io.moderne.ai.research.GetRecommendations Maven / Gradle / Ivy
/*
* Copyright 2024 the original author or authors.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* https://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package io.moderne.ai.research;
import io.moderne.ai.AgentGenerativeModelClient;
import io.moderne.ai.ClusteringClient;
import io.moderne.ai.EmbeddingModelClient;
import io.moderne.ai.table.Recommendations;
import lombok.EqualsAndHashCode;
import lombok.Value;
import org.jspecify.annotations.Nullable;
import org.openrewrite.ExecutionContext;
import org.openrewrite.Option;
import org.openrewrite.ScanningRecipe;
import org.openrewrite.TreeVisitor;
import org.openrewrite.java.JavaIsoVisitor;
import org.openrewrite.java.tree.J;
import org.openrewrite.java.tree.JavaSourceFile;
import java.util.ArrayList;
import java.util.List;
import java.util.Random;
import java.util.stream.Collectors;
@Value
@EqualsAndHashCode(callSuper = false)
public class GetRecommendations extends ScanningRecipe {
@Option(displayName = "random sampling",
description = "Do random sampling or use clusters based on embeddings to sample.")
@Nullable
Boolean randomSampling;
@Option(displayName = "number of centers to sample"
, description = "Number of diverse centers to sample if you don't do random sampling per repository.",
example = "3",
required = false)
int numberOfCenters;
transient Recommendations recommendationsTable = new Recommendations(this);
private static final Random random = new Random(13);
@Override
public String getDisplayName() {
return "Get recommendations";
}
@Override
public String getDescription() {
return "This recipe calls an AI model to get recommendations for modernizing" +
" the code base by looking at a sample of method declarations.";
}
@Value
public class Method {
String method;
String name;
String file;
}
public class Accumulator {
List methods = new ArrayList<>();
List embeddings = new ArrayList<>();
@Nullable
int[] centers;
public int[] getCenters(int numberOfCenters) {
if (this.centers == null) {
this.centers = ClusteringClient.getInstance().getCenters(this.embeddings, numberOfCenters);
}
return this.centers;
}
public Method[] getMethodsToSample(int numberOfCenters) {
int[] centersIndex = getCenters(numberOfCenters);
Method[] methodsToSample = new Method[centersIndex.length];
for (int i = 0; i < getCenters(numberOfCenters).length; i++) {
methodsToSample[i] = this.methods.get(centersIndex[i]);
}
return methodsToSample;
}
public void addMethodToSample(String method, String methodName, String file) {
this.methods.add(new Method(method, methodName, file));
this.embeddings.add(EmbeddingModelClient.getInstance().getEmbedding(method));
}
}
@Override
public Accumulator getInitialValue(ExecutionContext ctx) {
return new Accumulator();
}
@Override
public TreeVisitor getScanner(Accumulator acc) {
return new JavaIsoVisitor() {
@Override
public J.MethodDeclaration visitMethodDeclaration(J.MethodDeclaration method, ExecutionContext ctx) {
J.MethodDeclaration md = super.visitMethodDeclaration(method, ctx);
String methodName = md.getSimpleName();
JavaSourceFile javaSourceFile = getCursor().firstEnclosing(JavaSourceFile.class);
String source = javaSourceFile.getSourcePath().toString();
acc.addMethodToSample(md.printTrimmed(getCursor()), methodName, source);
return md;
}
};
}
@Override
public TreeVisitor getVisitor(Accumulator acc) {
return new JavaIsoVisitor() {
@Override
public J.MethodDeclaration visitMethodDeclaration(J.MethodDeclaration method, ExecutionContext ctx) {
J.MethodDeclaration md = super.visitMethodDeclaration(method, ctx);
boolean isMethodToSample = false;
JavaSourceFile javaSourceFile = getCursor().firstEnclosing(JavaSourceFile.class);
String source = javaSourceFile.getSourcePath().toString();
boolean randomSampling = getRandomSampling() != null ? getRandomSampling() : false;
if (randomSampling) {
isMethodToSample = random.nextInt(200) <= 1;
} else {
for (Method methodToSample : acc.getMethodsToSample(numberOfCenters)) {
if (methodToSample.file.equals(source) &&
methodToSample.name.equals(md.getSimpleName()) &&
methodToSample.method.equals(md.printTrimmed(getCursor()))) {
isMethodToSample = true;
break;
}
}
}
if (isMethodToSample) { // samples based on the results from running GetCodeEmbedding and clustering
long time = System.nanoTime();
// Get recommendations
List recommendations;
recommendations = AgentGenerativeModelClient.getInstance().getRecommendations(md.printTrimmed(getCursor()));
List recommendationsQuoted = recommendations.stream()
.map(element -> "\"" + element + "\"")
.collect(Collectors.toList());
String recommendationsAsString = "[" + String.join(", ", recommendationsQuoted) + "]";
int tokenSize = (int) ((md.printTrimmed(getCursor())).length() / 3.5 + recommendations.toString().length() / 3.5);
double elapsedTime = (System.nanoTime() - time) / 1e9;
recommendationsTable.insertRow(ctx, new Recommendations.Row(md.getSimpleName(),
elapsedTime,
tokenSize,
recommendationsAsString));
}
return md;
}
};
}
}