Spaces:
Sleeping
Sleeping
File size: 11,790 Bytes
93eb4a4 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 | """Multi-LLM Provider Support
Unified interface for OpenAI, Ollama, and HuggingFace Inference API.
Enables seamless switching between LLM backends with parameterized configuration.
"""
import os
from abc import ABC, abstractmethod
from typing import Optional, List, Dict, Any
from dataclasses import dataclass
@dataclass
class LLMResponse:
"""Standardized LLM response"""
text: str
model: str
provider: str
tokens_used: Optional[int] = None
metadata: Optional[Dict[str, Any]] = None
class LLMProvider(ABC):
"""Abstract base class for LLM providers"""
def __init__(self, model_name: str, **kwargs):
self.model_name = model_name
self.config = kwargs
@abstractmethod
def generate(self, prompt: str, **kwargs) -> LLMResponse:
"""Generate text from prompt"""
pass
@abstractmethod
def score_response(self, prompt: str, response: str) -> float:
"""Score quality of a response (0-1)"""
pass
class OpenAIProvider(LLMProvider):
"""OpenAI API provider"""
def __init__(self, model_name: str = "gpt-4", api_key: Optional[str] = None, **kwargs):
super().__init__(model_name, **kwargs)
self.api_key = api_key or os.getenv("OPENAI_API_KEY")
try:
from openai import OpenAI
self.client = OpenAI(api_key=self.api_key)
self.available = True
except ImportError:
print("Warning: openai package not installed. Install with: pip install openai")
self.available = False
except Exception as e:
print(f"Warning: Could not initialize OpenAI client: {e}")
self.available = False
def generate(self, prompt: str, temperature: float = 0.7, max_tokens: int = 1000, **kwargs) -> LLMResponse:
if not self.available:
return LLMResponse(
text="OpenAI provider not available. Please install openai package and set API key.",
model=self.model_name,
provider="openai"
)
try:
response = self.client.chat.completions.create(
model=self.model_name,
messages=[{"role": "user", "content": prompt}],
temperature=temperature,
max_tokens=max_tokens,
**kwargs
)
return LLMResponse(
text=response.choices[0].message.content,
model=self.model_name,
provider="openai",
tokens_used=response.usage.total_tokens if hasattr(response, 'usage') else None,
metadata={"finish_reason": response.choices[0].finish_reason}
)
except Exception as e:
return LLMResponse(
text=f"Error generating response: {str(e)}",
model=self.model_name,
provider="openai"
)
def score_response(self, prompt: str, response: str) -> float:
"""Score response quality using a simple prompt"""
if not self.available:
return 0.5
try:
score_prompt = f"""Rate the quality of this response on a scale of 0.0 to 1.0.
Prompt: {prompt}
Response: {response}
Provide only a number between 0.0 and 1.0:"""
result = self.generate(score_prompt, temperature=0.1, max_tokens=10)
# Extract number from response
try:
score = float(result.text.strip())
return max(0.0, min(1.0, score))
except ValueError:
return 0.5
except Exception:
return 0.5
class OllamaProvider(LLMProvider):
"""Ollama local LLM provider"""
def __init__(self, model_name: str = "llama2", base_url: str = "http://localhost:11434", **kwargs):
super().__init__(model_name, **kwargs)
self.base_url = base_url
try:
import requests
self.requests = requests
# Test connection
response = requests.get(f"{base_url}/api/tags", timeout=2)
self.available = response.status_code == 200
except ImportError:
print("Warning: requests package not installed. Install with: pip install requests")
self.available = False
except Exception as e:
print(f"Warning: Could not connect to Ollama at {base_url}: {e}")
self.available = False
def generate(self, prompt: str, temperature: float = 0.7, max_tokens: int = 1000, **kwargs) -> LLMResponse:
if not self.available:
return LLMResponse(
text="Ollama provider not available. Please ensure Ollama is running at " + self.base_url,
model=self.model_name,
provider="ollama"
)
try:
response = self.requests.post(
f"{self.base_url}/api/generate",
json={
"model": self.model_name,
"prompt": prompt,
"temperature": temperature,
"num_predict": max_tokens,
"stream": False
},
timeout=60
)
if response.status_code == 200:
data = response.json()
return LLMResponse(
text=data.get("response", ""),
model=self.model_name,
provider="ollama",
metadata=data
)
else:
return LLMResponse(
text=f"Ollama error: {response.status_code}",
model=self.model_name,
provider="ollama"
)
except Exception as e:
return LLMResponse(
text=f"Error generating response: {str(e)}",
