add logs and some bugs fixes
This commit is contained in:
queukat
parent
0c4ae18064
commit
525e794f8b
5 changed files with 565 additions and 224 deletions
316
src/gpt.py
316
src/gpt.py
|
|
@ -2,20 +2,21 @@ import json
|
|||
import os
|
||||
import re
|
||||
import textwrap
|
||||
import time
|
||||
from datetime import datetime
|
||||
from typing import Dict, List
|
||||
from functools import wraps
|
||||
from pathlib import Path
|
||||
from typing import Dict, List
|
||||
|
||||
import httpx
|
||||
from Levenshtein import distance
|
||||
from dotenv import load_dotenv
|
||||
from httpx import HTTPStatusError
|
||||
from langchain_core.messages.ai import AIMessage
|
||||
from langchain_core.output_parsers import StrOutputParser
|
||||
from langchain_core.prompt_values import StringPromptValue
|
||||
from langchain_core.prompts import ChatPromptTemplate
|
||||
from langchain_openai import ChatOpenAI
|
||||
from Levenshtein import distance
|
||||
import time
|
||||
from functools import wraps
|
||||
from openai import RateLimitError, OpenAIError, APIError
|
||||
|
||||
|
||||
import src.strings as strings
|
||||
from src.utils import logger
|
||||
|
|
@ -42,156 +43,209 @@ def global_rate_limiter(min_interval):
|
|||
|
||||
return decorator
|
||||
|
||||
def parse_wait_time_from_error_message(error_message: str) -> int:
|
||||
logger.debug("Parsing wait time from error message: %s", error_message)
|
||||
match = re.search(r"Please try again in (\d+)([smhd])", error_message)
|
||||
if match:
|
||||
value, unit = int(match.group(1)), match.group(2)
|
||||
logger.debug("Extracted wait time: %d %s", value, unit)
|
||||
if unit == 's':
|
||||
return value
|
||||
elif unit == 'm':
|
||||
return value * 60
|
||||
elif unit == 'h':
|
||||
return value * 3600
|
||||
elif unit == 'd':
|
||||
return value * 86400
|
||||
logger.debug("Default wait time applied: 30 seconds")
|
||||
return 30
|
||||
|
||||
|
||||
class LLMLogger:
|
||||
|
||||
def __init__(self, llm: ChatOpenAI):
|
||||
logger.debug("Initializing LLMLogger with LLM: %s", llm)
|
||||
self.llm = llm
|
||||
logger.debug("LLMLogger initialized with LLM: %s", llm)
|
||||
logger.debug("LLMLogger successfully initialized with LLM: %s", llm)
|
||||
|
||||
@staticmethod
|
||||
def log_request(prompts, parsed_reply: Dict[str, Dict]):
|
||||
logger.debug("Logging request with prompts: %s", prompts)
|
||||
calls_log = os.path.join(Path("data_folder/output"), "open_ai_calls.json")
|
||||
logger.debug("Starting log_request method")
|
||||
logger.debug("Prompts received: %s", prompts)
|
||||
logger.debug("Parsed reply received: %s", parsed_reply)
|
||||
|
||||
# Определяем путь к файлу для записи логов
|
||||
try:
|
||||
calls_log = os.path.join(Path("data_folder/output"), "open_ai_calls.json")
|
||||
logger.debug("Logging path determined: %s", calls_log)
|
||||
except Exception as e:
|
||||
logger.error("Error determining the log path: %s", str(e))
|
||||
raise
|
||||
|
||||
# Преобразование prompts в текст или словарь
|
||||
if isinstance(prompts, StringPromptValue):
|
||||
logger.debug("Prompts are of type StringPromptValue")
|
||||
prompts = prompts.text
|
||||
logger.debug("Prompts converted to text: %s", prompts)
|
||||
elif isinstance(prompts, Dict):
|
||||
# Convert prompts to a dictionary if they are not in the expected format
|
||||
prompts = {
|
||||
f"prompt_{i+1}": prompt.content
|
||||
