Update app.py
Browse files
app.py
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
from flask import Flask, render_template, request, Response, jsonify
|
| 2 |
from mistralai import Mistral
|
| 3 |
import logging
|
| 4 |
import time
|
|
@@ -14,6 +14,7 @@ import random
|
|
| 14 |
|
| 15 |
app = Flask(__name__)
|
| 16 |
app.secret_key = 'super_secret_key'
|
|
|
|
| 17 |
|
| 18 |
message_queue = queue.Queue()
|
| 19 |
|
|
@@ -21,12 +22,12 @@ message_queue = queue.Queue()
|
|
| 21 |
MISTRAL_MODEL = "mistral-large-latest"
|
| 22 |
N_CTX = 32768
|
| 23 |
MAX_RESULTS = 5
|
| 24 |
-
MAX_CONTENT_LENGTH = 10000
|
|
|
|
| 25 |
|
| 26 |
-
#
|
| 27 |
mistral_client = Mistral(api_key=os.getenv("MISTRAL_API_KEY"))
|
| 28 |
|
| 29 |
-
|
| 30 |
SYSTEM_PROMPT = """
|
| 31 |
Ты PrintMaster, сервисный инженер по печатной технике. Правила:
|
| 32 |
1. Формат ответа строго:
|
|
@@ -48,6 +49,14 @@ SYSTEM_PROMPT = """
|
|
| 48 |
6. Будь лаконичен, но технически точен
|
| 49 |
"""
|
| 50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
BLACKLISTED_DOMAINS = [
|
| 52 |
'reddit.com',
|
| 53 |
'stackoverflow.com',
|
|
@@ -91,31 +100,26 @@ def get_random_headers():
|
|
| 91 |
}
|
| 92 |
|
| 93 |
def extract_main_content(html, url):
|
| 94 |
-
"""Извлекает основной контент страницы
|
| 95 |
try:
|
| 96 |
-
# Пробуем trafilatura
|
| 97 |
content = trafilatura.extract(html, include_links=False, include_tables=False)
|
| 98 |
if content and len(content) > 500:
|
| 99 |
return content[:MAX_CONTENT_LENGTH]
|
| 100 |
except Exception as e:
|
| 101 |
logging.error(f"Trafilatura error: {str(e)}")
|
| 102 |
|
| 103 |
-
# Fallback на BeautifulSoup
|
| 104 |
try:
|
| 105 |
soup = BeautifulSoup(html, 'html.parser')
|
| 106 |
|
| 107 |
-
# Удаляем ненужные элементы
|
| 108 |
for element in soup(['script', 'style', 'header', 'footer', 'nav', 'aside', 'form']):
|
| 109 |
element.decompose()
|
| 110 |
|
| 111 |
-
# Пытаемся найти основной контент
|
| 112 |
main_content = soup.find('main') or soup.find('article') or soup.find('div', class_=re.compile('content|main|article|post', re.I))
|
| 113 |
|
| 114 |
if main_content:
|
| 115 |
text = main_content.get_text(separator='\n', strip=True)
|
| 116 |
return text[:MAX_CONTENT_LENGTH] if text else None
|
| 117 |
|
| 118 |
-
# Fallback: весь текст body
|
| 119 |
return soup.body.get_text(separator='\n', strip=True)[:MAX_CONTENT_LENGTH]
|
| 120 |
except Exception as e:
|
| 121 |
logging.error(f"BeautifulSoup error: {str(e)}")
|
|
@@ -210,7 +214,6 @@ def web_search(query: str) -> tuple:
|
|
| 210 |
sources = []
|
| 211 |
full_contents = []
|
| 212 |
|
| 213 |
-
# Обработка featured snippet
|
| 214 |
featured_snippet = data.get("featured_snippet", {})
|
| 215 |
if featured_snippet:
|
| 216 |
snippet = featured_snippet.get("snippet", "")
|
|
@@ -222,9 +225,8 @@ def web_search(query: str) -> tuple:
|
|
| 222 |
"content": snippet
|
| 223 |
