214 lines
8.4 KiB
Python
214 lines
8.4 KiB
Python
"""Gemini model provider implementation."""
|
|
|
|
from typing import Optional
|
|
|
|
from google import genai
|
|
from google.genai import types
|
|
|
|
from .base import ModelCapabilities, ModelProvider, ModelResponse, ProviderType, RangeTemperatureConstraint
|
|
|
|
|
|
class GeminiModelProvider(ModelProvider):
|
|
"""Google Gemini model provider implementation."""
|
|
|
|
# Model configurations
|
|
SUPPORTED_MODELS = {
|
|
"gemini-2.5-flash-preview-05-20": {
|
|
"context_window": 1_048_576, # 1M tokens
|
|
"supports_extended_thinking": True,
|
|
"max_thinking_tokens": 24576, # Flash 2.5 thinking budget limit
|
|
},
|
|
"gemini-2.5-pro-preview-06-05": {
|
|
"context_window": 1_048_576, # 1M tokens
|
|
"supports_extended_thinking": True,
|
|
"max_thinking_tokens": 32768, # Pro 2.5 thinking budget limit
|
|
},
|
|
# Shorthands
|
|
"flash": "gemini-2.5-flash-preview-05-20",
|
|
"pro": "gemini-2.5-pro-preview-06-05",
|
|
}
|
|
|
|
# Thinking mode configurations - percentages of model's max_thinking_tokens
|
|
# These percentages work across all models that support thinking
|
|
THINKING_BUDGETS = {
|
|
"minimal": 0.005, # 0.5% of max - minimal thinking for fast responses
|
|
"low": 0.08, # 8% of max - light reasoning tasks
|
|
"medium": 0.33, # 33% of max - balanced reasoning (default)
|
|
"high": 0.67, # 67% of max - complex analysis
|
|
"max": 1.0, # 100% of max - full thinking budget
|
|
}
|
|
|
|
def __init__(self, api_key: str, **kwargs):
|
|
"""Initialize Gemini provider with API key."""
|
|
super().__init__(api_key, **kwargs)
|
|
self._client = None
|
|
self._token_counters = {} # Cache for token counting
|
|
|
|
@property
|
|
def client(self):
|
|
"""Lazy initialization of Gemini client."""
|
|
if self._client is None:
|
|
self._client = genai.Client(api_key=self.api_key)
|
|
return self._client
|
|
|
|
def get_capabilities(self, model_name: str) -> ModelCapabilities:
|
|
"""Get capabilities for a specific Gemini model."""
|
|
# Resolve shorthand
|
|
resolved_name = self._resolve_model_name(model_name)
|
|
|
|
if resolved_name not in self.SUPPORTED_MODELS:
|
|
raise ValueError(f"Unsupported Gemini model: {model_name}")
|
|
|
|
config = self.SUPPORTED_MODELS[resolved_name]
|
|
|
|
# Gemini models support 0.0-2.0 temperature range
|
|
temp_constraint = RangeTemperatureConstraint(0.0, 2.0, 0.7)
|
|
|
|
return ModelCapabilities(
|
|
provider=ProviderType.GOOGLE,
|
|
model_name=resolved_name,
|
|
friendly_name="Gemini",
|
|
context_window=config["context_window"],
|
|
supports_extended_thinking=config["supports_extended_thinking"],
|
|
supports_system_prompts=True,
|
|
supports_streaming=True,
|
|
supports_function_calling=True,
|
|
temperature_constraint=temp_constraint,
|
|
)
|
|
|
|
def generate_content(
|
|
self,
|
|
prompt: str,
|
|
model_name: str,
|
|
system_prompt: Optional[str] = None,
|
|
temperature: float = 0.7,
|
|
max_output_tokens: Optional[int] = None,
|
|
thinking_mode: str = "medium",
|
|
**kwargs,
|
|
) -> ModelResponse:
|
|
"""Generate content using Gemini model."""
