241 lines
6.9 KiB
Python
241 lines
6.9 KiB
Python
from dataclasses import dataclass, is_dataclass, fields, MISSING
|
|
from typing import Any, Tuple, Type, Optional
|
|
import yaml
|
|
from pathlib import Path
|
|
from typing import Dict
|
|
import os
|
|
|
|
from loguru import logger
|
|
from dotenv import load_dotenv
|
|
|
|
load_dotenv()
|
|
|
|
|
|
def resolve_llm_api_key(api_key: Optional[str]) -> Optional[str]:
|
|
"""Resolve the API key for OpenAI-compatible providers."""
|
|
if api_key not in (None, "", "wrong-key"):
|
|
resolved_key = api_key
|
|
else:
|
|
resolved_key = os.environ.get("ALI_API_KEY") or os.environ.get("OPENAI_API_KEY")
|
|
|
|
# Some OpenAI-compatible integrations still read OPENAI_API_KEY from env.
|
|
if resolved_key and not os.environ.get("OPENAI_API_KEY"):
|
|
os.environ["OPENAI_API_KEY"] = resolved_key
|
|
|
|
return resolved_key
|
|
|
|
## NOTE: base classes taken from nerfstudio
|
|
class PrintableConfig:
|
|
"""
|
|
Printable Config defining str function
|
|
定义 __str__ 方法的可打印配置类
|
|
|
|
"""
|
|
|
|
def __str__(self):
|
|
lines = [self.__class__.__name__ + ":"]
|
|
for key, val in vars(self).items():
|
|
|
|
if self.is_secrete(key):
|
|
val = str(val)
|
|
val = val[:3] + "*"*(len(val) - 6) + val[-3:]
|
|
|
|
if isinstance(val, Tuple):
|
|
flattened_val = "["
|
|
for item in val:
|
|
flattened_val += str(item) + "\n"
|
|
flattened_val = flattened_val.rstrip("\n")
|
|
val = flattened_val + "]"
|
|
lines += f"{key}: {str(val)}".split("\n")
|
|
return "\n" + "\n ".join(lines)
|
|
|
|
def is_secrete(self, inp:str):
|
|
sec_list = ["secret", "api_key"]
|
|
for sec in sec_list:
|
|
if sec in inp:
|
|
return True
|
|
|
|
return False
|
|
|
|
|
|
# Base instantiate configs
|
|
@dataclass
|
|
class InstantiateConfig(PrintableConfig):
|
|
"""
|
|
Config class for instantiating an the class specified in the _target attribute.
|
|
|
|
用于实例化 _target 属性指定的类的配置类
|
|
|
|
"""
|
|
|
|
_target: Type
|
|
|
|
def setup(self, **kwargs) -> Any:
|
|
"""
|
|
Returns the instantiated object using the config.
|
|
|
|
使用配置返回实例化的对象
|
|
|
|
"""
|
|
return self._target(self, **kwargs)
|
|
|
|
def save_config(self, filename: str) -> None:
|
|
"""
|
|
Save the config to a YAML file.
|
|
|
|
将配置保存到 YAML 文件
|
|
|
|
"""
