mirror of
https://github.com/snailyp/gemini-balance.git
synced 2026-09-05 15:36:45 +08:00
添加API密钥管理、模型服务和安全服务，并优化FastAPI应用程序配置
This commit is contained in:
yinpeng
@@ -0,0 +1,299 @@
|
||||
import httpx
|
||||
import json
|
||||
import time
|
||||
import uuid
|
||||
import logging
|
||||
from typing import Dict, Any, Optional, AsyncGenerator, Union
|
||||
import openai
|
||||
from app.core.config import settings
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class ChatService:
|
||||
def __init__(self, base_url: str, key_manager=None):
|
||||
self.base_url = base_url
|
||||
self.key_manager = key_manager
|
||||
|
||||
def convert_messages_to_gemini_format(self, messages: list) -> list:
|
||||
"""Convert OpenAI message format to Gemini format"""
|
||||
converted_messages = []
|
||||
for msg in messages:
|
||||
role = "user" if msg["role"] == "user" else "model"
|
||||
parts = []
|
||||
|
||||
# 处理文本内容
|
||||
if isinstance(msg["content"], str):
|
||||
parts.append({"text": msg["content"]})
|
||||
# 处理包含图片的消息
|
||||
elif isinstance(msg["content"], list):
|
||||
for content in msg["content"]:
|
||||
if isinstance(content, str):
|
||||
parts.append({"text": content})
|
||||
elif isinstance(content, dict) and content["type"] == "text":
|
||||
parts.append({"text": content["text"]})
|
||||
elif isinstance(content, dict) and content["type"] == "image_url":
|
||||
# 处理图片URL
|
||||
image_url = content["image_url"]["url"]
|
||||
if image_url.startswith("data:image"):
|
||||
# 处理base64图片
|
||||
parts.append(
|
||||
{
|
||||
"inline_data": {
|
||||
"mime_type": "image/jpeg",
|
||||
"data": image_url.split(",")[1],
|
||||
}
|
||||
}
|
||||
)
|
||||
else:
|
||||
# 处理普通URL图片
|
||||
parts.append(
|
||||
{
|
||||
"inline_data": {
|
||||
"mime_type": "image/jpeg",
|
||||
"data": image_url,
|
||||
}
|
||||
}
|
||||
)
|
||||
|
||||
converted_messages.append({"role": role, "parts": parts})
|
||||
|
||||
return converted_messages
|
||||
|
||||
def convert_gemini_response_to_openai(
|
||||
self, response: Dict[str, Any], model: str, stream: bool = False
|
||||
) -> Optional[Dict[str, Any]]:
|
||||
"""Convert Gemini response to OpenAI format"""
|
||||
if stream:
|
||||
if not response.get("candidates"):
|
||||
return None
|
||||
|
||||
try:
|
||||
candidate = response["candidates"][0]
|
||||
content = candidate.get("content", {})
|
||||
parts = content.get("parts", [])
|
||||
|
||||
if not parts:
|
||||
return None
|
||||
|
||||
if "text" in parts[0]:
|
||||
text = parts[0].get("text")
|
||||
elif "executableCode" in parts[0]:
|
||||
text = self.format_code_block(parts[0]["executableCode"])
|
||||
elif "executableCodeResult" in parts[0]:
|
||||
text = self.format_execution_result(parts[0]["executableCodeResult"])
|
||||
else:
|
||||
text = ""
|
||||
|
||||
return {
|
||||
"id": f"chatcmpl-{uuid.uuid4()}",
|
||||
"object": "chat.completion.chunk",
|
||||
"created": int(time.time()),
|
||||
"model": model,
|
||||
"choices": [
|
||||
{
|
||||
"index": 0,
|
||||
"delta": {"content": text},
|
||||
"finish_reason": None,
|
||||
}
|
||||
],
|
||||
}
|
||||
except Exception as e:
|
||||
logger.error(f"Error converting Gemini response: {str(e)}")
|
||||
logger.debug(f"Raw response: {response}")
|
||||
return None
|
||||
else:
|
||||
return {
|
||||
"id": f"chatcmpl-{uuid.uuid4()}",
|
||||
"object": "chat.completion",
|
||||
"created": int(time.time()),
|
||||
"model": model,
