Spaces:
Sleeping
Sleeping
File size: 11,685 Bytes
2fd8593 2c0acc5 2fd8593 97a9e25 2fd8593 97a9e25 2fd8593 97a9e25 2fd8593 2c0acc5 2fd8593 757c114 2fd8593 97a9e25 2fd8593 757c114 2fd8593 97a9e25 2fd8593 757c114 2fd8593 2c0acc5 2fd8593 2c0acc5 2fd8593 97a9e25 2fd8593 2c0acc5 2fd8593 97a9e25 2c0acc5 2fd8593 97a9e25 2fd8593 2c0acc5 2fd8593 97a9e25 | 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 | """
LLM Provider abstraction layer for Blog2Code.
Supports multiple LLM providers: OpenAI, Google Gemini, NVIDIA Gemma
"""
import os
from typing import Dict, List, Any, Optional
from abc import ABC, abstractmethod
class LLMProvider(ABC):
"""Base class for LLM providers"""
@abstractmethod
def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
"""Create a chat completion"""
pass
@abstractmethod
def get_response_text(self, completion: Any) -> str:
"""Extract text from completion response"""
pass
@abstractmethod
def get_usage_info(self, completion: Any) -> Dict:
"""Extract token usage information"""
pass
@abstractmethod
def calculate_cost(self, usage: Dict, model: str) -> float:
"""Calculate cost based on usage"""
pass
class OpenAIProvider(LLMProvider):
"""OpenAI API implementation"""
def __init__(self, api_key: Optional[str] = None):
from openai import OpenAI
self.client = OpenAI(api_key=api_key or os.environ.get("OPENAI_API_KEY"))
def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
"""Create OpenAI chat completion"""
return self.client.chat.completions.create(
model=model,
messages=messages,
**kwargs
)
def get_response_text(self, completion: Any) -> str:
"""Extract text from OpenAI response"""
return completion.choices[0].message.content
def get_usage_info(self, completion: Any) -> Dict:
"""Extract usage from OpenAI response"""
return {
'prompt_tokens': completion.usage.prompt_tokens,
'completion_tokens': completion.usage.completion_tokens,
'total_tokens': completion.usage.total_tokens,
'cached_tokens': getattr(completion.usage.prompt_tokens_details, 'cached_tokens', 0) if hasattr(completion.usage, 'prompt_tokens_details') else 0
}
def calculate_cost(self, usage: Dict, model: str) -> float:
"""Calculate OpenAI cost"""
model_costs = {
"gpt-4o-mini": {"input": 0.150, "cached": 0.075, "output": 0.600},
"gpt-4o": {"input": 2.50, "cached": 1.25, "output": 10.00},
"gpt-3.5-turbo": {"input": 0.50, "cached": 0.25, "output": 1.50},
"o3-mini": {"input": 1.10, "cached": 0.55, "output": 4.40},
}
costs = model_costs.get(model, model_costs["gpt-4o-mini"])
prompt_tokens = usage['prompt_tokens']
cached_tokens = usage.get('cached_tokens', 0)
completion_tokens = usage['completion_tokens']
actual_input_tokens = prompt_tokens - cached_tokens
input_cost = (actual_input_tokens / 1_000_000) * costs["input"]
cached_cost = (cached_tokens / 1_000_000) * costs["cached"]
output_cost = (completion_tokens / 1_000_000) * costs["output"]
return input_cost + cached_cost + output_cost
class GeminiProvider(LLMProvider):
"""Google Gemini API implementation"""
def __init__(self, api_key: Optional[str] = None):
try:
import google.generativeai as genai
self.genai = genai
genai.configure(api_key=api_key or os.environ.get("GEMINI_API_KEY"))
except ImportError:
raise ImportError(
"google-generativeai not installed. "
"Install with: pip install google-generativeai"
)
def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
"""Create Gemini chat completion"""
gemini_messages = self._convert_messages(messages)
# Do NOT add models/ prefix - pass model name directly
gemini_model = self.genai.GenerativeModel(model)
response = gemini_model.generate_content(
gemini_messages,
generation_config=self._get_generation_config(**kwargs)
)
return response
def _convert_messages(self, messages: List[Dict]) -> str:
"""Convert OpenAI messages to Gemini prompt format"""
prompt_parts = []
for msg in messages:
role = msg['role']
content = msg['content']
if role == 'system':
prompt_parts.append(f"System Instructions:\n{content}\n")
elif role == 'user':
prompt_parts.append(f"User:\n{content}\n")
elif role == 'assistant':
prompt_parts.append(f"Assistant:\n{content}\n")
return "\n".join(prompt_parts)
def _get_generation_config(self, **kwargs):
"""Convert OpenAI kwargs to Gemini generation config"""
config = {}
if 'temperature' in kwargs:
config['temperature'] = kwargs['temperature']
if 'max_tokens' in kwargs:
config['max_output_tokens'] = kwargs['max_tokens']
if 'top_p' in kwargs:
config['top_p'] = kwargs['top_p']
return config
def get_response_text(self, completion: Any) -> str:
"""Extract text from Gemini response"""
return completion.text
def get_usage_info(self, completion: Any) -> Dict:
"""Extract usage from Gemini response"""
try:
metadata = completion.usage_metadata
return {
'prompt_tokens': metadata.prompt_token_count,
'completion_tokens': metadata.candidates_token_count,