for i, prompt in enumerate(prompts.messages)
|
||||
}
|
||||
logger.debug("Prompts are of type Dict")
|
||||
try:
|
||||
prompts = {
|
||||
f"prompt_{i+1}": prompt.content
|
||||
for i, prompt in enumerate(prompts.messages)
|
||||
}
|
||||
logger.debug("Prompts converted to dictionary: %s", prompts)
|
||||
except Exception as e:
|
||||
logger.error("Error converting prompts to dictionary: %s", str(e))
|
||||
raise
|
||||
else:
|
||||
prompts = {
|
||||
f"prompt_{i+1}": prompt.content
|
||||
for i, prompt in enumerate(prompts.messages)
|
||||
logger.debug("Prompts are of unknown type, attempting default conversion")
|
||||
try:
|
||||
prompts = {
|
||||
f"prompt_{i+1}": prompt.content
|
||||
for i, prompt in enumerate(prompts.messages)
|
||||
}
|
||||
logger.debug("Prompts converted to dictionary using default method: %s", prompts)
|
||||
except Exception as e:
|
||||
logger.error("Error converting prompts using default method: %s", str(e))
|
||||
raise
|
||||
|
||||
# Получение текущего времени
|
||||
try:
|
||||
current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
||||
logger.debug("Current time obtained: %s", current_time)
|
||||
except Exception as e:
|
||||
logger.error("Error obtaining current time: %s", str(e))
|
||||
raise
|
||||
|
||||
# Извлечение информации о токенах
|
||||
try:
|
||||
token_usage = parsed_reply["usage_metadata"]
|
||||
output_tokens = token_usage["output_tokens"]
|
||||
input_tokens = token_usage["input_tokens"]
|
||||
total_tokens = token_usage["total_tokens"]
|
||||
logger.debug("Token usage - Input: %d, Output: %d, Total: %d", input_tokens, output_tokens, total_tokens)
|
||||
except KeyError as e:
|
||||
logger.error("KeyError in parsed_reply structure: %s", str(e))
|
||||
raise
|
||||
|
||||
# Извлечение имени модели
|
||||
try:
|
||||
model_name = parsed_reply["response_metadata"]["model_name"]
|
||||
logger.debug("Model name: %s", model_name)
|
||||
except KeyError as e:
|
||||
logger.error("KeyError in response_metadata: %s", str(e))
|
||||
raise
|
||||
|
||||
# Вычисление стоимости использования API
|
||||
try:
|
||||
prompt_price_per_token = 0.00000015
|
||||
completion_price_per_token = 0.0000006
|
||||
total_cost = (input_tokens * prompt_price_per_token) + (output_tokens * completion_price_per_token)
|
||||
logger.debug("Total cost calculated: %f", total_cost)
|
||||
except Exception as e:
|
||||
logger.error("Error calculating total cost: %s", str(e))
|
||||
raise
|
||||
|
||||
# Формирование записи лога
|
||||
try:
|
||||
log_entry = {
|
||||
"model": model_name,
|
||||
"time": current_time,
|
||||
"prompts": prompts,
|
||||
"replies": parsed_reply["content"], # Контент ответа
|
||||
"total_tokens": total_tokens,
|
||||
"input_tokens": input_tokens,
|
||||
"output_tokens": output_tokens,
|
||||
"total_cost": total_cost,
|
||||
}
|
||||
logger.debug("Log entry created: %s", log_entry)
|
||||
except KeyError as e:
|
||||
logger.error("Error creating log entry: missing key %s in parsed_reply", str(e))
|
||||
raise
|
||||
|
||||
current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
||||
logger.debug("Current time: %s", current_time)
|
||||
|
||||
# Extract token usage details from the response
|
||||
token_usage = parsed_reply["usage_metadata"]
|
||||
output_tokens = token_usage["output_tokens"]
|
||||
input_tokens = token_usage["input_tokens"]
|
||||
total_tokens = token_usage["total_tokens"]
