Package com.tnsai.agents.variant

Class VariantSelector

java.lang.Object

com.tnsai.agents.variant.VariantSelector

public class VariantSelector
extends Object

Intelligent variant selector for automatic quality tier selection.

Combines multiple factors to select the optimal variant:

• Task Complexity: Analyzes input for complexity signals
• Historical Performance: Tracks success/failure rates per variant
• Cost Budget: Respects token budget constraints
• User Preferences: Learns from explicit feedback

Author:

TnsAI Team

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	VariantSelector.Builder	Builder for VariantSelector.
static enum	VariantSelector.SelectionReason	Reasons for variant selection.
static final record	VariantSelector.SelectionResult	Selection result with reasoning.
static class	VariantSelector.TokenBudget	Token budget for cost control.
static enum	VariantSelector.VariantPreference	User preference for quality/speed/cost trade-offs.
static class	VariantSelector.VariantStats	Statistics for a variant.

Method Summary

Modifier and Type	Method	Description
static VariantSelector.Builder	builder()	Creates a new builder.
static VariantSelector	defaults()	Creates a default selector with balanced settings.
Map<AgentVariant,VariantSelector.VariantStats>	getAllStats()	Gets all variant statistics.
long	getRemainingBudget()	Gets remaining budget.
VariantSelector.VariantStats	getStats(AgentVariant variant)	Gets statistics for a specific variant.
void	recordFailure(AgentVariant variant, int tokensUsed)	Records a failed execution for learning.
void	recordSuccess(AgentVariant variant, int tokensUsed)	Records a successful execution for learning.
void	resetStats()	Resets learning statistics.
VariantSelector.SelectionResult	select(String taskDescription)	Selects the optimal variant for the given task.
VariantSelector.SelectionResult	select(String taskDescription, AgentVariant override)	Selects the optimal variant considering action metadata.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

builder

public static VariantSelector.Builder builder()

Creates a new builder.

Returns:

builder instance

defaults

public static VariantSelector defaults()

Creates a default selector with balanced settings.

Returns:

default selector

select

public VariantSelector.SelectionResult select(String taskDescription)

Selects the optimal variant for the given task.

Parameters:

taskDescription - the task description

Returns:

selected variant

select

public VariantSelector.SelectionResult select(String taskDescription,
 AgentVariant override)

Selects the optimal variant considering action metadata.

Parameters:

taskDescription - the task description

override - explicit variant override (from @Variant annotation)

Returns:

selected variant

recordSuccess

public void recordSuccess(AgentVariant variant,
 int tokensUsed)

Records a successful execution for learning.

Parameters:

variant - the variant used

tokensUsed - tokens consumed

recordFailure

public void recordFailure(AgentVariant variant,
 int tokensUsed)

Records a failed execution for learning.

Parameters:

variant - the variant used

tokensUsed - tokens consumed before failure

getStats

public VariantSelector.VariantStats getStats(AgentVariant variant)

Gets statistics for a specific variant.

Parameters:

variant - the variant

Returns:

statistics

getAllStats

public Map<AgentVariant,VariantSelector.VariantStats> getAllStats()

Gets all variant statistics.

Returns:

map of variant to stats

getRemainingBudget

public long getRemainingBudget()

Gets remaining budget.

Returns:

remaining tokens, or -1 if no budget set

resetStats

public void resetStats()

Resets learning statistics.