'total_tokens': metadata.total_token_count,
'cached_tokens': getattr(metadata, 'cached_content_token_count', 0)
}
except:
return {
'prompt_tokens': 0,
'completion_tokens': 0,
'total_tokens': 0,
'cached_tokens': 0
}
def calculate_cost(self, usage: Dict, model: str) -> float:
"""Calculate Gemini cost"""
model_costs = {
"gemini-1.5-flash": {"input": 0.075, "cached": 0.01875, "output": 0.30},
"gemini-1.5-pro": {"input": 1.25, "cached": 0.3125, "output": 5.00},
"gemini-2.0-flash": {"input": 0.0, "cached": 0.0, "output": 0.0},
"gemini-2.0-flash-lite": {"input": 0.0, "cached": 0.0, "output": 0.0},
}
costs = model_costs.get(model, {"input": 0.0, "cached": 0.0, "output": 0.0})
prompt_tokens = usage['prompt_tokens']
cached_tokens = usage.get('cached_tokens', 0)
completion_tokens = usage['completion_tokens']
actual_input_tokens = prompt_tokens - cached_tokens
input_cost = (actual_input_tokens / 1_000_000) * costs["input"]
cached_cost = (cached_tokens / 1_000_000) * costs["cached"]
output_cost = (completion_tokens / 1_000_000) * costs["output"]
return input_cost + cached_cost + output_cost
class GemmaProvider(LLMProvider):
"""NVIDIA API implementation — supports Gemma, Llama, and other NVIDIA-hosted models"""
def __init__(self, api_key: Optional[str] = None):
import requests
self.requests = requests
self.api_key = api_key or os.environ.get("NVIDIA_API_KEY")
if not self.api_key:
raise ValueError(
"NVIDIA_API_KEY not found. "
"Set it as an environment variable or pass it to the constructor."
)
self.invoke_url = "https://integrate.api.nvidia.com/v1/chat/completions"
def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
"""Create NVIDIA API chat completion with retry logic"""
import time
headers = {
"Authorization": f"Bearer {self.api_key}",
"Accept": "application/json"
}
payload = {
"model": model,
"messages": messages,
"max_tokens": kwargs.get('max_tokens', 8192),
"temperature": kwargs.get('temperature', 0.20),
"top_p": kwargs.get('top_p', 0.70),
"stream": False
}
max_retries = 5
for attempt in range(max_retries):
try:
response = self.requests.post(self.invoke_url, headers=headers, json=payload)
response.raise_for_status()
return response.json()
except Exception as e:
if attempt < max_retries - 1:
wait = 10 * (attempt + 1) # 10s, 20s, 30s, 40s
print(f"[RETRY] Attempt {attempt+1} failed: {e}. Retrying in {wait}s...")
time.sleep(wait)
else:
raise
def get_response_text(self, completion: Any) -> str:
"""Extract text from NVIDIA API response"""
if isinstance(completion, dict):
return completion['choices'][0]['message']['content']
return str(completion)
def get_usage_info(self, completion: Any) -> Dict:
"""Extract usage from NVIDIA API response"""
try:
usage = completion.get('usage', {})
return {
'prompt_tokens': usage.get('prompt_tokens', 0),
'completion_tokens': usage.get('completion_tokens', 0),
'total_tokens': usage.get('total_tokens', 0),
'cached_tokens': 0
}
except:
return {
'prompt_tokens': 0,
'completion_tokens': 0,
'total_tokens': 0,
'cached_tokens': 0
}
def calculate_cost(self, usage: Dict, model: str) -> float:
"""Calculate NVIDIA API cost"""
model_costs = {
"google/gemma-3-27b-it": {"input": 0.0, "output": 0.0},
"meta/llama-3.3-70b-instruct": {"input": 0.0, "output": 0.0},
"meta/llama-3.1-8b-instruct": {"input": 0.0, "output": 0.0},
}
costs = model_costs.get(model, {"input": 0.0, "output": 0.0})
prompt_tokens = usage['prompt_tokens']
completion_tokens = usage['completion_tokens']
input_cost = (prompt_tokens / 1_000_000) * costs["input"]
output_cost = (completion_tokens / 1_000_000) * costs["output"]
return input_cost + output_cost
def get_provider(provider_name: str, api_key: Optional[str] = None) -> LLMProvider:
"""Factory function to get LLM provider."""
providers = {
'openai': OpenAIProvider,
'gemini': GeminiProvider,
'gemma': GemmaProvider,
}
if provider_name not in providers:
raise ValueError(
f"Unknown provider: {provider_name}. "
f"Available providers: {list(providers.keys())}"
)
return providers[provider_name](api_key=api_key)
def get_default_model(provider_name: str) -> str:
"""Get default model for a provider"""
defaults = {
'openai': 'gpt-4o-mini',
'gemini': 'gemini-1.5-flash',
'gemma': 'meta/llama-3.3-70b-instruct', # Llama via NVIDIA API
}
return defaults.get(provider_name, 'gpt-4o-mini')
if __name__ == "__main__":
print("Testing LLM Provider abstraction...")
try:
provider = get_provider('openai')
print("✅ OpenAI provider initialized")
except Exception as e:
print(f"❌ OpenAI provider failed: {e}")
try:
provider = get_provider('gemini')
print("✅ Gemini provider initialized")
except Exception as e:
print(f"❌ Gemini provider failed: {e}")
try:
provider = get_provider('gemma')
print("✅ Gemma provider initialized")
except Exception as e:
print(f"❌ Gemma provider failed: {e}") |