|
||||
|
||||
logger.debug("Token usage - Input: %d, Output: %d, Total: %d", input_tokens, output_tokens, total_tokens)
|
||||
|
||||
model_name = parsed_reply["response_metadata"]["model_name"]
|
||||
prompt_price_per_token = 0.00000015
|
||||
completion_price_per_token = 0.0000006
|
||||
|
||||
# Calculate the total cost of the API call
|
||||
total_cost = (input_tokens * prompt_price_per_token) + (
|
||||
output_tokens * completion_price_per_token
|
||||
)
|
||||
|
||||
logger.debug("Total cost calculated: %f", total_cost)
|
||||
|
||||
log_entry = {
|
||||
"model": model_name,
|
||||
"time": current_time,
|
||||
"prompts": prompts,
|
||||
"replies": parsed_reply["content"], # Response content
|
||||
"total_tokens": total_tokens,
|
||||
"input_tokens": input_tokens,
|
||||
"output_tokens": output_tokens,
|
||||
"total_cost": total_cost,
|
||||
}
|
||||
|
||||
logger.debug("Log entry created: %s", log_entry)
|
||||
|
||||
with open(calls_log, "a", encoding="utf-8") as f:
|
||||
json_string = json.dumps(log_entry, ensure_ascii=False, indent=4)
|
||||
f.write(json_string + "\n")
|
||||
logger.debug("Log entry written to file: %s", calls_log)
|
||||
# Запись в файл
|
||||
try:
|
||||
with open(calls_log, "a", encoding="utf-8") as f:
|
||||
json_string = json.dumps(log_entry, ensure_ascii=False, indent=4)
|
||||
f.write(json_string + "\n")
|
||||
logger.debug("Log entry written to file: %s", calls_log)
|
||||
except Exception as e:
|
||||
logger.error("Error writing log entry to file: %s", str(e))
|
||||
raise
|
||||
|
||||
|
||||
class LoggerChatModel:
|
||||
|
||||
def __init__(self, llm: ChatOpenAI):
|
||||
logger.debug("Initializing LoggerChatModel with LLM: %s", llm)
|
||||
self.llm = llm
|
||||
logger.debug("LoggerChatModel initialized with LLM: %s", llm)
|
||||
logger.debug("LoggerChatModel successfully initialized with LLM: %s", llm)
|
||||
|
||||
def __call__(self, messages: List[Dict[str, str]]) -> str:
|
||||
logger.debug("Calling LoggerChatModel with messages: %s", messages)
|
||||
while True:
|
||||
logger.debug("Entering __call__ method with messages: %s", messages)
|
||||
while True: # Бесконечный цикл до успешного выполнения
|
||||
try:
|
||||
# Попытка вызвать модель
|
||||
reply = self.llm(messages)
|
||||
logger.debug("Model reply received: %s", reply)
|
||||
logger.debug("Attempting to call the LLM with messages")
|
||||
reply = self.llm(messages) # Вызов LLM
|
||||
logger.debug("LLM response received: %s", reply)
|
||||
|
||||
parsed_reply = self.parse_llmresult(reply)
|
||||
logger.debug("Parsed LLM reply: %s", parsed_reply)
|
||||
|
||||
# Логируем запрос и ответ
|
||||
LLMLogger.log_request(prompts=messages, parsed_reply=parsed_reply)
|
||||
return reply
|
||||
except RateLimitError as err:
|
||||
# Handle RateLimitError
|
||||
wait_time = self.parse_wait_time_from_error_message(str(err))
|
||||
logger.warning("Rate limit exceeded. Waiting for %d seconds before retrying...", wait_time)
|
||||
time.sleep(wait_time)
|
||||
logger.debug("Request successfully logged")
|
||||
|
||||
return reply # Возвращаем корректный ответ, завершаем цикл
|
||||
|
||||
except httpx.HTTPStatusError as e:
|
||||
logger.error("HTTPStatusError encountered: %s", str(e))
|
||||
if e.response.status_code == 429:
|
||||
retry_after = e.response.headers.get('retry-after')
|
||||
retry_after_ms = e.response.headers.get('retry-after-ms')
