Skip to content
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

add more information into the eval outputs #96

Merged
merged 1 commit into from
Oct 3, 2024
Merged

add more information into the eval outputs #96

Show file tree
Hide file tree
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
34 changes: 28 additions & 6 deletions evals/bananalyzer-ts/index.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ export async function evaluateExample(
launchServer: true,
serverPort: 6778,
},
): Promise<boolean> {
): Promise<any> {
await new Promise((resolve) => setTimeout(resolve, 2000));

const examples = JSON.parse(
Expand Down Expand Up @@ -162,15 +162,25 @@ export async function evaluateExample(
if (evalItem.expected) {
if (!validateJsonMatch(evalItem.expected, extractionResult)) {
console.log("❌ JSON match failed");
return false;
return {
_success: false,
case: "json_mismatch_1",
expected: evalItem.expected,
actual: extractionResult,
};
}
} else if (evalItem.options) {
const matchesAny = evalItem.options.some((option) =>
validateJsonMatch(option, extractionResult),
);
if (!matchesAny) {
console.log("❌ No JSON match found in options");
return false;
return {
_success: false,
case: "json_mismatch_2",
expected: evalItem.expected,
actual: extractionResult,
};
}
}
} else if (
Expand All @@ -181,16 +191,28 @@ export async function evaluateExample(
!validateEndUrlMatch(evalItem.expected, await stagehand.page.url())
) {
console.log("❌ URL match failed");
return false;
return {
_success: false,
case: "url_mismatch",
expected: evalItem.expected,
actual: await stagehand.page.url(),
};
}
}
}

console.log("✅ All evaluations passed");
return true;
return {
_success: true,
expected: extractionResult,
actual: extractionResult,
};
} catch (error) {
console.error("Error during evaluation:", error);
return false;
return {
_success: false,
error: error,
};
} finally {
try {
const deleteResponse = await fetch(
Expand Down
71 changes: 54 additions & 17 deletions evals/index.eval.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,11 @@ const vanta = async () => {

await stagehand.context.close();

return observationResult == expectedResult;
return {
_success: observationResult == expectedResult,
expected: expectedResult,
actual: observationResult,
};
};

const vanta_h = async () => {
Expand All @@ -59,7 +63,10 @@ const vanta_h = async () => {
await stagehand.context.close();

// we should have no saved observation since the element shouldn't exist
return observation === null;
return {
_success: observation === null,
observation,
};
};

const simple_google_search = async () => {
Expand All @@ -80,7 +87,10 @@ const simple_google_search = async () => {

await stagehand.context.close();

return currentUrl.startsWith(expectedUrl);
return {
_success: currentUrl.startsWith(expectedUrl),
currentUrl,
};
};

const peeler_simple = async () => {
Expand All @@ -101,7 +111,9 @@ const peeler_simple = async () => {
const isVisible = await successMessageLocator.isVisible();

await stagehand.context.close();
return isVisible;
return {
_success: isVisible,
};
};

const peeler_complex = async () => {
Expand Down Expand Up @@ -130,7 +142,10 @@ const peeler_complex = async () => {

await stagehand.context.close();

return price !== null;
return {
_success: price !== null,
price,
};
};

const extract_collaborators_from_github_repository = async () => {
Expand Down Expand Up @@ -164,7 +179,10 @@ const extract_collaborators_from_github_repository = async () => {

console.log("Extracted collaborators:", contributors);
await stagehand.context.close();
return contributors.length === 20;
return {
_success: contributors.length === 20,
contributors,
};
} catch (error) {
console.error("Error or timeout occurred:", error);
await stagehand.context.close();
Expand Down Expand Up @@ -201,7 +219,10 @@ const extract_last_twenty_github_commits = async () => {

console.log("Extracted commits:", commits);
await stagehand.context.close();
return commits.length === 20;
return {
_success: commits.length === 20,
commits,
};
} catch (error) {
console.error("Error or timeout occurred:", error);
await stagehand.context.close();
Expand All @@ -226,7 +247,11 @@ const wikipedia = async () => {
const currentUrl = await stagehand.page.url();
await stagehand.context.close();

return currentUrl === url;
return {
_success: currentUrl === url,
expected: url,
actual: currentUrl,
};
};

const costar = async () => {
Expand Down Expand Up @@ -268,10 +293,10 @@ const costar = async () => {

await stagehand.context.close();

return isTitleValid;
return { title: articleTitle.title, _success: isTitleValid };
} catch (error) {
console.error(`Error in costar function: ${error.message}`);
return { title: null };
return { title: null, _success: false } as any;
} finally {
await stagehand.context.close();
}
Expand Down Expand Up @@ -349,7 +374,7 @@ const google_jobs = async () => {

console.log("Job Details valid:", isJobDetailsValid);

return isJobDetailsValid;
return { _success: isJobDetailsValid, jobDetails };
};

const tasks = {
Expand All @@ -365,12 +390,20 @@ const tasks = {
google_jobs,
};

const exactMatch = (args: { input; output; expected? }) => {
const exactMatch = (args: { input: any; output: any; expected?: any }) => {
console.log(`Task "${args.input.name}" returned: ${args.output}`);

const expected = args.expected ?? true;
if (expected === true) {
return {
name: "Exact match",
score: args.output === true || args.output?._success == true,
};
}

return {
name: "Exact match",
score: args.output === true || args.output?.success == true,
score: args.output === expected,
};
};

Expand All @@ -395,9 +428,13 @@ const testcases = [
},
{ input: { name: "peeler_complex" } },
{ input: { name: "simple_google_search" } },
{ input: { name: "extract_collaborators_from_github_repository" } },
{
input: {
name: "extract_collaborators_from_github_repository",
},
},
{ input: { name: "extract_last_twenty_github_commits" } },
// { input: { name: "costar" } },
// { input: { name: "costar", expected: true } },
{ input: { name: "google_jobs" } },
...chosenBananalyzerEvals.map((evalItem: any) => ({
input: {
Expand All @@ -416,7 +453,7 @@ Eval("stagehand", {
data: () => {
return testcases;
},
task: async (input) => {
task: async (input: any) => {
// console.log("input", input);
try {
if ("source" in input && input.source === "bananalyzer-ts") {
Expand All @@ -440,7 +477,7 @@ Eval("stagehand", {
return result;
} else {
// Handle predefined tasks
const result = await tasks[input.name](input);
const result = await (tasks as any)[input.name](input);
if (result) {
console.log(`✅ ${input.name}: Passed`);
} else {
Expand Down