Spaces:
Sleeping
Sleeping
File size: 14,063 Bytes
226b286 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 |
import os
import tiktoken
from typing import Union
from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from langchain_openai import AzureChatOpenAI, ChatOpenAI, AzureOpenAIEmbeddings, OpenAIEmbeddings
from agents import OpenAIChatCompletionsModel
from openai import AsyncOpenAI, AsyncAzureOpenAI
from huggingface_hub import login
from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFaceEmbeddings
from langchain_ollama import ChatOllama, OllamaEmbeddings
class ModelFactory:
"""
A static utility class to create and return LLM instances based on the input type.
"""
@staticmethod
def get_model(framework: str = "openai-sdk-agent", # openai-sdk-agent, langchain, autogen
provider: str = "openai", # openai, azure, google, groq, huggingface, ollama
model_name: str = "gpt-4o-mini", # gpt-4o-mini, gemini-flash-1.5, groq/compound
model_info: dict = None, # additional info (e.g. backend provider for autogen/langchain)
temperature: float = 0
) -> Union[AzureChatOpenAI, ChatOpenAI, OpenAIChatCompletionsModel, ChatHuggingFace, ChatOllama]:
"""
Returns an LLM instance based on the specified parameters.
Parameters:
framework (str): The framework to use ('langchain', 'openai-sdk-agent', 'autogen').
provider (str): The model provider ('openai', 'azure', 'google', 'groq', 'huggingface', 'ollama').
model_name (str): The specific model name.
model_info (dict): Additional model info.
temperature (float): The temperature for generation (default 0).
Returns:
Union[...]: The model instance.
"""
# ----------------------------------------------------------------------
# AUTOGEN SUPPORT
# ----------------------------------------------------------------------
if framework.lower() == "autogen":
# Lazy import to avoid dependency issues if autogen is not installed
try:
from autogen_ext.models.openai import OpenAIChatCompletionClient
except ImportError as e:
raise ImportError("AutoGen libraries (autogen-agentchat, autogen-ext[openai]) are not installed.") from e
# Azure Backend
if provider.lower() == "azure":
token_provider = get_bearer_token_provider(
DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
return OpenAIChatCompletionClient(
model=model_name,
azure_endpoint=os.environ["AZURE_OPENAI_API_URI"],
api_version=os.environ["AZURE_OPENAI_API_VERSION"],
azure_ad_token_provider=token_provider,
temperature=temperature,
)
# OpenAI Backend
elif provider.lower() == "openai":
return OpenAIChatCompletionClient(
model=model_name,
api_key=os.environ["OPENAI_API_KEY"],
temperature=temperature,
)
# Google Backend (Gemini via OpenAI compat)
elif provider.lower() == "google" or provider.lower() == "gemini":
return OpenAIChatCompletionClient(
model=model_name,
base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
api_key=os.environ["GOOGLE_API_KEY"],
model_info=model_info, # Pass full model_info for capabilities
temperature=temperature,
)
# Groq Backend
elif provider.lower() == "groq":
return OpenAIChatCompletionClient(
model=model_name,
base_url="https://api.groq.com/openai/v1",
api_key=os.environ["GROQ_API_KEY"],
temperature=temperature,
)
# Ollama Backend
elif provider.lower() == "ollama":
# Ensure model_info defaults to empty dict if None
info = model_info if model_info is not None else {}
return OpenAIChatCompletionClient(
model=model_name,
base_url="http://localhost:11434/v1",
api_key="ollama", # dummy key
model_info=info,
temperature=temperature,
)
else:
raise ValueError(f"Unsupported AutoGen provider: {provider}")
# ----------------------------------------------------------------------
# LANGCHAIN SUPPORT
# ----------------------------------------------------------------------
elif framework.lower() == "langchain":
if provider.lower() == "azure":
token_provider = get_bearer_token_provider(
DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
return AzureChatOpenAI(
azure_endpoint=os.environ["AZURE_OPENAI_API_URI"],
azure_deployment=os.environ["AZURE_OPENAI_API_BASE_MODEL"],
api_version=os.environ["AZURE_OPENAI_API_VERSION"],
azure_ad_token_provider=token_provider,
model_name=model_name,
temperature=temperature,
)
elif provider.lower() == "openai":
return ChatOpenAI(
api_key=os.environ["OPENAI_API_KEY"],
model_name=model_name,
temperature=temperature,
)
elif provider.lower() == "huggingface":
if os.environ.get("HF_TOKEN"):
login(token=os.environ.get("HF_TOKEN"))
llm = HuggingFaceEndpoint(
repo_id=model_name,
task="text-generation",
temperature=temperature,
max_new_tokens=512,
huggingfacehub_api_token=os.environ.get("HF_TOKEN")
)
return ChatHuggingFace(llm=llm)
elif provider.lower() == "ollama":
return ChatOllama(model=model_name, temperature=temperature)
else:
raise ValueError(f"Unsupported LangChain provider: {provider}")
# ----------------------------------------------------------------------
# STANDARD LOGIC (Agents Lib / OpenAI SDK)
# ----------------------------------------------------------------------
elif framework.lower() == "openai-sdk-agent" or framework.lower() == "openai-sdk" or framework.lower() == "openai":
if provider.lower() == "azure":
token_provider = get_bearer_token_provider(
DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
client = AsyncAzureOpenAI(
azure_endpoint=os.environ["AZURE_OPENAI_API_URI"],
api_version=os.environ["AZURE_OPENAI_API_VERSION"],
azure_ad_token_provider=token_provider,
)
return OpenAIChatCompletionsModel(model=model_name, openai_client=client)
elif provider.lower() == "openai":
client = AsyncOpenAI(api_key=os.environ["OPENAI_API_KEY"])
return OpenAIChatCompletionsModel(model=model_name, openai_client=client)
elif provider.lower() == "google":
GEMINI_BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
client = AsyncOpenAI(
base_url=GEMINI_BASE_URL,
api_key=os.environ["GOOGLE_API_KEY"]
)
return OpenAIChatCompletionsModel(model=model_name, openai_client=client)
elif provider.lower() == "groq":
GROQ_BASE_URL = "https://api.groq.com/openai/v1"
client = AsyncOpenAI(
base_url=GROQ_BASE_URL,
api_key=os.environ["GROQ_API_KEY"]
)
return OpenAIChatCompletionsModel(model=model_name, openai_client=client)
elif provider.lower() == "ollama":
client = AsyncOpenAI(
base_url="http://localhost:11434/v1",
api_key="ollama"
)
return OpenAIChatCompletionsModel(model=model_name, openai_client=client)
elif provider.lower() == "huggingface":
# Agents lib doesn't have native HF support in the same way
raise ValueError("For Hugging Face, please use framework='langchain'")
else:
raise ValueError(f"Unsupported provider for openai-sdk-agent: {provider}")
else:
raise ValueError(f"Unsupported framework: {framework}")
@staticmethod
def num_tokens_from_messages(messages, model: str = "gpt-4o"):
"""
Return the number of tokens used by a list of messages.
"""
try:
encoding = tiktoken.encoding_for_model(model)
except KeyError:
encoding = tiktoken.get_encoding("cl100k_base")
tokens_per_message = 3
num_tokens = 0
for message in messages:
num_tokens += tokens_per_message
for key, value in message.items():
if key == "name":
num_tokens += 1
# Encode values if they are strings
if isinstance(value, str):
num_tokens += len(encoding.encode(value))
elif isinstance(value, list) and key == "content":
for part in value:
if isinstance(part, dict) and part.get("type") == "text":
num_tokens += len(encoding.encode(part.get("text", "")))
elif isinstance(part, dict) and part.get("type") == "image_url":
num_tokens += 85
num_tokens += 3
return num_tokens
class EmbeddingFactory:
"""
A static utility class to create and return Embedding Model instances.
"""
@staticmethod
def get_embedding_model(provider: str = "openai",
model_name: str = "text-embedding-3-small"
) -> Union[AzureOpenAIEmbeddings, OpenAIEmbeddings, OllamaEmbeddings, HuggingFaceEmbeddings]:
if provider.lower() == "azure":
token_provider = get_bearer_token_provider(
DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
return AzureOpenAIEmbeddings(
azure_endpoint=os.environ["AZURE_OPENAI_API_URI"],
azure_deployment=os.environ.get("AZURE_OPENAI_EMBEDDING_DEPLOYMENT", model_name),
api_version=os.environ["AZURE_OPENAI_API_VERSION"],
azure_ad_token_provider=token_provider,
)
elif provider.lower() == "openai":
return OpenAIEmbeddings(
api_key=os.environ["OPENAI_API_KEY"],
model=model_name
)
elif provider.lower() == "ollama":
return OllamaEmbeddings(model=model_name)
elif provider.lower() == "huggingface":
if os.environ.get("HF_TOKEN"):
login(token=os.environ.get("HF_TOKEN"))
return HuggingFaceEmbeddings(model_name=model_name)
else:
raise ValueError(f"Unsupported embedding provider: {provider}")
# =================================================================================================
# GLOBAL HELPER FUNCTIONS (for agents)
# =================================================================================================
# model used for orchestrator or executor
# def get_model(provider:str = "google", framework:str = "openai-sdk", model_name:str = "gemini-2.5-flash"):
def get_model(provider:str = "openai", framework:str = "openai", model_name:str = "gpt-4-turbo"):
# def get_model(provider:str = "groq", framework:str = "openai-sdk", model_name:str = "openai/gpt-oss-120b"):
model_info = None
if provider in list["gemini", "google"]:
model_info = {
"family": "gemini",
"vision": True,
"function_calling": True,
"json_output": True,
"structured_output": True,
}
return ModelFactory.get_model( framework=framework,
provider=provider,
model_name=model_name,
model_info=model_info,
temperature=0)
# else:
# return ModelFactory.get_model( framework="openai-sdk",
# provider="openai",
# model_name="gpt-4o-mini",
# temperature=0)
# Use this model where agent executing tool and returning JSON
def get_model_json(model_name: str = "gpt-4.1-mini", provider: str = "openai"):
return ModelFactory.get_model( framework="openai-sdk",
provider=provider,
model_name=model_name,
temperature=0)
|