|
|
# Persist the full config object (including type tags) so it can be
|
|
# deserialized back into config instances with methods like .setup().
|
|
# Secret masking is intentionally handled by __str__ for printing/logging,
|
|
# not when writing to disk.
|
|
with open(filename, "w", encoding="utf-8") as f:
|
|
yaml.dump(self, f)
|
|
logger.info(f"[yellow]config saved to: {filename}[/yellow]")
|
|
|
|
def get_name(self):
|
|
return self.__class__.__name__
|
|
|
|
|
|
@dataclass
|
|
class LLMKeyConfig(InstantiateConfig):
|
|
llm_name: str = "qwen-plus"
|
|
"""name of llm"""
|
|
|
|
llm_provider:str = "openai"
|
|
"""provider of the llm"""
|
|
|
|
base_url:str = "https://dashscope.aliyuncs.com/compatible-mode/v1"
|
|
"""base url; could be used to overwrite the baseurl in llm provider"""
|
|
|
|
api_key:str = None
|
|
"""api key for llm"""
|
|
|
|
def __post_init__(self):
|
|
original_api_key = self.api_key
|
|
self.api_key = resolve_llm_api_key(self.api_key)
|
|
if self.api_key is None:
|
|
logger.error("no ALI_API_KEY or OPENAI_API_KEY provided for embedding")
|
|
elif original_api_key in (None, "", "wrong-key"):
|
|
logger.info("LLM API key loaded from environment")
|
|
|
|
|
|
@dataclass
|
|
class LLMNodeConfig(LLMKeyConfig):
|
|
"""
|
|
class is for LLM nodes that has system prompt config
|
|
"""
|
|
|
|
pipeline_id: Optional[str] = None
|
|
"""If set, load prompts from database (with file fallback)"""
|
|
|
|
prompt_set_id: Optional[str] = None
|
|
"""If set, load from this specific prompt set instead of the active one"""
|
|
|
|
|
|
@dataclass
|
|
class ToolConfig(InstantiateConfig):
|
|
use_tool:bool = True
|
|
"""
|
|
specify to use tool or not
|
|
|
|
指定是否使用工具
|
|
"""
|
|
|
|
def load_tyro_conf(filename: str, inp_conf = None) -> InstantiateConfig:
|
|
"""
|
|
load and overwrite config from file
|
|
|
|
从文件加载并覆盖配置
|
|
|
|
"""
|
|
config = yaml.load(Path(filename).read_text(), Loader=yaml.Loader)
|
|
|
|
# config = ovewrite_config(config, inp_conf) if inp_conf is not None else config
|
|
return config
|
|
|
|
def is_default(instance, field_):
|
|
"""
|
|
Check if the value of a field in a dataclass instance is the default value.
|
|
|
|
检查数据类实例中字段的值是否为默认值
|
|
|
|
"""
|
|
value = getattr(instance, field_.name)
|
|
|
|
if field_.default is not MISSING:
|
|
# Compare with default value
|
|
"""
|
|
与默认值进行比较
|
|
|
|
如果字段有默认值，则将当前值与默认值进行比较
|
|
|
|
"""
|
|
return value == field_.default
|
|
elif field_.default_factory is not MISSING:
|
|
# Compare with value generated by the default factory
|
|
"""
|
|
与默认工厂生成的值进行比较
|
|
|
|
如果字段有默认工厂，则将当前值与默认工厂生成的值进行比较
|
|
|
|
"""
|
|
return value == field_.default_factory()
|
|
else:
|
|
# No default value specified
|
|
return False
|
|
|
|
def ovewrite_config(loaded_conf, inp_conf):
|
|
"""
|
|
for non-default values in inp_conf, overwrite the corresponding values in loaded_conf
|
|
|
|
对于 inp_conf 中的非默认值，覆盖 loaded_conf 中对应的配置
|
|
|
|
"""
|
|
if not (is_dataclass(loaded_conf) and is_dataclass(inp_conf)):
|
|
return loaded_conf
|
|
|
|
for field_ in fields(loaded_conf):
|
|
field_name = field_.name
|
|
"""
|
|
if field_name in inp_conf:
|
|
|
|
如果字段名在 inp_conf 中，则进行覆盖
|
|
|
|
"""
|
|
current_value = getattr(inp_conf, field_name)
|
|
new_value = getattr(inp_conf, field_name)
|
|
|
|
"""
|
|
inp_conf[field_name]
|
|
从 inp_conf 中获取字段值
|
|
|
|
如果字段名在 inp_conf 中，则获取其值
|
|
|
|
"""
|
|
|
|
if is_dataclass(current_value):
|
|
|
|
"""
|
|
Recurse for nested dataclasses
|
|
|
|
递归处理嵌套的数据类
|
|
|
|
如果当前值是数据类，则递归调用 ovewrite_config 进行合并
|
|
|
|
"""
|
|
merged_value = ovewrite_config(current_value, new_value)
|
|
setattr(loaded_conf, field_name, merged_value)
|
|
elif not is_default(inp_conf, field_):
|
|
"""
|
|
Overwrite only if the current value is not default
|
|
|
|
仅在当前值不是默认值时进行覆盖
|
|
|
|
如果 inp_conf 中的字段值不是默认值，则覆盖 loaded_conf 中的对应值
|
|
|
|
"""
|
|
setattr(loaded_conf, field_name, new_value)
|
|
|
|
return loaded_conf |