})
|
| 224 |
|
| 225 |
-
# Обработка organic results
|
| 226 |
organic_results = data.get("organic_results", [])
|
| 227 |
-
for i, res in enumerate(organic_results[:5]):
|
| 228 |
title = res.get("title", "Без заголовка")
|
| 229 |
link = res.get("link", "#")
|
| 230 |
snippet = res.get("snippet", "") or ""
|
|
@@ -232,7 +234,6 @@ def web_search(query: str) -> tuple:
|
|
| 232 |
if any(domain in link for domain in BLACKLISTED_DOMAINS):
|
| 233 |
continue
|
| 234 |
|
| 235 |
-
# Загрузка полного контента
|
| 236 |
content = None
|
| 237 |
try:
|
| 238 |
headers = get_random_headers()
|
|
@@ -243,13 +244,11 @@ def web_search(query: str) -> tuple:
|
|
| 243 |
logging.error(f"Ошибка загрузки {link}: {str(e)}")
|
| 244 |
|
| 245 |
if not content:
|
| 246 |
-
content = snippet
|
| 247 |
|
| 248 |
-
# Форматирование контента
|
| 249 |
cleaned_content = re.sub(r'\s+', ' ', content).strip()
|
| 250 |
combined_content += f"[[Источник {i+1}]] {title}\n{cleaned_content}\n\n"
|
| 251 |
|
| 252 |
-
# Сохранение источника
|
| 253 |
source_data = {
|
| 254 |
"title": title,
|
| 255 |
"url": link,
|
|
@@ -269,39 +268,20 @@ def web_search(query: str) -> tuple:
|
|
| 269 |
|
| 270 |
|
| 271 |
def clean_response(response: str, sources: list) -> str:
|
| 272 |
-
# Удаление служебных тегов
|
| 273 |
response = re.sub(r'</?assistant>|<\|system\|>|</s>', '', response, flags=re.IGNORECASE)
|
| 274 |
-
|
| 275 |
-
# Удаление лишних разделителей
|
| 276 |
response = re.sub(r'^-{3,}\s*', '', response)
|
| 277 |
-
|
| 278 |
-
# Удаление дублирования разделов
|
| 279 |
response = re.sub(r'(\*\*Проблема:\*\*.+?)(\*\*Проблема:\*\*)', r'\1', response, flags=re.DOTALL)
|
| 280 |
response = re.sub(r'(\*\*Решение:\*\*.+?)(\*\*Решение:\*\*)', r'\1', response, flags=re.DOTALL)
|
| 281 |
-
|
| 282 |
-
# Удаление лишних переносов
|
| 283 |
response = re.sub(r'\n\s*\n', '\n\n', response)
|
| 284 |
response = re.sub(r'[ \t]{2,}', ' ', response)
|
| 285 |
-
|
| 286 |
-
# Удаление начальных фраз
|
| 287 |
response = re.sub(r'^Вот исправленный ответ[^:]+:\s*', '', response)
|
| 288 |
-
|
| 289 |
-
# Форматирование примечаний
|
| 290 |
response = re.sub(r'^---\s*Примечания:\s*', '**Примечания:**\n', response)
|
| 291 |
-
|
| 292 |
-
# Форматирование источников
|
| 293 |
response = re.sub(r'^---\s*Источники:\s*', '**Источники информации:**\n', response)
|
| 294 |
-
|
| 295 |
-
# Удаление лишних маркеров
|
| 296 |
response = re.sub(r'^---\s*', '', response, flags=re.MULTILINE)
|
| 297 |
-
|
| 298 |
-
# Очистка завершающих символов
|
| 299 |
response = re.sub(r'\s*\.{3,}\s*$', '', response)
|
| 300 |
-
|
| 301 |
return response.strip()
|
| 302 |
|
| 303 |
def verify_with_sources(response: str, sources: list) -> str:
|
| 304 |
-
"""Проверяет соответствие ответа источникам с помощью LLM"""
|
| 305 |
try:
|
| 306 |
message_queue.put(('log', "🔍 Проверяю соответствие ответа источникам..."))