|
||||
"choices": [
|
||||
{
|
||||
"index": 0,
|
||||
"message": {
|
||||
"role": "assistant",
|
||||
"content": response["candidates"][0]["content"]["parts"][0][
|
||||
"text"
|
||||
],
|
||||
},
|
||||
"finish_reason": "stop",
|
||||
}
|
||||
],
|
||||
"usage": {
|
||||
"prompt_tokens": 0,
|
||||
"completion_tokens": 0,
|
||||
"total_tokens": 0,
|
||||
},
|
||||
}
|
||||
|
||||
async def create_chat_completion(
|
||||
self,
|
||||
messages: list,
|
||||
model: str,
|
||||
temperature: float,
|
||||
stream: bool,
|
||||
api_key: str,
|
||||
tools: Optional[list] = None,
|
||||
tool_choice: Optional[str] = None,
|
||||
) -> Union[Dict[str, Any], AsyncGenerator[str, None]]:
|
||||
"""Create chat completion using either Gemini or OpenAI API"""
|
||||
|
||||
if tools is None:
|
||||
tools = []
|
||||
if settings.TOOLS_CODE_EXECUTION_ENABLED:
|
||||
tools.append({"code_execution": {}})
|
||||
if model.endswith("-search"):
|
||||
tools.append({"googleSearch": {}})
|
||||
return await self._gemini_chat_completion(
|
||||
messages, model, temperature, stream, api_key, tools
|
||||
)
|
||||
# else:
|
||||
# return await self._openai_chat_completion(
|
||||
# messages, model, temperature, stream, api_key, tools
|
||||
# )
|
||||
|
||||
async def _gemini_chat_completion(
|
||||
self,
|
||||
messages: list,
|
||||
model: str,
|
||||
temperature: float,
|
||||
stream: bool,
|
||||
api_key: str,
|
||||
tools: Optional[list] = None,
|
||||
) -> Union[Dict[str, Any], AsyncGenerator[str, None]]:
|
||||
"""Handle Gemini API chat completion"""
|
||||
if model.endswith("-search"):
|
||||
gemini_model = model[:-7] # Remove -search suffix
|
||||
else:
|
||||
gemini_model = model
|
||||
gemini_messages = self.convert_messages_to_gemini_format(messages)
|
||||
|
||||
payload = {
|
||||
"contents": gemini_messages,
|
||||
"generationConfig": {"temperature": temperature},
|
||||
"tools": tools,
|
||||
}
|
||||
|
||||
if stream:
|
||||
|
||||
async def generate():
|
||||
retries = 0
|
||||
MAX_RETRIES = 3
|
||||
current_api_key = api_key
|
||||
|
||||
while retries < MAX_RETRIES:
|
||||
try:
|
||||
async with httpx.AsyncClient() as client:
|
||||
stream_url = f"https://generativelanguage.googleapis.com/v1beta/models/{gemini_model}:streamGenerateContent?alt=sse&key={current_api_key}"
|
||||
async with client.stream(
|
||||
"POST", stream_url, json=payload
|
||||
) as response:
|
||||
if response.status_code != 200:
|
||||
if retries < MAX_RETRIES - 1:
|
||||
logger.warning(
|
||||
f"API error: {response.status_code}, attempting retry {retries + 1}"
|
||||
)
|
||||
current_api_key = (
|
||||
await self.key_manager.handle_api_failure(
|
||||
current_api_key
|
||||
)
|
||||
)
|
||||
logger.info(
|
||||
f"Switched to new API key: {current_api_key}"
|
||||
)
|
||||
retries += 1
|
||||
continue
|
||||
else:
|
||||
logger.error(
|
||||
f"Max retries reached. Final error: {response.status_code}"
|
||||
)
|
||||
yield f"data: {json.dumps({'error': f'API error: {response.status_code}'})}\n\n"
|
||||
return
|
||||
|
||||
async for line in response.aiter_lines():
|
||||
if line.startswith("data: "):
|
||||
try:
|
||||
chunk = json.loads(line[6:])
|
||||
openai_chunk = (
|
||||
self.convert_gemini_response_to_openai(
|
||||
chunk, model, stream=True
|
||||
)
|
||||
)
|
||||
if openai_chunk:
|
||||
yield f"data: {json.dumps(openai_chunk)}\n\n"
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
