} | |
response = requests.post(API_URL, files=files) | |
# Обработка ответа от сервера | |
if response.status_code == 200: | |
log_message("Файл успешно загружен.") | |
return "Файл успешно загружен." | |
else: | |
log_message(f"Ошибка при загрузке файла: {response.json().get('detail')}") | |
return f"Ошибка: {response.json().get('detail')}" | |
def generate_response(user_query): | |
log_message(f"Генерация ответа на запрос: {user_query}") | |
prompt = ChatPromptTemplate.from_template(template) | |
documents = retrieve_from_vdb(user_query) | |
context = "\n".join(documents) | |
log_message(f"Контекст из базы данных: {context[:100]}...") | |
full_prompt = prompt.format(context=context, input=user_query) | |
response = llm.invoke(full_prompt) | |
log_message(f"Сгенерированный ответ: {response}") | |
return response | |
def process_comments(ACCESS_TOKEN): | |
log_message("Начинаем процесс скрытия отрицательных комментариев.") | |
hidden_comments_data = hide_negative_comments(ACCESS_TOKEN) | |
log_message(f"Количество постов с скрытыми комментариями: {len(hidden_comments_data)}") | |
log_message("Получение неотвеченных комментариев.") | |
posts_with_unanswered_comments = get_unanswered_comments(ACCESS_TOKEN) | |
page_id = get_page_id(ACCESS_TOKEN) | |
if not page_id: | |
log_message("Не удалось получить ID страницы.") | |
return {"status": "failed", "reason": "Не удалось получить ID страницы."} | |
log_message(f"ID страницы: {page_id}") | |
processed_posts = [] | |
processed_comment_ids = set() # Отслеживание обработанных комментариев | |
for post_data in posts_with_unanswered_comments: | |
post_id = post_data['post_id'] | |
post_message = post_data['post_message'] | |
unanswered_comments = post_data['unanswered_comments'] | |
post_replies = [] | |
for comment in unanswered_comments: | |
comment_id = comment['id'] | |
if comment_id in processed_comment_ids: | |
log_message(f"Комментарий {comment_id} уже обработан. Пропуск.") | |
continue | |
processed_comment_ids.add(comment_id) | |
message = comment['message'] | |
log_message(f"Обработка комментария: {message}") | |
classification = classify_comment(message) | |
log_message(f"Классификация комментария: {classification}") | |
if classification == "interrogative": | |
response_message = generate_response(message) | |
log_message(f"Ответ на комментарий: {response_message}") | |
success = reply_comment(comment_id=comment['id'], message=response_message, token=ACCESS_TOKEN) | |
if success: | |
post_replies.append({ | |
'comment_id': comment['id'], | |
'comment_message': comment['message'], | |
'reply_message': response_message | |
}) | |
processed_posts.append({ | |
'post_id': post_id, | |
'post_message': post_message, | |
'hidden_comments': next((item['hidden_comments'] for item in hidden_comments_data if item['post_id'] == post_id), []), | |
'replies': post_replies | |
}) | |
return { | |
"status": "completed", | |
"posts": processed_posts | |
} | |
with gr.Blocks() as demo: | |
with gr.Tab("Главная страница"): | |
gr.Markdown("# Facebook Comment Filter") | |
token_input = gr.Textbox(label="Access Token") | |
output_main = gr.JSON() | |
process_btn = gr.Button("Процессировать комментарии") | |
process_btn.click(process_comments, inputs=token_input, outputs=output_main) | |
with gr.Tab("Загрузить данные"): | |
gr.Markdown("# Отправь excel файл") | |
file_input = gr.File(label="Загрузите Excel файл (.xlsx)") | |
output_second = gr.Text() | |
second_page_btn = gr.Button("Отправить файл") | |
second_page_btn.click(upload_file_vdb, inputs=file_input, outputs=output_second) | |
delete_btn = gr.Button("Удалить FAISS индекс") | |
delete_btn.click(delete_faiss_index, outputs=output_second) | |
if __name__ == "__main__": | |
demo.launch( | |
debug=True, | |
server_port=7860, | |
server_name="0.0.0.0", | |
) |