|
| 307 |
|
|
@@ -350,37 +330,59 @@ def process_query(prompt: str):
|
|
| 350 |
try:
|
| 351 |
start_time = time.time()
|
| 352 |
message_queue.put(('log', f"👤 Запрос: {prompt}"))
|
| 353 |
-
message_queue.put(('log', f"⚙️ Извлекаю параметры из входящего запроса"))
|
| 354 |
-
|
| 355 |
-
norm_data = generate_search_query(prompt)
|
| 356 |
-
message_queue.put(('log', f"⏏️ Извлечено: {json.dumps(norm_data, ensure_ascii=False)}"))
|
| 357 |
|
| 358 |
-
|
| 359 |
-
|
| 360 |
-
|
| 361 |
-
|
| 362 |
-
|
| 363 |
-
|
| 364 |
-
|
| 365 |
-
|
| 366 |
-
|
| 367 |
-
|
| 368 |
-
|
| 369 |
-
|
| 370 |
-
|
| 371 |
-
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
-
|
| 383 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 384 |
Контекст:
|
| 385 |
Бренд: {norm_data['brand']}
|
| 386 |
Модель: {norm_data['model']}
|
|
@@ -388,13 +390,26 @@ def process_query(prompt: str):
|
|
| 388 |
Суть проблемы (на основе поиска): {extracted_problem}
|
| 389 |
Данные поиска:
|
| 390 |
{search_data}
|
| 391 |
-
"""
|
| 392 |
-
|
| 393 |
-
|
| 394 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 395 |
message_queue.put(('log', "🧠 На основе полученных данных генерирую ответ..."))
|
| 396 |
-
|
| 397 |
-
|
| 398 |
full_response = ""
|
| 399 |
for chunk in mistral_client.chat.stream(
|
| 400 |
model=MISTRAL_MODEL,
|
|
@@ -405,31 +420,45 @@ def process_query(prompt: str):
|
|
| 405 |
if chunk.data.choices[0].delta.content is not None:
|
| 406 |
chunk_text = chunk.data.choices[0].delta.content
|
| 407 |
full_response += chunk_text
|
| 408 |
-
message_queue.put(('response_chunk', chunk_text))
|
| 409 |
|
| 410 |
-
#
|
| 411 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 412 |
|
| 413 |
-
# Очистка и форматирование ответа
|
| 414 |
-
final_response = clean_response(verified_response, sources)
|
| 415 |
-
|
| 416 |
-
message_queue.put(('response_end', final_response))
|
| 417 |
-
message_queue.put(('sources', json.dumps(sources)))
|
| 418 |
-
|
| 419 |
total_time = time.time() - start_time
|
| 420 |
-
message_queue.put(('log', f"💡 Ответ
|
| 421 |
-
message_queue.put(('log', f"⏱ Время: {total_time:.1f}с"))
|
| 422 |
message_queue.put(('done', ''))
|
| 423 |
|
| 424 |
except Exception as e:
|
| 425 |
-
error_msg = f"❌ Ошибка: {str(e)}"
|
| 426 |
message_queue.put(('log', error_msg))
|
| 427 |
-
message_queue.put(('response', "\n⚠️
|
| 428 |
message_queue.put(('done', ''))
|
| 429 |
|
| 430 |
|
| 431 |
@app.route('/')
|
| 432 |
def index():
|
|
|
|
|
|
|
|
|
|
| 433 |
return render_template('index.html')
|
| 434 |
|
| 435 |
|
|
@@ -442,6 +471,25 @@ def ask():
|
|
| 442 |
