feat:添加了一个异步HTTP接口的gemini客户端

This commit is contained in:
minecraft1024a
2025-08-12 15:56:12 +08:00
parent 9530dda6ec
commit 837631d1aa

View File

@@ -0,0 +1,555 @@
import asyncio
import json
import io
import base64
from typing import Callable, Any, Coroutine, Optional, List, AsyncIterator
import aiohttp
from json_repair import repair_json
from src.config.api_ada_configs import ModelInfo, APIProvider
from src.common.logger import get_logger
from .base_client import APIResponse, UsageRecord, BaseClient, client_registry
from ..exceptions import (
RespParseException,
NetworkConnectionError,
RespNotOkException,
ReqAbortException,
)
from ..payload_content.message import Message, RoleType
from ..payload_content.resp_format import RespFormat, RespFormatType
from ..payload_content.tool_option import ToolOption, ToolParam, ToolCall
logger = get_logger("AioHTTP-Gemini客户端")
def _convert_messages(messages: list[Message]) -> tuple[list[dict], list[str] | None]:
"""
转换消息格式 - 将消息转换为Gemini REST API所需的格式
:param messages: 消息列表
:return: (contents, system_instructions)
"""
def _convert_message_item(message: Message) -> dict:
"""转换单个消息格式"""
# 转换角色名称
if message.role == RoleType.Assistant:
role = "model"
elif message.role == RoleType.User:
role = "user"
else:
raise ValueError(f"不支持的消息角色: {message.role}")
# 转换内容
parts = []
if isinstance(message.content, str):
parts.append({"text": message.content})
elif isinstance(message.content, list):
for item in message.content:
if isinstance(item, tuple): # (format, base64_data)
parts.append({
"inline_data": {
"mime_type": f"image/{item[0].lower()}",
"data": item[1]
}
})
elif isinstance(item, str):
parts.append({"text": item})
else:
raise RuntimeError("无法触及的代码请使用MessageBuilder类构建消息对象")
return {
"role": role,
"parts": parts
}
contents = []
system_instructions = []
for message in messages:
if message.role == RoleType.System:
if isinstance(message.content, str):
system_instructions.append(message.content)
else:
raise ValueError("System消息不支持非文本内容")
elif message.role == RoleType.Tool:
# 工具调用结果处理
if not message.tool_call_id:
raise ValueError("工具调用消息缺少tool_call_id")
contents.append({
"role": "function",
"parts": [{"text": str(message.content)}]
})
else:
contents.append(_convert_message_item(message))
return contents, system_instructions if system_instructions else None
def _convert_tool_options(tool_options: list[ToolOption]) -> list[dict]:
"""
转换工具选项格式 - 将工具选项转换为Gemini REST API所需的格式
"""
def _convert_tool_param(param: ToolParam) -> dict:
"""转换工具参数"""
result = {
"type": param.param_type.value,
"description": param.description,
}
if param.enum_values:
result["enum"] = param.enum_values
return result
def _convert_tool_option_item(tool_option: ToolOption) -> dict:
"""转换单个工具选项"""
function_declaration = {
"name": tool_option.name,
"description": tool_option.description,
}
if tool_option.params:
function_declaration["parameters"] = {
"type": "object",
"properties": {
param.name: _convert_tool_param(param)
for param in tool_option.params
},
"required": [
param.name
for param in tool_option.params
if param.required
],
}
return {
"function_declarations": [function_declaration]
}
return [_convert_tool_option_item(tool_option) for tool_option in tool_options]
def _build_generation_config(
max_tokens: int,
temperature: float,
response_format: RespFormat | None = None,
extra_params: dict | None = None
) -> dict:
"""构建生成配置"""
config = {
"maxOutputTokens": max_tokens,
"temperature": temperature,
}
# 处理响应格式
if response_format:
if response_format.format_type == RespFormatType.JSON_OBJ:
config["responseMimeType"] = "application/json"
elif response_format.format_type == RespFormatType.JSON_SCHEMA:
config["responseMimeType"] = "application/json"
config["responseSchema"] = response_format.to_dict()
# 合并额外参数
if extra_params:
config.update(extra_params)
return config
class AiohttpGeminiStreamParser:
"""流式响应解析器"""
def __init__(self):
self.content_buffer = io.StringIO()
self.reasoning_buffer = io.StringIO()
self.tool_calls_buffer = []
self.usage_record = None
def parse_chunk(self, chunk_text: str):
"""解析单个流式数据块"""
try:
if not chunk_text.strip():
return
# 移除data:前缀
if chunk_text.startswith("data: "):
chunk_text = chunk_text[6:].strip()
if chunk_text == "[DONE]":
return
chunk_data = json.loads(chunk_text)
# 解析候选项
if "candidates" in chunk_data and chunk_data["candidates"]:
candidate = chunk_data["candidates"][0]
# 解析内容
if "content" in candidate and "parts" in candidate["content"]:
for part in candidate["content"]["parts"]:
if "text" in part:
self.content_buffer.write(part["text"])
# 解析工具调用
if "functionCall" in candidate:
func_call = candidate["functionCall"]
call_id = f"gemini_call_{len(self.tool_calls_buffer)}"
self.tool_calls_buffer.append({
"id": call_id,
"name": func_call.get("name", ""),
"args": func_call.get("args", {})
})
# 解析使用统计
if "usageMetadata" in chunk_data:
usage = chunk_data["usageMetadata"]
self.usage_record = (
usage.get("promptTokenCount", 0),
usage.get("candidatesTokenCount", 0),
usage.get("totalTokenCount", 0)
)
except json.JSONDecodeError as e:
logger.warning(f"解析流式数据块失败: {e}, 数据: {chunk_text}")
except Exception as e:
logger.error(f"处理流式数据块时出错: {e}")
def get_response(self) -> APIResponse:
"""获取最终响应"""
response = APIResponse()
if self.content_buffer.tell() > 0:
response.content = self.content_buffer.getvalue()
if self.reasoning_buffer.tell() > 0:
response.reasoning_content = self.reasoning_buffer.getvalue()
if self.tool_calls_buffer:
response.tool_calls = []
for call_data in self.tool_calls_buffer:
response.tool_calls.append(ToolCall(
call_data["id"],
call_data["name"],
call_data["args"]
))
# 清理缓冲区
self.content_buffer.close()
self.reasoning_buffer.close()
return response
async def _default_stream_response_handler(
response: aiohttp.ClientResponse,
interrupt_flag: asyncio.Event | None,
) -> tuple[APIResponse, Optional[tuple[int, int, int]]]:
"""默认流式响应处理器"""
parser = AiohttpGeminiStreamParser()
try:
async for line in response.content:
if interrupt_flag and interrupt_flag.is_set():
raise ReqAbortException("请求被外部信号中断")
line_text = line.decode('utf-8').strip()
if line_text:
parser.parse_chunk(line_text)
api_response = parser.get_response()
return api_response, parser.usage_record
except Exception as e:
if not isinstance(e, ReqAbortException):
raise RespParseException(None, f"流式响应解析失败: {e}") from e
raise
def _default_normal_response_parser(
response_data: dict,
) -> tuple[APIResponse, Optional[tuple[int, int, int]]]:
"""默认普通响应解析器"""
api_response = APIResponse()
try:
# 解析候选项
if "candidates" in response_data and response_data["candidates"]:
candidate = response_data["candidates"][0]
# 解析文本内容
if "content" in candidate and "parts" in candidate["content"]:
content_parts = []
for part in candidate["content"]["parts"]:
if "text" in part:
content_parts.append(part["text"])
if content_parts:
api_response.content = "".join(content_parts)
# 解析工具调用
if "functionCall" in candidate:
func_call = candidate["functionCall"]
api_response.tool_calls = [ToolCall(
f"gemini_call_0",
func_call.get("name", ""),
func_call.get("args", {})
)]
# 解析使用统计
usage_record = None
if "usageMetadata" in response_data:
usage = response_data["usageMetadata"]
usage_record = (
usage.get("promptTokenCount", 0),
usage.get("candidatesTokenCount", 0),
usage.get("totalTokenCount", 0)
)
api_response.raw_data = response_data
return api_response, usage_record
except Exception as e:
raise RespParseException(response_data, f"响应解析失败: {e}") from e
@client_registry.register_client_class("aiohttp_gemini")
class AiohttpGeminiClient(BaseClient):
"""使用aiohttp的Gemini客户端"""
def __init__(self, api_provider: APIProvider):
super().__init__(api_provider)
self.base_url = "https://generativelanguage.googleapis.com/v1beta"
self.session: aiohttp.ClientSession | None = None
self.api_key = api_provider.api_key
# 如果提供了自定义base_url使用它
if api_provider.base_url:
self.base_url = api_provider.base_url.rstrip('/')
async def _get_session(self) -> aiohttp.ClientSession:
"""获取或创建aiohttp会话"""
if self.session is None or self.session.closed:
timeout = aiohttp.ClientTimeout(total=300) # 5分钟超时
self.session = aiohttp.ClientSession(
timeout=timeout,
headers={
"Content-Type": "application/json",
"User-Agent": "MMC-AioHTTP-Gemini-Client/1.0"
}
)
return self.session
async def _make_request(
self,
method: str,
endpoint: str,
data: dict | None = None,
stream: bool = False
) -> aiohttp.ClientResponse:
"""发起HTTP请求"""
session = await self._get_session()
url = f"{self.base_url}/{endpoint}?key={self.api_key}"
try:
if method.upper() == "POST":
response = await session.post(
url,
json=data,
headers={"Accept": "text/event-stream" if stream else "application/json"}
)
else:
response = await session.get(url)
# 检查HTTP状态码
if response.status >= 400:
error_text = await response.text()
raise RespNotOkException(response.status, error_text)
return response
except aiohttp.ClientError as e:
raise NetworkConnectionError() from e
async def get_response(
self,
