Package com.tnsai.refinement

Class RefinementLoop

java.lang.Object

com.tnsai.refinement.RefinementLoop

public final class RefinementLoop
extends Object

Iterative refinement loop for persistent output improvement.

Inspired by Claude's Ralph plugin, this executor repeatedly processes outputs until they meet predefined standards through iterative refinement.

Usage

 RefinementLoop loop = RefinementLoop.builder()
     .task("Convert Python to TypeScript")
     .completionCriteria(CompletionCriteria.builder()
         .compilerCheck("tsc --noEmit")
         .testCommand("npm test")
         .mustNotContain("def ", "import ")
         .build())
     .maxIterations(10)
     .stopHook(StopHook.onAllTestsPass())
     .onIteration(iter -> log.info("Iteration {}", iter.iterationNumber()))
     .build();

 RefinementResult result = loop.execute(agent, pythonCode);
 System.out.println(result.getFinalOutput());
 

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final class	RefinementLoop.Builder	Builder for RefinementLoop.

Method Summary

Modifier and Type	Method	Description
static RefinementLoop.Builder	builder()	Creates a new builder for RefinementLoop.
RefinementResult	execute(com.tnsai.agents.Agent agent, String input)	Executes the refinement loop with an agent.
RefinementResult	execute(com.tnsai.llm.LLMClient llm, String input)	Executes the refinement loop with an LLM client.

Methods inherited from class java.lang.Object

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

Method Details

builder

public static RefinementLoop.Builder builder()

Creates a new builder for RefinementLoop.

execute

public RefinementResult execute(com.tnsai.agents.Agent agent,
 String input)

Executes the refinement loop with an agent.

Parameters:

agent - The agent to use for refinement

input - The initial input to refine

Returns:

The refinement result

execute

public RefinementResult execute(com.tnsai.llm.LLMClient llm,
 String input)

Executes the refinement loop with an LLM client.