model=self.model_name,
provider="ollama"
)
def score_response(self, prompt: str, response: str) -> float:
"""Simple heuristic scoring for Ollama"""
# Length-based heuristic (can be improved)
if len(response) < 10:
return 0.2
elif len(response) < 50:
return 0.5
elif len(response) < 200:
return 0.7
else:
return 0.9
class HuggingFaceProvider(LLMProvider):
"""HuggingFace Inference API provider"""
def __init__(self, model_name: str = "HuggingFaceH4/zephyr-7b-beta", api_token: Optional[str] = None, **kwargs):
super().__init__(model_name, **kwargs)
self.api_token = api_token or os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_TOKEN")
try:
from huggingface_hub import InferenceClient
self.client = InferenceClient(token=self.api_token)
self.available = True
except ImportError:
print("Warning: huggingface_hub not installed. Install with: pip install huggingface_hub")
self.available = False
except Exception as e:
print(f"Warning: Could not initialize HuggingFace client: {e}")
self.available = False
def generate(self, prompt: str, temperature: float = 0.7, max_tokens: int = 1000, **kwargs) -> LLMResponse:
if not self.available:
return LLMResponse(
text="HuggingFace provider not available. Please install huggingface_hub.",
model=self.model_name,
provider="huggingface"
)
try:
response = self.client.text_generation(
prompt,
model=self.model_name,
max_new_tokens=max_tokens,
temperature=temperature,
**kwargs
)
return LLMResponse(
text=response,
model=self.model_name,
provider="huggingface"
)
except Exception as e:
return LLMResponse(
text=f"Error generating response: {str(e)}",
model=self.model_name,
provider="huggingface"
)
def score_response(self, prompt: str, response: str) -> float:
"""Simple heuristic scoring"""
# Basic quality metrics
score = 0.5
# Length check
if 50 <= len(response) <= 2000:
score += 0.2
# Contains relevant terms from prompt
prompt_words = set(prompt.lower().split())
response_words = set(response.lower().split())
overlap = len(prompt_words & response_words) / max(len(prompt_words), 1)
score += overlap * 0.3
return min(1.0, score)
class LLMManager:
"""Unified LLM manager for seamless provider switching"""
def __init__(self, default_provider: str = "huggingface"):
self.providers: Dict[str, LLMProvider] = {}
self.default_provider = default_provider
self._initialize_providers()
def _initialize_providers(self):
"""Initialize all available providers"""
# OpenAI
openai_model = os.getenv("OPENAI_MODEL", "gpt-4")
self.providers["openai"] = OpenAIProvider(model_name=openai_model)
# Ollama
ollama_model = os.getenv("OLLAMA_MODEL", "llama2")
ollama_url = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
self.providers["ollama"] = OllamaProvider(model_name=ollama_model, base_url=ollama_url)
# HuggingFace
hf_model = os.getenv("HF_MODEL", "HuggingFaceH4/zephyr-7b-beta")
self.providers["huggingface"] = HuggingFaceProvider(model_name=hf_model)
def generate(
self,
prompt: str,
provider: Optional[str] = None,
**kwargs
) -> LLMResponse:
"""Generate text using specified or default provider"""
provider_name = provider or self.default_provider
if provider_name not in self.providers:
return LLMResponse(
text=f"Provider '{provider_name}' not found. Available: {list(self.providers.keys())}",
model="unknown",
provider="unknown"
)
return self.providers[provider_name].generate(prompt, **kwargs)
def score_response(self, prompt: str, response: str, provider: Optional[str] = None) -> float:
"""Score response quality"""
provider_name = provider or self.default_provider
if provider_name not in self.providers:
return 0.5
return self.providers[provider_name].score_response(prompt, response)
def get_available_providers(self) -> List[str]:
"""Get list of available providers"""
return [
name for name, provider in self.providers.items()
if getattr(provider, 'available', True)
]
def set_default_provider(self, provider: str):
"""Set default provider"""
if provider in self.providers:
self.default_provider = provider
else:
raise ValueError(f"Provider '{provider}' not found")
# Singleton instance
_llm_manager = None
def get_llm_manager(default_provider: Optional[str] = None) -> LLMManager:
"""Get or create LLM manager singleton"""
global _llm_manager
if _llm_manager is None:
provider = default_provider or os.getenv("LLM_PROVIDER", "huggingface")
_llm_manager = LLMManager(default_provider=provider)
elif default_provider:
_llm_manager.set_default_provider(default_provider)
return _llm_manager
|