mirror of
https://github.com/crewAIInc/crewAI.git
synced 2026-01-22 06:18:14 +00:00
9945da7dbe
Some checks failed
Notify Downstream / notify-downstream (push) Has been cancelled
- Add `HallucinationGuardrail` class as enterprise feature placeholder - Update LLM guardrail events to support `HallucinationGuardrail` instances - Add comprehensive tests for `HallucinationGuardrail` initialization and behavior - Add integration tests for `HallucinationGuardrail` with task execution system - Ensure no-op behavior always returns True
97 lines
3.4 KiB
Python
97 lines
3.4 KiB
Python
"""Hallucination Guardrail Placeholder for CrewAI.
|
|
|
|
This is a no-op version of the HallucinationGuardrail for the open-source repository.
|
|
|
|
Classes:
|
|
HallucinationGuardrail: Placeholder guardrail that validates task outputs.
|
|
"""
|
|
|
|
from typing import Any, Optional, Tuple
|
|
|
|
from crewai.llm import LLM
|
|
from crewai.tasks.task_output import TaskOutput
|
|
from crewai.utilities.logger import Logger
|
|
|
|
|
|
class HallucinationGuardrail:
|
|
"""Placeholder for the HallucinationGuardrail feature.
|
|
|
|
Attributes:
|
|
context: The reference context that outputs would be checked against.
|
|
llm: The language model that would be used for evaluation.
|
|
threshold: Optional minimum faithfulness score that would be required to pass.
|
|
tool_response: Optional tool response information that would be used in evaluation.
|
|
|
|
Examples:
|
|
>>> # Basic usage with default verdict logic
|
|
>>> guardrail = HallucinationGuardrail(
|
|
... context="AI helps with various tasks including analysis and generation.",
|
|
... llm=agent.llm
|
|
... )
|
|
|
|
>>> # With custom threshold for stricter validation
|
|
>>> strict_guardrail = HallucinationGuardrail(
|
|
... context="Quantum computing uses qubits in superposition.",
|
|
... llm=agent.llm,
|
|
... threshold=8.0 # Would require score >= 8 to pass in enterprise version
|
|
... )
|
|
|
|
>>> # With tool response for additional context
|
|
>>> guardrail_with_tools = HallucinationGuardrail(
|
|
... context="The current weather data",
|
|
... llm=agent.llm,
|
|
... tool_response="Weather API returned: Temperature 22°C, Humidity 65%"
|
|
... )
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
context: str,
|
|
llm: LLM,
|
|
threshold: Optional[float] = None,
|
|
tool_response: str = "",
|
|
):
|
|
"""Initialize the HallucinationGuardrail placeholder.
|
|
|
|
Args:
|
|
context: The reference context that outputs would be checked against.
|
|
llm: The language model that would be used for evaluation.
|
|
threshold: Optional minimum faithfulness score that would be required to pass.
|
|
tool_response: Optional tool response information that would be used in evaluation.
|
|
"""
|
|
self.context = context
|
|
self.llm: LLM = llm
|
|
self.threshold = threshold
|
|
self.tool_response = tool_response
|
|
self._logger = Logger(verbose=True)
|
|
self._logger.log(
|
|
"warning",
|
|
"""Hallucination detection is a no-op in open source, use it for free at https://app.crewai.com\n""",
|
|
color="red",
|
|
)
|
|
|
|
@property
|
|
def description(self) -> str:
|
|
"""Generate a description of this guardrail for event logging."""
|
|
return "HallucinationGuardrail (no-op)"
|
|
|
|
def __call__(self, task_output: TaskOutput) -> Tuple[bool, Any]:
|
|
"""Validate a task output against hallucination criteria.
|
|
|
|
In the open source, this method always returns that the output is valid.
|
|
|
|
Args:
|
|
task_output: The output to be validated.
|
|
|
|
Returns:
|
|
A tuple containing:
|
|
- True
|
|
- The raw task output
|
|
"""
|
|
self._logger.log(
|
|
"warning",
|
|
"Premium hallucination detection skipped (use for free at https://app.crewai.com)\n",
|
|
color="red",
|
|
)
|
|
return True, task_output.raw
|