return jsonify({'status': 'processing'})
|
| 443 |
|
| 444 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 445 |
@app.route('/stream')
|
| 446 |
def stream():
|
| 447 |
def generate():
|
|
|
|
| 1 |
+
from flask import Flask, render_template, request, Response, jsonify, session
|
| 2 |
from mistralai import Mistral
|
| 3 |
import logging
|
| 4 |
import time
|
|
|
|
| 14 |
|
| 15 |
app = Flask(__name__)
|
| 16 |
app.secret_key = 'super_secret_key'
|
| 17 |
+
app.config['SESSION_TYPE'] = 'filesystem'
|
| 18 |
|
| 19 |
message_queue = queue.Queue()
|
| 20 |
|
|
|
|
| 22 |
MISTRAL_MODEL = "mistral-large-latest"
|
| 23 |
N_CTX = 32768
|
| 24 |
MAX_RESULTS = 5
|
| 25 |
+
MAX_CONTENT_LENGTH = 10000
|
| 26 |
+
MAX_HISTORY_LENGTH = 5 # Максимальное количество сообщений в истории
|
| 27 |
|
| 28 |
+
# Клиент Mistral
|
| 29 |
mistral_client = Mistral(api_key=os.getenv("MISTRAL_API_KEY"))
|
| 30 |
|
|
|
|
| 31 |
SYSTEM_PROMPT = """
|
| 32 |
Ты PrintMaster, сервисный инженер по печатной технике. Правила:
|
| 33 |
1. Формат ответа строго:
|
|
|
|
| 49 |
6. Будь лаконичен, но технически точен
|
| 50 |
"""
|
| 51 |
|
| 52 |
+
SYSTEM_PROMPT_CONTINUE = """
|
| 53 |
+
Ты PrintMaster, сервисный инженер. Продолжаешь диалог с пользователем. Правила:
|
| 54 |
+
1. Отвечай на основе предыдущего контекста диалога
|
| 55 |
+
2. Будь лаконичным и технически точным
|
| 56 |
+
3. Если вопрос не связан с предыдущей проблемой, вежливо предложи начать новый диалог
|
| 57 |
+
4. Сохраняй профессиональный стиль общения
|
| 58 |
+
"""
|
| 59 |
+
|
| 60 |
BLACKLISTED_DOMAINS = [
|
| 61 |
'reddit.com',
|
| 62 |
'stackoverflow.com',
|
|
|
|
| 100 |
}
|
| 101 |
|
| 102 |
def extract_main_content(html, url):
|
| 103 |
+
"""Извлекает основной контент страницы"""
|
| 104 |
try:
|
|
|
|
| 105 |
content = trafilatura.extract(html, include_links=False, include_tables=False)
|
| 106 |
if content and len(content) > 500:
|
| 107 |
return content[:MAX_CONTENT_LENGTH]
|
| 108 |
except Exception as e:
|
| 109 |
logging.error(f"Trafilatura error: {str(e)}")
|
| 110 |
|
|
|
|
| 111 |
try:
|
| 112 |
soup = BeautifulSoup(html, 'html.parser')
|
| 113 |
|
|
|
|
| 114 |
for element in soup(['script', 'style', 'header', 'footer', 'nav', 'aside', 'form']):
|
| 115 |
element.decompose()
|
| 116 |
|
|
|
|
| 117 |
main_content = soup.find('main') or soup.find('article') or soup.find('div', class_=re.compile('content|main|article|post', re.I))
|
| 118 |
|
| 119 |
if main_content:
|
| 120 |
text = main_content.get_text(separator='\n', strip=True)
|
| 121 |
return text[:MAX_CONTENT_LENGTH] if text else None
|
| 122 |
