AxACE

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L139

AxACE implements the Agentic Context Engineering loop (Generator → Reflector → Curator). The implementation mirrors the paper’s architecture while integrating with the Ax optimizer ergonomics (unified optimized program artifacts, metrics, and checkpointing).

Extends

• AxBaseOptimizer

Constructors

Constructor

new AxACE(args: Readonly<AxOptimizerArgs>, options?: Readonly<AxACEOptions>): AxACE;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L151

Parameters

Returns

AxACE

Overrides

AxBaseOptimizer.constructor

Methods

applyOnlineUpdate()

applyOnlineUpdate(args: Readonly<{
  example: AxExample;
  feedback?: string;
  prediction: unknown;
}>): Promise<undefined | AxACECuratorOutput>;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L454

Apply ACE updates after each online inference. Mirrors the online adaptation flow described in the paper; can be called by user-land code between queries.

Parameters

Returns

Promise<undefined | AxACECuratorOutput>

compile()

compile<IN, OUT>(
   program: Readonly<AxGen<IN, OUT>>, 
   examples: readonly AxTypedExample<IN>[], 
   metricFn: AxMetricFn, 
options?: AxACECompileOptions): Promise<AxACEResult<OUT>>;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L195

Abstract method that must be implemented by concrete optimizers

Type Parameters

Parameters

Returns

Promise<AxACEResult<OUT>>

Overrides

AxBaseOptimizer.compile

compilePareto()

compilePareto<IN, OUT>(
   program: Readonly<AxGen<IN, OUT>>, 
   examples: readonly AxTypedExample<IN>[], 
   metricFn: AxMultiMetricFn, 
options?: AxCompileOptions): Promise<AxParetoResult<OUT>>;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizer.ts#L1561

Multi-objective optimization using Pareto frontier Default implementation that leverages the single-objective compile method

Type Parameters

Parameters

Returns

Promise<AxParetoResult<OUT>>

Pareto optimization result with frontier of non-dominated solutions

Inherited from

AxBaseOptimizer.compilePareto

compileStream()

compileStream<IN, OUT>(
   program: Readonly<AxGen<IN, OUT>>, 
   examples: readonly AxTypedExample<IN>[], 
   metricFn: AxMetricFn, 
options?: AxCompileOptions): AsyncIterableIterator<AxOptimizationProgress>;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizer.ts#L1448

Optimize a program with real-time streaming updates

Type Parameters

Parameters

Returns

AsyncIterableIterator<AxOptimizationProgress>

Async iterator yielding optimization progress

Inherited from

AxBaseOptimizer.compileStream

configureAuto()

configureAuto(level: "light" | "medium" | "heavy"): void;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L178

Parameters

Returns

void

getStats()

getStats(): AxOptimizationStats;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizer.ts#L2494

Get current optimization statistics

Returns

AxOptimizationStats

Current optimization statistics

Inherited from

AxBaseOptimizer.getStats

reset()

reset(): void;

Defined in: https://github.com/ax-llm/ax/blob/0aa4aa2ceed1ba61106711baed6ce962cf2eb604/src/ax/dsp/optimizers/ace.ts#L168

Reset optimizer state for reuse with different programs

Returns

void

Overrides

AxBaseOptimizer.reset