"""Configuration models for LLM providers."""
from __future__ import annotations
from typing import Literal
from pydantic import BaseModel, Field
class OllamaConfig(BaseModel):
"""Configuration for Ollama provider."""
base_url: str = "http://localhost:11434"
model: str = "gemma3"
timeout_seconds: float = 30.0
max_retries: int = 3
retry_delay_seconds: float = 1.0
retry_backoff_multiplier: float = 2.0
class OpenAIConfig(BaseModel):
"""Configuration for OpenAI provider."""
api_key: str | None = None
model: str = "gpt-4"
timeout_seconds: float = 30.0
max_retries: int = 3
class GeminiConfig(BaseModel):
"""Configuration for Gemini provider."""
api_key: str | None = None
model: str = "gemini-pro"
timeout_seconds: float = 30.0
max_retries: int = 3
class LLMConfig(BaseModel):
"""Main LLM configuration."""
provider: Literal["ollama", "openai", "gemini"] = "ollama"
ollama: OllamaConfig = Field(default_factory=OllamaConfig)
openai: OpenAIConfig | None = None
gemini: GeminiConfig | None = None
def get_model(self) -> str:
"""Return the configured model name for the active provider.
Today we primarily use Ollama. For other providers, raise a clear error
if their nested config isn't set to avoid silently using the wrong model.
"""
if self.provider == "ollama":
return self.ollama.model
if self.provider == "openai":
if not self.openai:
raise ValueError("LLM provider is openai but llm.openai is not configured")
return self.openai.model
if self.provider == "gemini":
if not self.gemini:
raise ValueError("LLM provider is gemini but llm.gemini is not configured")
return self.gemini.model
# Defensive: Literal should prevent this, but keep a stable error.
raise ValueError(f"Unknown LLM provider: {self.provider}")
