from typing import Any, Optional
from ...key_management import SmartKeyPool
from ...prompts import default_instruction_generation_prompt
from ...monitoring import GenerationMonitor
from ...types import ModelProviderName
from .llm_backed import LLMBackedInstructionGeneratorCallback
[docs]
class SimpleInstructionGeneratorCallback(LLMBackedInstructionGeneratorCallback):
"""Generates instructions from the correspondent prompt only (no document context).
Omits a ``provider`` attribute so :class:`~afterimage.sampling.SamplingStrategy` does not
treat this callback as document-backed.
Args:
api_key: API key for the generative AI service.
prompt: System instruction for instruction generation. If None, uses the default.
model_name: Model name to use.
model_provider_name: Model provider name to use.
n_instructions: Number of instructions to generate in each round.
safety_settings: Safety settings for the model (mainly for Gemini).
monitor: Optional :class:`~afterimage.monitoring.GenerationMonitor`.
"""
def __init__(
self,
api_key: str | SmartKeyPool,
prompt: str | None = None,
model_name: str | None = None,
model_provider_name: ModelProviderName = "gemini",
n_instructions: int = 3,
safety_settings: Optional[dict] = None,
monitor: GenerationMonitor | None = None,
llm_create_extras: dict[str, Any] | None = None,
):
base_prompt = (
prompt if prompt is not None else default_instruction_generation_prompt
)
super().__init__(
api_key=api_key,
prompt=base_prompt,
model_name=model_name,
model_provider_name=model_provider_name,
n_instructions=n_instructions,
safety_settings=safety_settings,
monitor=monitor,
llm_create_extras=llm_create_extras,
)
[docs]
def generate(self, original_prompt):
"""Generate instructions using ``original_prompt`` as the user message (no context block)."""
model = self._create_model()
return self._execute_generation(
model=model,
prompt=original_prompt,
full_context="",
context_id=None,
context_ids=[],
)
[docs]
async def agenerate(self, original_prompt):
"""Async variant of :meth:`generate`."""
model = self._create_model()
return await self._aexecute_generation(
model=model,
prompt=original_prompt,
full_context="",
context_id=None,
context_ids=[],
)