Implementation:Microsoft Agent framework Edge Condition Function Pattern

Property	Value
Implementation Name	Edge Condition Function Pattern
SDK	Microsoft Agent Framework
Repository	Microsoft Agent Framework
Type	Pattern Doc (user-defined interface)
Domains	Workflow_Engine, Control_Flow
Interface	Callable[[Any], bool | Awaitable[bool]]

Overview

The Edge Condition Function Pattern describes the user-defined callable interface used to control conditional edge traversal in Microsoft Agent Framework workflows. An edge condition function receives the message data produced by the source executor and returns a boolean (True to traverse the edge, False to skip it). Both synchronous and asynchronous condition functions are supported.

This is a Pattern Doc -- it documents a user-defined interface rather than a specific framework API. The framework defines the contract; users provide the implementation.

Interface Contract

Callable Signature

Callable[[Any], bool | Awaitable[bool]]

Synchronous Form

def condition(data: Any) -> bool:
    ...

Asynchronous Form

async def condition(data: Any) -> bool:
    ...

Parameter Details

Parameter	Type	Description
data	Any	The message data emitted by the source executor. In most workflow configurations this is an AgentExecutorResponse, but the type depends on the source executor's output contract. The condition function should be typed to match the expected source output.

Return Value

Return Value	Effect
True	The edge is traversed and the target executor is scheduled for execution.
False	The edge is skipped; the target executor is not reached via this edge.

Data Flow

The condition function sits between a source executor and a target executor in the workflow graph:

Step	Component	Action
1	Source Executor	Completes execution and produces output data.
2	Workflow Engine	Passes the source output to each outgoing edge's condition function.
3	Condition Function	Inspects the data and returns True or False.
4	Workflow Engine	Traverses edges where the condition returned True; skips edges where it returned False.
5	Target Executor	Receives the source output and begins execution (only if the edge was traversed).

Usage Examples

Simple Value Check

Inspect a boolean field on the executor's structured output to determine whether to proceed:

from agent_framework.workflows import AgentExecutorResponse

def is_approved(data: AgentExecutorResponse) -> bool:
    return data.agent_response.value.approved == True

Pydantic Structured Output Check

When the source agent uses a Pydantic model as its structured output type, the condition function can inspect typed fields directly:

from pydantic import BaseModel
from agent_framework.workflows import AgentExecutorResponse

class Decision(BaseModel):
    action: str

def route_to_action(data: AgentExecutorResponse) -> bool:
    return data.agent_response.value.action == "process"

Async Condition with External Lookup

When the routing decision depends on external state (such as a database or API), use an asynchronous condition function:

async def check_database(data: str) -> bool:
    result = await db.check(data)
    return result.is_valid

Composing Multiple Conditions

Standard Python techniques can be used to combine conditions into more complex predicates:

from agent_framework.workflows import AgentExecutorResponse

def is_approved(data: AgentExecutorResponse) -> bool:
    return data.agent_response.value.approved == True

def is_high_priority(data: AgentExecutorResponse) -> bool:
    return data.agent_response.value.priority >= 5

def is_approved_and_high_priority(data: AgentExecutorResponse) -> bool:
    return is_approved(data) and is_high_priority(data)

Implementation Guidelines

1. Type the data parameter -- Use the specific type produced by the source executor (e.g., AgentExecutorResponse) rather than Any to gain type-checking benefits.
2. Return explicit booleans -- Always return True or False. Avoid relying on Python's truthy/falsy coercion, which can lead to subtle routing bugs.
3. Guard against missing attributes -- If the source executor's output structure may vary, use defensive access patterns (e.g., getattr, hasattr, or try/except) inside the condition function.
4. Minimize side effects -- Condition functions should ideally be pure predicates. If side effects are necessary (e.g., logging), ensure they do not alter the data or the workflow state.
5. Keep async conditions lightweight -- Asynchronous conditions introduce I/O latency into the routing path. Use them only when external lookups are genuinely required for the routing decision.

Error Handling

Scenario	Behavior
Condition raises an exception	The workflow engine propagates the exception. Users should handle errors within the condition function to avoid workflow-level failures.
Condition returns a non-boolean value	Behavior depends on the workflow engine's coercion rules. It is strongly recommended to return an explicit bool.

Related Pages

• Principle:Microsoft_Agent_framework_Edge_Condition_Pattern -- The principle-level documentation for the Edge Condition Pattern.