|
||||
|
||||
if retry_after:
|
||||
wait_time = int(retry_after)
|
||||
logger.warning("Rate limit exceeded. Waiting for %d seconds before retrying (extracted from 'retry-after' header)...", wait_time)
|
||||
time.sleep(wait_time)
|
||||
elif retry_after_ms:
|
||||
wait_time = int(retry_after_ms) / 1000.0
|
||||
logger.warning("Rate limit exceeded. Waiting for %f seconds before retrying (extracted from 'retry-after-ms' header)...", wait_time)
|
||||
time.sleep(wait_time)
|
||||
else:
|
||||
wait_time = 30 # Время ожидания по умолчанию
|
||||
logger.warning("'retry-after' header not found. Waiting for %d seconds before retrying (default)...", wait_time)
|
||||
time.sleep(wait_time)
|
||||
else:
|
||||
logger.error("HTTP error occurred with status code: %d, waiting 30 seconds before retrying", e.response.status_code)
|
||||
time.sleep(30)
|
||||
|
||||
except Exception as e:
|
||||
logger.error("Unexpected error occurred: %s", str(e))
|
||||
raise
|
||||
logger.info("Waiting for 30 seconds before retrying due to an unexpected error.")
|
||||
time.sleep(30)
|
||||
continue # Продолжаем цикл
|
||||
|
||||
def parse_llmresult(self, llmresult: AIMessage) -> Dict[str, Dict]:
|
||||
logger.debug("Parsing LLM result: %s", llmresult)
|
||||
content = llmresult.content
|
||||
response_metadata = llmresult.response_metadata
|
||||
id_ = llmresult.id
|
||||
usage_metadata = llmresult.usage_metadata
|
||||
parsed_result = {
|
||||
"content": content,
|
||||
"response_metadata": {
|
||||
"model_name": response_metadata.get("model_name", ""),
|
||||
"system_fingerprint": response_metadata.get("system_fingerprint", ""),
|
||||
"finish_reason": response_metadata.get("finish_reason", ""),
|
||||
"logprobs": response_metadata.get("logprobs", None),
|
||||
},
|
||||
"id": id_,
|
||||
"usage_metadata": {
|
||||
"input_tokens": usage_metadata.get("input_tokens", 0),
|
||||
"output_tokens": usage_metadata.get("output_tokens", 0),
|
||||
"total_tokens": usage_metadata.get("total_tokens", 0),
|
||||
},
|
||||
}
|
||||
logger.debug("Parsed LLM result: %s", parsed_result)
|
||||
return parsed_result
|
||||
|
||||
def parse_wait_time_from_error_message(self, error_message: str) -> int:
|
||||
logger.debug("Parsing wait time from error message: %s", error_message)
|
||||
match = re.search(r"Please try again in (\d+)([smhd])", error_message)
|
||||
if match:
|
||||
value, unit = match.groups()
|
||||
value = int(value)
|
||||
logger.debug("Extracted wait time: %d %s", value, unit)
|
||||
if unit == "s":
|
||||
return value
|
||||
elif unit == "m":
|
||||
return value * 60
|
||||
elif unit == "h":
|
||||
return value * 3600
|
||||
elif unit == "d":
|
||||
return value * 86400
|
||||
logger.debug("Default wait time applied: 30 seconds")
|
||||
return 30
|
||||
# Извлечение данных из ответа
|
||||
try:
|
||||
content = llmresult.content
|
||||
response_metadata = llmresult.response_metadata
|
||||
id_ = llmresult.id
|
||||
usage_metadata = llmresult.usage_metadata
|
||||
|
||||
parsed_result = {
|
||||
"content": content,
|
||||
"response_metadata": {
|
||||
"model_name": response_metadata.get("model_name", ""),
|
||||
"system_fingerprint": response_metadata.get("system_fingerprint", ""),
|
||||
"finish_reason": response_metadata.get("finish_reason", ""),
|
||||
"logprobs": response_metadata.get("logprobs", None),
|
||||
},
|
||||
"id": id_,
|
||||
"usage_metadata": {
|