model_info: ModelInfo,
message_list: list[Message],
tool_options: list[ToolOption] | None = None,
max_tokens: int = 1024,
temperature: float = 0.7,
response_format: RespFormat | None = None,
stream_response_handler: Optional[
Callable[
[aiohttp.ClientResponse, asyncio.Event | None],
Coroutine[Any, Any, tuple[APIResponse, Optional[tuple[int, int, int]]]],
]
] = None,
async_response_parser: Optional[
Callable[[dict], tuple[APIResponse, Optional[tuple[int, int, int]]]]
] = None,
interrupt_flag: asyncio.Event | None = None,
extra_params: dict[str, Any] | None = None,
) -> APIResponse:
"""
获取对话响应
"""
if stream_response_handler is None:
stream_response_handler = _default_stream_response_handler
if async_response_parser is None:
async_response_parser = _default_normal_response_parser
# 转换消息格式
contents, system_instructions = _convert_messages(message_list)
# 构建请求体
request_data = {
"contents": contents,
"generationConfig": _build_generation_config(
max_tokens, temperature, response_format, extra_params
)
}
# 添加系统指令
if system_instructions:
request_data["systemInstruction"] = {
"parts": [{"text": instr} for instr in system_instructions]
}
# 添加工具定义
if tool_options:
request_data["tools"] = _convert_tool_options(tool_options)
try:
if model_info.force_stream_mode:
# 流式请求
endpoint = f"models/{model_info.model_identifier}:streamGenerateContent"
req_task = asyncio.create_task(
self._make_request("POST", endpoint, request_data, stream=True)
)
while not req_task.done():
if interrupt_flag and interrupt_flag.is_set():
req_task.cancel()
raise ReqAbortException("请求被外部信号中断")
await asyncio.sleep(0.1)
response = req_task.result()
api_response, usage_record = await stream_response_handler(response, interrupt_flag)
else:
# 普通请求
endpoint = f"models/{model_info.model_identifier}:generateContent"
req_task = asyncio.create_task(
self._make_request("POST", endpoint, request_data)
)
while not req_task.done():
if interrupt_flag and interrupt_flag.is_set():
req_task.cancel()
raise ReqAbortException("请求被外部信号中断")
await asyncio.sleep(0.1)
response = req_task.result()
response_data = await response.json()
api_response, usage_record = async_response_parser(response_data)
except (ReqAbortException, NetworkConnectionError, RespNotOkException, RespParseException):
# 直接重抛项目定义的异常
raise
except Exception as e:
# 其他异常转换为网络连接错误
raise NetworkConnectionError() from e
# 设置使用统计
if usage_record:
api_response.usage = UsageRecord(
model_name=model_info.name,
provider_name=model_info.api_provider,
prompt_tokens=usage_record[0],
completion_tokens=usage_record[1],
total_tokens=usage_record[2],
)
return api_response
async def get_embedding(
self,
model_info: ModelInfo,
embedding_input: str,
extra_params: dict[str, Any] | None = None,
) -> APIResponse:
"""
获取文本嵌入 - 此客户端不支持嵌入功能
"""
raise NotImplementedError("AioHTTP Gemini客户端不支持文本嵌入功能")
async def get_audio_transcriptions(
self,
model_info: ModelInfo,
audio_base64: str,
extra_params: dict[str, Any] | None = None,
) -> APIResponse:
"""
获取音频转录
"""
# 构建包含音频的内容
contents = [{
"role": "user",
"parts": [
{"text": "Generate a transcript of the speech. The language of the transcript should match the language of the speech."},
{
"inline_data": {
"mime_type": "audio/wav",
"data": audio_base64
}
}
]
}]
request_data = {
"contents": contents,
"generationConfig": _build_generation_config(2048, 0.1, None, extra_params)
}
try:
endpoint = f"models/{model_info.model_identifier}:generateContent"
response = await self._make_request("POST", endpoint, request_data)
response_data = await response.json()
api_response, usage_record = _default_normal_response_parser(response_data)
if usage_record:
api_response.usage = UsageRecord(
model_name=model_info.name,
provider_name=model_info.api_provider,
prompt_tokens=usage_record[0],
completion_tokens=usage_record[1],
total_tokens=usage_record[2],
)
return api_response
except (NetworkConnectionError, RespNotOkException, RespParseException):
raise
except Exception as e:
raise NetworkConnectionError() from e
def get_support_image_formats(self) -> list[str]:
"""
获取支持的图片格式
"""
return ["png", "jpg", "jpeg", "webp", "heic", "heif"]
async def __aenter__(self):
"""异步上下文管理器入口"""
return self
async def __aexit__(self, exc_type, exc_val, exc_tb):
"""异步上下文管理器出口"""
if self.session and not self.session.closed:
await self.session.close()
def __del__(self):
"""析构函数,确保会话被正确关闭"""
if hasattr(self, 'session') and self.session and not self.session.closed:
asyncio.create_task(self.session.close())