Skip to content

Move automodel version split further up #2676

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
charisk merged 1 commit into from
Aug 7, 2023
Merged

Move automodel version split further up #2676

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
263 changes: 144 additions & 119 deletions extensions/ql-vscode/src/data-extensions-editor/data-extensions-editor-view.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -160,15 +160,20 @@ export class DataExtensionsEditorView extends AbstractWebview<

break;
case "generateExternalApiFromLlm":
await this.generateModeledMethodsFromLlm(
msg.externalApiUsages,
msg.modeledMethods,
);

if (useLlmGenerationV2()) {
await this.generateModeledMethodsFromLlmV2(
msg.externalApiUsages,
msg.modeledMethods,
);
} else {
await this.generateModeledMethodsFromLlmV1(
msg.externalApiUsages,
msg.modeledMethods,
);
}
break;
case "modelDependency":
await this.modelDependency();

break;
case "switchMode":
this.mode = msg.mode;
Expand Down Expand Up @@ -362,7 +367,7 @@ export class DataExtensionsEditorView extends AbstractWebview<
);
}

private async generateModeledMethodsFromLlm(
private async generateModeledMethodsFromLlmV1(
externalApiUsages: ExternalApiUsage[],
modeledMethods: Record<string, ModeledMethod>,
): Promise<void> {
Expand All @@ -375,135 +380,155 @@ export class DataExtensionsEditorView extends AbstractWebview<
message: "Retrieving usages",
});

let predictedModeledMethods: Record<string, ModeledMethod>;
const usages = await getAutoModelUsages({
cliServer: this.cliServer,
queryRunner: this.queryRunner,
queryStorageDir: this.queryStorageDir,
queryDir: this.queryDir,
databaseItem: this.databaseItem,
progress: (update) => progress({ ...update, maxStep }),
});

if (useLlmGenerationV2()) {
// Fetch the candidates to send to the model
const candidateMethods = getCandidates(
this.mode,
externalApiUsages,
modeledMethods,
);
progress({
step: 1800,
maxStep,
message: "Creating request",
});

// If there are no candidates, there is nothing to model and we just return
if (candidateMethods.length === 0) {
void extLogger.log("No candidates to model. Stopping.");
return;
}
const request = createAutoModelRequest(
this.databaseItem.language,
externalApiUsages,
modeledMethods,
usages,
this.mode,
);

const usages = await runAutoModelQueries({
mode: this.mode,
candidateMethods,
cliServer: this.cliServer,
queryRunner: this.queryRunner,
queryStorageDir: this.queryStorageDir,
databaseItem: this.databaseItem,
progress: (update) => progress({ ...update, maxStep }),
});
if (!usages) {
return;
}
progress({
step: 2000,
maxStep,
message: "Sending request",
});

progress({
step: 1800,
maxStep,
message: "Creating request",
});
const response = await this.callAutoModelApi(request);
if (!response) {
return;
}

const request = await createAutoModelV2Request(this.mode, usages);
progress({
step: 2500,
maxStep,
message: "Parsing response",
});

progress({
step: 2000,
maxStep,
message: "Sending request",
});
const predictedModeledMethods = parsePredictedClassifications(
response.predicted || [],
);

const response = await this.callAutoModelApiV2(request);
if (!response) {
return;
}
progress({
step: 2800,
maxStep,
message: "Applying results",
});

progress({
step: 2500,
maxStep,
message: "Parsing response",
});
await this.postMessage({
t: "addModeledMethods",
modeledMethods: predictedModeledMethods,
});
});
}

const models = loadYaml(response.models, {
filename: "auto-model.yml",
});
private async generateModeledMethodsFromLlmV2(
externalApiUsages: ExternalApiUsage[],
modeledMethods: Record<string, ModeledMethod>,
): Promise<void> {
await withProgress(async (progress) => {
const maxStep = 3000;

const loadedMethods = loadDataExtensionYaml(models);
if (!loadedMethods) {
return;
}
progress({
step: 0,
maxStep,
message: "Retrieving usages",
});

// Any candidate that was part of the response is a negative result
// meaning that the canidate is not a sink for the kinds that the LLM is checking for.
// For now we model this as a sink neutral method, however this is subject
// to discussion.
for (const candidate of candidateMethods) {
if (!(candidate.signature in loadedMethods)) {
loadedMethods[candidate.signature] = {
type: "neutral",
kind: "sink",
input: "",
output: "",
provenance: "ai-generated",
signature: candidate.signature,
packageName: candidate.packageName,
typeName: candidate.typeName,
methodName: candidate.methodName,
methodParameters: candidate.methodParameters,
};
}
}
// Fetch the candidates to send to the model
const candidateMethods = getCandidates(
this.mode,
externalApiUsages,
modeledMethods,
);

predictedModeledMethods = loadedMethods;
} else {
const usages = await getAutoModelUsages({
cliServer: this.cliServer,
queryRunner: this.queryRunner,
queryStorageDir: this.queryStorageDir,
queryDir: this.queryDir,
databaseItem: this.databaseItem,
progress: (update) => progress({ ...update, maxStep }),
});
// If there are no candidates, there is nothing to model and we just return
if (candidateMethods.length === 0) {
void extLogger.log("No candidates to model. Stopping.");
return;
}

progress({
step: 1800,
maxStep,
message: "Creating request",
});
const usages = await runAutoModelQueries({
mode: this.mode,
candidateMethods,
cliServer: this.cliServer,
queryRunner: this.queryRunner,
queryStorageDir: this.queryStorageDir,
databaseItem: this.databaseItem,
progress: (update) => progress({ ...update, maxStep }),
});
if (!usages) {
return;
}

const request = createAutoModelRequest(
this.databaseItem.language,
externalApiUsages,
modeledMethods,
usages,
this.mode,
);
progress({
step: 1800,
maxStep,
message: "Creating request",
});

progress({
step: 2000,
maxStep,
message: "Sending request",
});
const request = await createAutoModelV2Request(this.mode, usages);

const response = await this.callAutoModelApi(request);
if (!response) {
return;
}
progress({
step: 2000,
maxStep,
message: "Sending request",
});

progress({
step: 2500,
maxStep,
message: "Parsing response",
});
const response = await this.callAutoModelApiV2(request);
if (!response) {
return;
}

predictedModeledMethods = parsePredictedClassifications(
response.predicted || [],
);
progress({
step: 2500,
maxStep,
message: "Parsing response",
});

const models = loadYaml(response.models, {
filename: "auto-model.yml",
});

const loadedMethods = loadDataExtensionYaml(models);
if (!loadedMethods) {
return;
}

// Any candidate that was part of the response is a negative result
// meaning that the canidate is not a sink for the kinds that the LLM is checking for.
// For now we model this as a sink neutral method, however this is subject
// to discussion.
for (const candidate of candidateMethods) {
if (!(candidate.signature in loadedMethods)) {
loadedMethods[candidate.signature] = {
type: "neutral",
kind: "sink",
input: "",
output: "",
provenance: "ai-generated",
signature: candidate.signature,
packageName: candidate.packageName,
typeName: candidate.typeName,
methodName: candidate.methodName,
methodParameters: candidate.methodParameters,
};
}
}

progress({
Expand All @@ -514,7 +539,7 @@ export class DataExtensionsEditorView extends AbstractWebview<

await this.postMessage({
t: "addModeledMethods",
modeledMethods: predictedModeledMethods,
modeledMethods: loadedMethods,
});
});
}
Expand Down