||||
"input_tokens": usage_metadata.get("input_tokens", 0),
|
||||
"output_tokens": usage_metadata.get("output_tokens", 0),
|
||||
"total_tokens": usage_metadata.get("total_tokens", 0),
|
||||
},
|
||||
}
|
||||
|
||||
logger.debug("Parsed LLM result successfully: %s", parsed_result)
|
||||
return parsed_result
|
||||
|
||||
except KeyError as e:
|
||||
logger.error("KeyError while parsing LLM result: missing key %s", str(e))
|
||||
raise # Повторно выбрасываем исключение, чтобы оно обрабатывалось выше
|
||||
|
||||
except Exception as e:
|
||||
logger.error("Unexpected error while parsing LLM result: %s", str(e))
|
||||
raise
|
||||
|
||||
|
||||
|
||||
class GPTAnswerer:
|
||||
|
|
@ -239,7 +293,7 @@ class GPTAnswerer:
|
|||
logger.debug("Setting job application profile: %s", job_application_profile)
|
||||
self.job_application_profile = job_application_profile
|
||||
|
||||
@global_rate_limiter(25)
|
||||
#@global_rate_limiter(25)
|
||||
def summarize_job_description(self, text: str) -> str:
|
||||
logger.debug("Summarizing job description: %s", text)
|
||||
strings.summarize_prompt_template = self._preprocess_template_string(
|
||||
|
|
@ -256,7 +310,7 @@ class GPTAnswerer:
|
|||
prompt = ChatPromptTemplate.from_template(template)
|
||||
return prompt | self.llm_cheap | StrOutputParser()
|
||||
|
||||
@global_rate_limiter(25)
|
||||
#@global_rate_limiter(25)
|
||||
def answer_question_textual_wide_range(self, question: str) -> str:
|
||||
logger.debug("Answering textual question: %s", question)
|
||||
chains = {
|
||||
|
|
@ -384,7 +438,7 @@ class GPTAnswerer:
|
|||
logger.debug("Question answered: %s", output)
|
||||
return output
|
||||
|
||||
@global_rate_limiter(25)
|
||||
#@global_rate_limiter(25)
|
||||
def answer_question_numeric(self, question: str, default_experience: int = 3) -> int:
|
||||
logger.debug("Answering numeric question: %s", question)
|
||||
func_template = self._preprocess_template_string(strings.numeric_question_template)
|
||||
|
|
@ -410,7 +464,7 @@ class GPTAnswerer:
|
|||
logger.error("No numbers found in the string")
|
||||
raise ValueError("No numbers found in the string")
|
||||
|
||||
@global_rate_limiter(25)
|
||||
#@global_rate_limiter(25)
|
||||
def answer_question_from_options(self, question: str, options: list[str]) -> str:
|
||||
logger.debug("Answering question from options: %s", question)
|
||||
func_template = self._preprocess_template_string(strings.options_template)
|
||||
|
|
@ -422,11 +476,11 @@ class GPTAnswerer:
|
|||
logger.debug("Best option determined: %s", best_option)
|
||||
return best_option
|
||||
|
||||
@global_rate_limiter(25)
|
||||
#@global_rate_limiter(25)
|
||||
def resume_or_cover(self, phrase: str) -> str:
|
||||
logger.debug("Determining if phrase refers to resume or cover letter: %s", phrase)
|
||||
prompt_template = """
|
||||
Given the following phrase, respond with only 'resume' if the phrase is about a resume, or 'cover' if it's about a cover letter. Do not provide any additional information or explanations.
|
||||
Given the following phrase, respond with only 'resume' if the phrase is about a resume, or 'cover' if it's about a cover letter. If the phrase contains only the word 'upload', consider it as 'cover'. Do not provide any additional information or explanations.
|
||||
|
||||
phrase: {phrase}
|
||||
"""
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue