Tools

LLM tool factories that expose TAC Memory and Knowledge to each connector.

Agent Framework tools

tac_microsoft.agent_framework_tools

TAC tools for the Agent Framework connector.

Each factory returns a plain async callable that Agent Framework auto-discovers from function name, docstring, and type annotations.

Usage::

from tac_microsoft.agent_framework_tools import (
    create_memory_tool,
    create_knowledge_tool,
)

tools = [
    create_memory_tool(tac, session),
    await create_knowledge_tool(tac, kb_id),
]
agent = client.as_agent(name="Agent", tools=[t for t in tools if t])

KnowledgeBaseInfo dataclass

KnowledgeBaseInfo(name: str, description: str)

Metadata fetched from a knowledge base.

fetch_knowledge_base_info async

fetch_knowledge_base_info(
    tac: TAC, knowledge_base_id: str
) -> KnowledgeBaseInfo

Fetch name and description from a knowledge base.

Useful when you want the LLM tool name/description to match the knowledge base configuration without hardcoding them.

Parameters:

Name	Type	Description	Default
tac	TAC	TAC instance.	required
knowledge_base_id	str	The knowledge base ID (format: know_knowledgebase_*).	required

Returns:

Type	Description
KnowledgeBaseInfo	class:KnowledgeBaseInfo with name and description.

Raises:

Type	Description
ValueError	If tac.knowledge_client is not initialised.

create_memory_tool

create_memory_tool(
    tac: TAC,
    session: ConversationSession,
    *,
    name: str | None = None,
    description: str | None = None,
) -> Any | None

Create a memory tool for Agent Framework.

Returns a plain async function, or None if prerequisites are not met (no memory client or profile_id).

create_knowledge_tool async

create_knowledge_tool(
    tac: TAC,
    knowledge_base_id: str,
    *,
    name: str | None = None,
    description: str | None = None,
    top_k: int = 5,
) -> Any

Create a knowledge base search tool for Agent Framework.

If name or description is omitted, fetches the knowledge base metadata to derive defaults. Returns a plain async function.

create_handoff_tool

create_handoff_tool(
    tac: TAC,
    session: ConversationSession,
    attributes: dict[str, Any] | None = None,
) -> Any

Create a Studio handoff tool for Agent Framework.

Returns a plain async function. Requires tac.config.studio_handoff_flow_sid to be set.

interstitial_filler async

interstitial_filler(
    filler_words: str,
) -> AsyncGenerator[dict[str, Any], None]

Provide a short, conversational filler sentence to fill dead air latency.

Use this tool to speak brief filler words while waiting for other tool results to keep the conversation flowing naturally.

Parameters:

Name	Type	Description	Default
filler_words	str	The creative, context-aware filler sentence to speak while waiting for tool results.	required

Voice Live tools

tac_microsoft.voice_live_tools

TAC tools for the Voice Live connector.

Each factory returns a TACTool instance that can be passed directly to :class:VoiceLiveConfig — no conversion step needed.

Usage::

from tac_microsoft.voice_live_tools import (
    create_memory_tool,
    create_knowledge_tool,
)

tools = [
    create_memory_tool(tac, session),
    await create_knowledge_tool(tac, kb_id),
    my_custom_tac_tool,  # TACTool from @function_tool
]

config = VoiceLiveConfig(..., tools=[t for t in tools if t])

KnowledgeBaseInfo dataclass

KnowledgeBaseInfo(name: str, description: str)

Metadata fetched from a knowledge base.

fetch_knowledge_base_info async

fetch_knowledge_base_info(
    tac: TAC, knowledge_base_id: str
) -> KnowledgeBaseInfo

Fetch name and description from a knowledge base.

Useful when you want the LLM tool name/description to match the knowledge base configuration without hardcoding them.

Parameters:

Name	Type	Description	Default
tac	TAC	TAC instance.	required
knowledge_base_id	str	The knowledge base ID (format: know_knowledgebase_*).	required

Returns:

Type	Description
KnowledgeBaseInfo	class:KnowledgeBaseInfo with name and description.

Raises:

Type	Description
ValueError	If tac.knowledge_client is not initialised.

create_memory_tool

create_memory_tool(
    tac: TAC,
    session: ConversationSession,
    *,
    name: str | None = None,
    description: str | None = None,
) -> TACTool | None

Create a memory TACTool for Voice Live.

Returns None if prerequisites are not met.

create_knowledge_tool async

create_knowledge_tool(
    tac: TAC,
    knowledge_base_id: str,
    *,
    name: str | None = None,
    description: str | None = None,
    top_k: int = 5,
) -> TACTool

Create a knowledge base search TACTool for Voice Live.

If name or description is omitted, fetches the knowledge base metadata to derive defaults.

create_handoff_tool

create_handoff_tool(
    tac: TAC,
    session: ConversationSession,
    attributes: dict[str, Any] | None = None,
) -> TACTool

Create a Studio handoff TACTool for Voice Live.

Requires tac.config.studio_handoff_flow_sid to be set.