yield "data: [DONE]\n\n"
|
||||
return # 成功完成,退出重试循环
|
||||
|
||||
except Exception as e:
|
||||
if retries < MAX_RETRIES - 1:
|
||||
logger.warning(
|
||||
f"Stream error: {str(e)}, attempting retry {retries + 1}"
|
||||
)
|
||||
current_api_key = await self.key_manager.handle_api_failure(
|
||||
current_api_key
|
||||
)
|
||||
retries += 1
|
||||
continue
|
||||
else:
|
||||
logger.error(f"Max retries reached. Final error: {str(e)}")
|
||||
yield f"data: {json.dumps({'error': str(e)})}\n\n"
|
||||
return
|
||||
|
||||
return generate()
|
||||
else:
|
||||
async with httpx.AsyncClient() as client:
|
||||
url = f"https://generativelanguage.googleapis.com/v1beta/models/{gemini_model}:generateContent?key={api_key}"
|
||||
response = await client.post(url, json=payload)
|
||||
gemini_response = response.json()
|
||||
return self.convert_gemini_response_to_openai(gemini_response, model)
|
||||
|
||||
async def _openai_chat_completion(
|
||||
self,
|
||||
messages: list,
|
||||
model: str,
|
||||
temperature: float,
|
||||
stream: bool,
|
||||
api_key: str,
|
||||
tools: Optional[list] = None,
|
||||
) -> Union[Dict[str, Any], AsyncGenerator[str, None]]:
|
||||
"""Handle OpenAI API chat completion"""
|
||||
client = openai.OpenAI(api_key=api_key, base_url=self.base_url)
|
||||
if tools:
|
||||
response = client.chat.completions.create(
|
||||
model=model,
|
||||
messages=messages,
|
||||
temperature=temperature,
|
||||
stream=stream,
|
||||
tools=tools,
|
||||
)
|
||||
else:
|
||||
response = client.chat.completions.create(
|
||||
model=model, messages=messages, temperature=temperature, stream=stream
|
||||
)
|
||||
|
||||
if stream:
|
||||
|
||||
async def generate():
|
||||
for chunk in response:
|
||||
yield f"data: {chunk.model_dump_json()}\n\n"
|
||||
|
||||
return generate()
|
||||
|
||||
return response
|
||||
|
||||
def format_code_block(self, code_data: dict) -> str:
|
||||
"""格式化代码块输出"""
|
||||
language = code_data.get("language", "").lower()
|
||||
code = code_data.get("code", "").strip()
|
||||
|
||||
return f"""\n```{language}\n{code}\n```\n"""
|
||||
|
||||
|
||||
def format_execution_result(result_data: dict) -> str:
|
||||
"""格式化执行结果输出"""
|
||||
outcome = result_data.get("outcome", "")
|
||||
output = result_data.get("output", "").strip()
|
||||
return f"""\n【执行结果】\n{output}\n"""
|
||||
@@ -0,0 +1,22 @@
|
||||
import logging
|
||||
import openai
|
||||
from typing import Union, List, Dict, Any
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class EmbeddingService:
|
||||
def __init__(self, base_url: str):
|
||||
self.base_url = base_url
|
||||
|
||||
async def create_embedding(
|
||||
self, input_text: Union[str, List[str]], model: str, api_key: str
|
||||
) -> Dict[str, Any]:
|
||||
"""Create embeddings using OpenAI API"""
|
||||
try:
|
||||
client = openai.OpenAI(api_key=api_key, base_url=self.base_url)
|
||||
response = client.embeddings.create(input=input_text, model=model)
|
||||
return response
|
||||
except Exception as e:
|
||||
logger.error(f"Error creating embedding: {str(e)}")
|
||||
raise
|
||||
@@ -0,0 +1,57 @@
|
||||
import asyncio
|
||||
from itertools import cycle
|
||||
import logging
|
||||
from typing import Dict
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class KeyManager:
|
||||
def __init__(self, api_keys: list):
|
||||
self.api_keys = api_keys
|
||||
self.key_cycle = cycle(api_keys)
