Semantic Kernel (Microsoft) integration

Astrocyte as a kernel plugin for Microsoft Semantic Kernel.

Module: astrocyte.integrations.semantic_kernel Pattern: Plugin with kernel functions — retain, recall, reflect, forget methods Framework dependency: semantic-kernel (optional)

Install

pip install astrocyte semantic-kernel

Usage

from astrocyte import Astrocyte
from astrocyte.integrations.semantic_kernel import AstrocytePlugin

brain = Astrocyte.from_config("astrocyte.yaml")
plugin = AstrocytePlugin(brain, bank_id="user-123")

# Register with Semantic Kernel
from semantic_kernel import Kernel
kernel = Kernel()
kernel.add_plugin(plugin, plugin_name="memory")

# Call functions directly
result = await plugin.retain("Calvin prefers dark mode", tags="preference,ui")
result = await plugin.recall("UI preferences")
answer = await plugin.reflect("What are Calvin's preferences?")

Integration pattern

Each method follows the Semantic Kernel @kernel_function convention: named method with docstring, type-annotated parameters, returns string.

Plugin method	Astrocyte call	Returns
retain(content, tags?)	brain.retain()	"Stored memory (id: ...)"
recall(query, max_results?)	brain.recall()	Formatted list of scored hits
reflect(query)	brain.reflect()	Synthesized answer
forget(memory_ids)	brain.forget()	"Deleted N memories."

End-to-end example

A Semantic Kernel agent that builds up knowledge about a project:

import asyncio
from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion
from astrocyte import Astrocyte
from astrocyte.integrations.semantic_kernel import AstrocytePlugin

brain = Astrocyte.from_config("astrocyte.yaml")
plugin = AstrocytePlugin(brain, bank_id="project-alpha")

kernel = Kernel()
kernel.add_plugin(plugin, plugin_name="memory")
kernel.add_service(AzureChatCompletion(
    deployment_name="gpt-4o",
    endpoint=endpoint,
    api_key=api_key,
))

async def main():
    # Store project facts
    await plugin.retain(
        "Project Alpha uses FastAPI with PostgreSQL and Redis caching",
        tags="architecture,tech-stack",
    )
    await plugin.retain(
        "Deployment target is AWS EKS with Helm charts",
        tags="architecture,deployment",
    )

    # Recall specific knowledge
    results = await plugin.recall("database setup", max_results=3)
    print(results)

    # Synthesize across all project knowledge
    answer = await plugin.reflect("Summarize the project architecture")
    print(answer)

asyncio.run(main())

Using with planner

Semantic Kernel’s planner can auto-select memory functions:

from semantic_kernel.planners import FunctionCallingStepwisePlanner

planner = FunctionCallingStepwisePlanner(kernel)
result = await planner.invoke(
    "What do we know about the deployment architecture? "
    "Check memory first, then summarize."
)

API reference

AstrocytePlugin(brain, bank_id, *, include_reflect=True, include_forget=False)

Methods: retain(), recall(), reflect(), forget(), get_functions().

get_functions() returns a list of {"name", "description", "function"} dicts for programmatic registration.