|
|
# Validate parameters
|
|
resolved_name = self._resolve_model_name(model_name)
|
|
self.validate_parameters(resolved_name, temperature)
|
|
|
|
# Combine system prompt with user prompt if provided
|
|
if system_prompt:
|
|
full_prompt = f"{system_prompt}\n\n{prompt}"
|
|
else:
|
|
full_prompt = prompt
|
|
|
|
# Prepare generation config
|
|
generation_config = types.GenerateContentConfig(
|
|
temperature=temperature,
|
|
candidate_count=1,
|
|
)
|
|
|
|
# Add max output tokens if specified
|
|
if max_output_tokens:
|
|
generation_config.max_output_tokens = max_output_tokens
|
|
|
|
# Add thinking configuration for models that support it
|
|
capabilities = self.get_capabilities(resolved_name)
|
|
if capabilities.supports_extended_thinking and thinking_mode in self.THINKING_BUDGETS:
|
|
# Get model's max thinking tokens and calculate actual budget
|
|
model_config = self.SUPPORTED_MODELS.get(resolved_name)
|
|
if model_config and "max_thinking_tokens" in model_config:
|
|
max_thinking_tokens = model_config["max_thinking_tokens"]
|
|
actual_thinking_budget = int(max_thinking_tokens * self.THINKING_BUDGETS[thinking_mode])
|
|
generation_config.thinking_config = types.ThinkingConfig(thinking_budget=actual_thinking_budget)
|
|
|
|
try:
|
|
# Generate content
|
|
response = self.client.models.generate_content(
|
|
model=resolved_name,
|
|
contents=full_prompt,
|
|
config=generation_config,
|
|
)
|
|
|
|
# Extract usage information if available
|
|
usage = self._extract_usage(response)
|
|
|
|
return ModelResponse(
|
|
content=response.text,
|
|
usage=usage,
|
|
model_name=resolved_name,
|
|
friendly_name="Gemini",
|
|
provider=ProviderType.GOOGLE,
|
|
metadata={
|
|
"thinking_mode": thinking_mode if capabilities.supports_extended_thinking else None,
|
|
"finish_reason": (
|
|
getattr(response.candidates[0], "finish_reason", "STOP") if response.candidates else "STOP"
|
|
),
|
|
},
|
|
)
|
|
|
|
except Exception as e:
|
|
# Log error and re-raise with more context
|
|
error_msg = f"Gemini API error for model {resolved_name}: {str(e)}"
|
|
raise RuntimeError(error_msg) from e
|
|
|
|
def count_tokens(self, text: str, model_name: str) -> int:
|
|
"""Count tokens for the given text using Gemini's tokenizer."""
|
|
self._resolve_model_name(model_name)
|
|
|
|
# For now, use a simple estimation
|
|
# TODO: Use actual Gemini tokenizer when available in SDK
|
|
# Rough estimation: ~4 characters per token for English text
|
|
return len(text) // 4
|
|
|
|
def get_provider_type(self) -> ProviderType:
|
|
"""Get the provider type."""
|
|
return ProviderType.GOOGLE
|
|
|
|
def validate_model_name(self, model_name: str) -> bool:
|
|
"""Validate if the model name is supported."""
|
|
resolved_name = self._resolve_model_name(model_name)
|
|
return resolved_name in self.SUPPORTED_MODELS and isinstance(self.SUPPORTED_MODELS[resolved_name], dict)
|
|
|
|
def supports_thinking_mode(self, model_name: str) -> bool:
|
|
"""Check if the model supports extended thinking mode."""
|
|
capabilities = self.get_capabilities(model_name)
|
|
return capabilities.supports_extended_thinking
|
|
|
|
def get_thinking_budget(self, model_name: str, thinking_mode: str) -> int:
|
|
"""Get actual thinking token budget for a model and thinking mode."""
|
|
resolved_name = self._resolve_model_name(model_name)
|
|
model_config = self.SUPPORTED_MODELS.get(resolved_name, {})
|
|
|
|
if not model_config.get("supports_extended_thinking", False):
|
|
return 0
|
|
|
|
if thinking_mode not in self.THINKING_BUDGETS:
|
|
return 0
|
|
|
|
max_thinking_tokens = model_config.get("max_thinking_tokens", 0)
|
|
if max_thinking_tokens == 0:
|
|
return 0
|
|
|
|
return int(max_thinking_tokens * self.THINKING_BUDGETS[thinking_mode])
|
|
|
|
def _resolve_model_name(self, model_name: str) -> str:
|
|
"""Resolve model shorthand to full name."""
|
|
# Check if it's a shorthand
|
|
shorthand_value = self.SUPPORTED_MODELS.get(model_name.lower())
|
|
if isinstance(shorthand_value, str):
|
|
return shorthand_value
|
|
return model_name
|
|
|
|
def _extract_usage(self, response) -> dict[str, int]:
|
|
"""Extract token usage from Gemini response."""
|
|
usage = {}
|
|
|
|
# Try to extract usage metadata from response
|
|
# Note: The actual structure depends on the SDK version and response format
|
|
if hasattr(response, "usage_metadata"):
|
|
metadata = response.usage_metadata
|
|
if hasattr(metadata, "prompt_token_count"):
|
|
usage["input_tokens"] = metadata.prompt_token_count
|
|
if hasattr(metadata, "candidates_token_count"):
|
|
usage["output_tokens"] = metadata.candidates_token_count
|
|
if "input_tokens" in usage and "output_tokens" in usage:
|
|
usage["total_tokens"] = usage["input_tokens"] + usage["output_tokens"]
|
|
|
|
return usage
|