Semantic Kernel Integration¶

Added in v3.11.0

Director-AI provides a function invocation filter for Microsoft Semantic Kernel.

Setup¶

from semantic_kernel import Kernel
from director_ai.integrations.semantic_kernel import DirectorAIFilter

kernel = Kernel()
kernel.add_filter("function_invocation", DirectorAIFilter(
    facts={"pricing": "Team plan costs $19/user/month."},
    threshold=0.5,
))

How It Works¶

The filter runs after each function invocation. If the LLM output coherence is below the threshold, it either raises HallucinationError (default) or annotates the result with approved=False.

Parameters¶

Parameter	Type	Default	Description
`facts`	`dict[str, str]`	`None`	Key-value facts for the knowledge base
`store`	`GroundTruthStore`	`None`	Pre-built store (overrides facts)
`threshold`	`float`	`0.5`	Minimum coherence to pass
`use_nli`	`bool \| None`	`None`	NLI mode (None=auto-detect)
`raise_on_fail`	`bool`	`True`	Raise on failure vs annotate result

Non-Raising Mode¶

filter = DirectorAIFilter(
    facts={"pricing": "Team plan costs $19/user/month."},
    raise_on_fail=False,
)
# Result will be a dict with approved=False instead of raising