|
||||
self.key_cycle_lock = asyncio.Lock()
|
||||
self.failure_count_lock = asyncio.Lock()
|
||||
self.key_failure_counts: Dict[str, int] = {key: 0 for key in api_keys}
|
||||
self.MAX_FAILURES = 10
|
||||
|
||||
async def get_next_key(self) -> str:
|
||||
"""获取下一个API key"""
|
||||
async with self.key_cycle_lock:
|
||||
return next(self.key_cycle)
|
||||
|
||||
async def is_key_valid(self, key: str) -> bool:
|
||||
"""检查key是否有效"""
|
||||
async with self.failure_count_lock:
|
||||
return self.key_failure_counts[key] < self.MAX_FAILURES
|
||||
|
||||
async def reset_failure_counts(self):
|
||||
"""重置所有key的失败计数"""
|
||||
async with self.failure_count_lock:
|
||||
for key in self.key_failure_counts:
|
||||
self.key_failure_counts[key] = 0
|
||||
|
||||
async def get_next_working_key(self) -> str:
|
||||
"""获取下一个可用的API key"""
|
||||
initial_key = await self.get_next_key()
|
||||
current_key = initial_key
|
||||
|
||||
while True:
|
||||
if await self.is_key_valid(current_key):
|
||||
return current_key
|
||||
|
||||
current_key = await self.get_next_key()
|
||||
if current_key == initial_key:
|
||||
await self.reset_failure_counts()
|
||||
return current_key
|
||||
|
||||
async def handle_api_failure(self, api_key: str) -> str:
|
||||
"""处理API调用失败"""
|
||||
async with self.failure_count_lock:
|
||||
self.key_failure_counts[api_key] += 1
|
||||
if self.key_failure_counts[api_key] >= self.MAX_FAILURES:
|
||||
logger.warning(
|
||||
f"API key {api_key} has failed {self.MAX_FAILURES} times"
|
||||
)
|
||||
|
||||
return await self.get_next_working_key()
|
||||
@@ -0,0 +1,55 @@
|
||||
import requests
|
||||
from datetime import datetime, timezone
|
||||
from typing import Optional, Dict, Any
|
||||
import logging
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class ModelService:
|
||||
def __init__(self, model_search: list):
|
||||
self.model_search = model_search
|
||||
|
||||
def get_gemini_models(self, api_key: str) -> Optional[Dict[str, Any]]:
|
||||
base_url = "https://generativelanguage.googleapis.com/v1beta"
|
||||
url = f"{base_url}/models?key={api_key}"
|
||||
|
||||
try:
|
||||
response = requests.get(url)
|
||||
if response.status_code == 200:
|
||||
gemini_models = response.json()
|
||||
return self.convert_to_openai_models_format(gemini_models)
|
||||
else:
|
||||
logger.error(f"Error: {response.status_code}")
|
||||
logger.error(response.text)
|
||||
return None
|
||||
|
||||
except requests.RequestException as e:
|
||||
logger.error(f"Request failed: {e}")
|
||||
return None
|
||||
|
||||
def convert_to_openai_models_format(
|
||||
self, gemini_models: Dict[str, Any]
|
||||
) -> Dict[str, Any]:
|
||||
openai_format = {"object": "list", "data": []}
|
||||
|
||||
for model in gemini_models.get("models", []):
|
||||
model_id = model["name"].split("/")[-1]
|
||||
openai_model = {
|
||||
"id": model_id,
|
||||
"object": "model",
|
||||
"created": int(datetime.now(timezone.utc).timestamp()),
|
||||
"owned_by": "google",
|
||||
"permission": [],
|
||||
"root": model["name"],
|
||||
"parent": None,
|
||||
"success": True,
|
||||
}
|
||||
openai_format["data"].append(openai_model)
|
||||
|
||||
if model_id in self.model_search:
|
||||
search_model = openai_model.copy()
|
||||
search_model["id"] = f"{model_id}-search"
|
||||
openai_format["data"].append(search_model)
|
||||
|
||||
return openai_format
|
||||
Reference in New Issue
Block a user