|
|
|
|
| 123 |
return soup.body.get_text(separator='\n', strip=True)[:MAX_CONTENT_LENGTH]
|
| 124 |
except Exception as e:
|
| 125 |
logging.error(f"BeautifulSoup error: {str(e)}")
|
|
|
|
| 214 |
sources = []
|
| 215 |
full_contents = []
|
| 216 |
|
|
|
|
| 217 |
featured_snippet = data.get("featured_snippet", {})
|
| 218 |
if featured_snippet:
|
| 219 |
snippet = featured_snippet.get("snippet", "")
|
|
|
|
| 225 |
"content": snippet
|
| 226 |
})
|
| 227 |
|
|
|
|
| 228 |
organic_results = data.get("organic_results", [])
|
| 229 |
+
for i, res in enumerate(organic_results[:5]):
|
| 230 |
title = res.get("title", "Без заголовка")
|
| 231 |
link = res.get("link", "#")
|
| 232 |
snippet = res.get("snippet", "") or ""
|
|
|
|
| 234 |
if any(domain in link for domain in BLACKLISTED_DOMAINS):
|
| 235 |
continue
|
| 236 |
|
|
|
|
| 237 |
content = None
|
| 238 |
try:
|
| 239 |
headers = get_random_headers()
|
|
|
|
| 244 |
logging.error(f"Ошибка загрузки {link}: {str(e)}")
|
| 245 |
|
| 246 |
if not content:
|
| 247 |
+
content = snippet
|
| 248 |
|
|
|
|
| 249 |
cleaned_content = re.sub(r'\s+', ' ', content).strip()
|
| 250 |
combined_content += f"[[Источник {i+1}]] {title}\n{cleaned_content}\n\n"
|
| 251 |
|
|
|
|
| 252 |
source_data = {
|
| 253 |
"title": title,
|
| 254 |
"url": link,
|
|
|
|
| 268 |
|
| 269 |
|
| 270 |
def clean_response(response: str, sources: list) -> str:
|
|
|
|
| 271 |
response = re.sub(r'</?assistant>|<\|system\|>|</s>', '', response, flags=re.IGNORECASE)
|
|
|
|
|
|
|
| 272 |
response = re.sub(r'^-{3,}\s*', '', response)
|
|
|
|
|
|
|
| 273 |
response = re.sub(r'(\*\*Проблема:\*\*.+?)(\*\*Проблема:\*\*)', r'\1', response, flags=re.DOTALL)
|
| 274 |
response = re.sub(r'(\*\*Решение:\*\*.+?)(\*\*Решение:\*\*)', r'\1', response, flags=re.DOTALL)
|
|
|
|
|
|
|
| 275 |
response = re.sub(r'\n\s*\n', '\n\n', response)
|
| 276 |
response = re.sub(r'[ \t]{2,}', ' ', response)
|
|
|
|
|
|
|
| 277 |
response = re.sub(r'^Вот исправленный ответ[^:]+:\s*', '', response)
|
|
|
|
|
|
|
| 278 |
response = re.sub(r'^---\s*Примечания:\s*', '**Примечания:**\n', response)
|
|
|
|
|
|
|
| 279 |
response = re.sub(r'^---\s*Источники:\s*', '**Источники информации:**\n', response)
|
|
|
|
|
|
|
| 280 |
response = re.sub(r'^---\s*', '', response, flags=re.MULTILINE)
|
|
|
|
|
|
|
| 281 |
response = re.sub(r'\s*\.{3,}\s*$', '', response)
|
|
|
|
| 282 |
return response.strip()
|
| 283 |
|
| 284 |
def verify_with_sources(response: str, sources: list) -> str:
|
|
|
|
| 285 |
try:
|
| 286 |
message_queue.put(('log', "🔍 Проверяю соответствие ответа источникам..."))
