Skip to content

Fix malformed snapshot Unicode #2345

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
monadoid merged 3 commits into from
Jul 9, 2026
+108 −3
Merged

Fix malformed snapshot Unicode #2345

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1eb90b2
Fix malformed snapshot Unicode
Jul 9, 2026
d2c8096
Add changeset for Unicode fix
Jul 9, 2026
143ac0e
Format Unicode regression test
Jul 9, 2026
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
5 changes: 5 additions & 0 deletions .changeset/fix-malformed-snapshot-unicode.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"@browserbasehq/stagehand": patch
---

Repair malformed UTF-16 snapshot text before it reaches model prompts.
11 changes: 8 additions & 3 deletions packages/core/lib/v3/understudy/a11y/snapshot/capture.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@ type Interval = { start: number; end: number };
type ExclusionIntervalsByFrame = Map<string, Interval[]>;
type ChildFrameHost = { childFrameId: string; hostBackendNodeId: number };
type ChildFramesByParent = Map<string, ChildFrameHost[]>;
type WellFormedString = string & { toWellFormed(): string };

/**
* Capture a hybrid DOM + Accessibility snapshot for the provided page.
Expand Down Expand Up @@ -278,14 +279,16 @@ export async function tryScopedSnapshot(

const scopedUrlMap: Record<string, string> = { ...urlMap };

const wellFormedOutline = (outline as WellFormedString).toWellFormed();

const snapshot: HybridSnapshot = {
combinedTree: outline,
combinedTree: wellFormedOutline,
combinedXpathMap: scopedXpathMap,
combinedUrlMap: scopedUrlMap,
perFrame: [
{
frameId: targetFrameId,
outline,
outline: wellFormedOutline,
xpathMap,
urlMap,
},
Expand Down Expand Up @@ -846,7 +849,9 @@ export function mergeFramesIntoSnapshot(
perFrameOutlines.find((o) => o.frameId === context.rootId)?.outline ??
perFrameOutlines[0]?.outline ??
"";
const combinedTree = injectSubtrees(rootOutline, idToTree);
const combinedTree = (
injectSubtrees(rootOutline, idToTree) as WellFormedString
).toWellFormed();
Comment thread
monadoid marked this conversation as resolved.

return {
combinedTree,
Expand Down
54 changes: 54 additions & 0 deletions packages/core/tests/integration/unicode-well-formed.spec.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
import { expect, test } from "@playwright/test";
import { V3 } from "../../lib/v3/v3.js";
import {
closeV3,
createScriptedAisdkTestLlmClient,
promptToText,
} from "./testUtils.js";
import { getV3TestConfig } from "./v3.config.js";

function encodeHtml(html: string): string {
return `data:text/html,${encodeURIComponent(html)}`;
}

test("repairs page text before local SDK model calls", async () => {
let observedPromptText = "";
const llmClient = createScriptedAisdkTestLlmClient({
modelId: "mock/unicode-well-formed",
jsonResponses: {
Observation: (options) => {
observedPromptText = promptToText(options.prompt);
expect(
(
observedPromptText as string & { isWellFormed(): boolean }
).isWellFormed(),
).toBe(true);
expect(observedPromptText).toContain("Draw Again. \uFFFD");
return { elements: [] };
},
},
});

const v3 = new V3(getV3TestConfig({ llmClient }));
await v3.init();

try {
const page = v3.context.pages()[0];
await page.goto(
encodeHtml(`<!doctype html>
<meta charset="utf-8">
<h1 id="target"></h1>
<script>
document.getElementById("target").textContent =
"Draw Again. " + String.fromCharCode(0xd83c);
</script>`),
);

const observed = await v3.observe("Find the promo banner text");

expect(observed).toEqual([]);
expect(observedPromptText).not.toBe("");
} finally {
await closeV3(v3);
}
});
41 changes: 41 additions & 0 deletions packages/core/tests/unit/snapshot-frame-merge.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -346,6 +346,47 @@ describe("mergeFramesIntoSnapshot", () => {
expect(snapshot.combinedTree).toBe("[child] frame2");
});

it("repairs malformed surrogate pairs in the combined tree", () => {
const context: FrameContext = {
rootId: "frame-1",
frames: ["frame-1"],
parentByFrame: new Map([["frame-1", null]]),
};

const perFrameMaps = new Map<string, FrameDomMaps>([
[
"frame-1",
{
tagNameMap: {},
scrollableMap: {},
urlMap: {},
xpathMap: {},
},
],
]);

const snapshot = mergeFramesIntoSnapshot(
context,
perFrameMaps,
[
{
frameId: "frame-1",
outline: `Draw Again. ${String.fromCharCode(0xd83c)}`,
},
],
new Map([["frame-1", ""]]),
new Map(),
context.frames,
);

expect(
(
snapshot.combinedTree as string & { isWellFormed(): boolean }
).isWellFormed(),
).toBe(true);
expect(snapshot.combinedTree).toBe("Draw Again. \uFFFD");
});

it("overwrites duplicate iframe host entries when multiple children map to the same parent", () => {
const context: FrameContext = {
rootId: "frame-1",
Expand Down
Loading