|
| 287 |
|
|
|
|
| 330 |
try:
|
| 331 |
start_time = time.time()
|
| 332 |
message_queue.put(('log', f"👤 Запрос: {prompt}"))
|
|
|
|
|
|
|
|
|
|
|
|
|
| 333 |
|
| 334 |
+
# Проверяем тип запроса
|
| 335 |
+
if prompt.strip().lower() == "/repeat":
|
| 336 |
+
message_queue.put(('log', "🔄 Пользователь запросил повтор последнего ответа"))
|
| 337 |
+
if 'last_response' in session:
|
| 338 |
+
message_queue.put(('response', session['last_response']))
|
| 339 |
+
message_queue.put(('sources', session.get('last_sources', '')))
|
| 340 |
+
message_queue.put(('done', ''))
|
| 341 |
+
return
|
| 342 |
+
else:
|
| 343 |
+
message_queue.put(('response', "❌ Нет предыдущего ответа для повтора"))
|
| 344 |
+
message_queue.put(('done', ''))
|
| 345 |
+
return
|
| 346 |
+
|
| 347 |
+
# Инициализируем историю диалога
|
| 348 |
+
if 'history' not in session:
|
| 349 |
+
session['history'] = []
|
| 350 |
+
new_dialog = True
|
| 351 |
+
else:
|
| 352 |
+
new_dialog = False
|
| 353 |
+
|
| 354 |
+
# Определяем, нужен ли поиск
|
| 355 |
+
need_search = new_dialog or not any(msg['role'] == 'assistant' for msg in session['history'])
|
| 356 |
+
|
| 357 |
+
# Для новых диалогов или первого сообщения - делаем поиск
|
| 358 |
+
if need_search:
|
| 359 |
+
message_queue.put(('log', "⚙️ Извлекаю параметры из входящего запроса"))
|
| 360 |
+
norm_data = generate_search_query(prompt)
|
| 361 |
+
message_queue.put(('log', f"⏏️ Извлечено: {json.dumps(norm_data, ensure_ascii=False)}"))
|
| 362 |
+
|
| 363 |
+
search_query = norm_data['search_query']
|
| 364 |
+
search_data, sources = web_search(search_query)
|
| 365 |
+
message_queue.put(('log', f"📚 Собрано: {len(search_data)} символов в {len(sources)} источнике(-ах)"))
|
| 366 |
+
|
| 367 |
+
message_queue.put(('log', f"⚙️ Определяю проблему"))
|
| 368 |
+
problem_response = mistral_client.chat.complete(
|
| 369 |
+
model=MISTRAL_MODEL,
|
| 370 |
+
messages=[
|
| 371 |
+
{"role": "system", "content": "Опиши СУТЬ проблемы в одном предложении. Только диагноз, без решений. Не более 12 слов. На русском."},
|
| 372 |
+
{"role": "user", "content": f"Запрос пользователя: {prompt}\nПоисковые данные:\n{search_data}"}
|
| 373 |
+
],
|
| 374 |
+
max_tokens=150,
|
| 375 |
+
temperature=0.2
|
| 376 |
+
)
|
| 377 |
+
extracted_problem = problem_response.choices[0].message.content.strip()
|
| 378 |
+
|
| 379 |
+
if not extracted_problem or len(extracted_problem) < 5:
|
| 380 |
+
extracted_problem = f"Неисправность {norm_data['brand']} {norm_data['model']}"
|
| 381 |
+
|
| 382 |
+
message_queue.put(('log', f"🧩 Определённая проблема: {extracted_problem}"))
|
| 383 |
+
|
| 384 |
+
# Формируем системный промпт с результатами поиска
|
| 385 |
+
system_prompt = SYSTEM_PROMPT + f"""
|
| 386 |
Контекст:
|
| 387 |
Бренд: {norm_data['brand']}
|
| 388 |
Модель: {norm_data['model']}
|
|
|
|
| 390 |
Суть проблемы (на основе поиска): {extracted_problem}
|
| 391 |
Данные поиска:
|
| 392 |
{search_data}
|
| 393 |
+
"""
|
| 394 |
+
else:
|
| 395 |
+
# Для продолжения диалога используем упрощенный системный промпт
|
| 396 |
+
system_prompt = SYSTEM_PROMPT_CONTINUE
|
| 397 |
+
sources = []
|
| 398 |
+
message_queue.put(('log', "💬 Продолжаю диалог на основе истории"))
|
| 399 |
+
|
| 400 |
+
# Собираем все сообщения для модели
|
| 401 |
+
messages = [{"role": "system", "content": system_prompt}]
|
| 402 |
+
|
| 403 |
+
# Добавляем историю диалога (ограниченную по длине)
|
| 404 |
+
history = session['history'][-MAX_HISTORY_LENGTH*2:] # Берем последние N пар сообщений
|
| 405 |
+
messages.extend(history)
|
| 406 |
+
|
| 407 |
+
# Добавляем текущий запрос пользователя
|
| 408 |
+
messages.append({"role": "user", "content": prompt})
|
| 409 |
+
|
| 410 |
message_queue.put(('log', "🧠 На основе полученных данных генерирую ответ..."))
|
| 411 |
+
|
| 412 |
+
# Генерация ответа
|
| 413 |
full_response = ""
|
| 414 |
for chunk in mistral_client.chat.stream(
|
| 415 |
model=MISTRAL_MODEL,
|
|
|
|
| 420 |
if chunk.data.choices[0].delta.content is not None:
|
| 421 |
chunk_text = chunk.data.choices[0].delta.content
|
| 422 |
full_response += chunk_text
|
|
|
|
| 423 |
|
| 424 |
+
# Для новых запросов делаем верификацию
|
| 425 |
+
if need_search and sources:
|
| 426 |
+
verified_response = verify_with_sources(full_response, sources)
|
| 427 |
+
final_response = clean_response(verified_response, sources)
|
| 428 |
+
else:
|
| 429 |
+
final_response = clean_response(full_response, [])
|
| 430 |
+
|
| 431 |
+
# Обновляем историю диалога
|
| 432 |
+
session['history'].append({"role": "user", "content": prompt})
|
| 433 |
+
session['history'].append({"role": "assistant", "content": final_response})
|
| 434 |
+
|
| 435 |
+
# Сохраняем последний ответ для возможного повтора
|
| 436 |
+
session['last_response'] = final_response
|
| 437 |
+
if sources:
|
| 438 |
+
session['last_sources'] = json.dumps(sources)
|
| 439 |
+
|
| 440 |
+
# Отправляем ответ
|
| 441 |
+
message_queue.put(('response', final_response))
|
| 442 |
+
if sources:
|
| 443 |
+
message_queue.put(('sources', json.dumps(sources)))
|
| 444 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 445 |
total_time = time.time() - start_time
|
| 446 |
+
message_queue.put(('log', f"💡 Ответ сгенерирован"))
|
| 447 |
+
message_queue.put(('log', f"⏱ Время обработки: {total_time:.1f}с"))
|
| 448 |
message_queue.put(('done', ''))
|
| 449 |
|
| 450 |
except Exception as e:
|
| 451 |
+
error_msg = f"❌ Ошибка обработки: {str(e)}"
|
| 452 |
message_queue.put(('log', error_msg))
|
| 453 |
+
message_queue.put(('response', "\n⚠️ Произошла ошибка при обработке запроса"))
|
| 454 |
message_queue.put(('done', ''))
|
| 455 |
|
| 456 |
|
| 457 |
@app.route('/')
|
| 458 |
def index():
|
| 459 |
+
# Инициализируем сессию при первом заходе
|
| 460 |
+
if 'history' not in session:
|
| 461 |
+
session['history'] = []
|
| 462 |
return render_template('index.html')
|
| 463 |
|
| 464 |
|
|
|
|
| 471 |
return jsonify({'status': 'processing'})
|
| 472 |
|
| 473 |
|
| 474 |
+
@app.route('/repeat', methods=['POST'])
|
| 475 |
+
def repeat_last():
|
| 476 |
+
"""Повтор последнего ответа"""
|
| 477 |
+
thread = threading.Thread(target=process_query, args=("/repeat",))
|
| 478 |
+
thread.daemon = True
|
| 479 |
+
thread.start()
|
| 480 |
+
return jsonify({'status': 'repeating'})
|
| 481 |
+
|
| 482 |
+
|
| 483 |
+
@app.route('/new', methods=['POST'])
|
| 484 |
+
def new_dialog():
|
| 485 |
+
"""Начало нового диалога"""
|
| 486 |
+
session.pop('history', None)
|
| 487 |
+
session.pop('last_response', None)
|
| 488 |
+
session.pop('last_sources', None)
|
| 489 |
+
session['history'] = []
|
| 490 |
+
return jsonify({'status': 'new_dialog_created'})
|
| 491 |
+
|
| 492 |
+
|
| 493 |
@app.route('/stream')
|
| 494 |
def stream():
|
| 495 |
def generate():
|