fix: Link HTML attributes (BLO-915)

packages/core/src/extensions/tiptap-extensions/Link/link.ts

@@ -10,13 +10,6 @@ import { UNICODE_WHITESPACE_REGEX_GLOBAL } from "./helpers/whitespace.js";
 
 const DEFAULT_PROTOCOL = "https";
 
-const HTML_ATTRIBUTES = {
-  target: "_blank",
-  rel: "noopener noreferrer nofollow",
-  className: "bn-inline-content-section",
-  "data-inline-content-type": "link",
-};
-
 // Pre-compiled regex for URI protocol validation.
 // Allows: http, https, ftp, ftps, mailto, tel, callto, sms, cid, xmpp
 const ALLOWED_URI_REGEX =
@@ -84,7 +77,12 @@ export const Link = Mark.create<LinkOptions>({
 
   addOptions() {
     return {
-      HTMLAttributes: {},
+      HTMLAttributes: {
+        target: "_blank",
+        rel: "noopener noreferrer nofollow",
+        className: "bn-inline-content-section",
+        "data-inline-content-type": "link",
+      },
       editor: undefined,
       onClick: undefined,
       isValidLink: isAllowedUri,
@@ -99,12 +97,6 @@ export const Link = Mark.create<LinkOptions>({
           return element.getAttribute("href");
         },
       },
-      target: {
-        default: HTML_ATTRIBUTES.target,
-      },
-      rel: {
-        default: HTML_ATTRIBUTES.rel,
-      },
     };
   },
 
@@ -128,12 +120,22 @@ export const Link = Mark.create<LinkOptions>({
     if (!this.options.isValidLink(HTMLAttributes.href)) {
       return [
         "a",
-        mergeAttributes(HTML_ATTRIBUTES, { ...HTMLAttributes, href: "" }),
+        mergeAttributes(
+          {
+            ...HTMLAttributes,
+            href: "",
+          },
+          this.options.HTMLAttributes,
+        ),
         0,
       ];
     }
 
-    return ["a", mergeAttributes(HTML_ATTRIBUTES, HTMLAttributes), 0];
+    return [
+      "a",
+      mergeAttributes(HTMLAttributes, this.options.HTMLAttributes),
+      0,
+    ];
   },
 
   addPasteRules() {

packages/xl-ai/package.json

@@ -66,6 +66,7 @@
     "lint": "eslint src --max-warnings 0",
     "test": "NODE_EXTRA_CA_CERTS=\"$(mkcert -CAROOT)/rootCA.pem\" vitest --run",
     "test-watch": "NODE_EXTRA_CA_CERTS=\"$(mkcert -CAROOT)/rootCA.pem\" vitest watch",
+    "rename-msw-snapshots": "node scripts/rename-msw-snapshots.mjs",
     "email": "email dev"
   },
   "dependencies": {

packages/xl-ai/scripts/rename-msw-snapshots.mjs

@@ -0,0 +1,129 @@
+#!/usr/bin/env node
+// Repair msw-snapshot files after a request-shape change.
+//
+// When the request body changes (e.g. a schema change in BlockNote alters the
+// HTML/JSON sent to the LLM), the md5 hash that msw-snapshot embeds in each
+// cached response filename no longer matches. msw-snapshot then treats the
+// snapshot as missing and (because of `updateSnapshots: "missing"`) falls
+// through to the real API, which fails in CI without credentials and writes a
+// new file at the *correct* new hash containing the failure response (e.g.
+// 401).
+//
+// After that failed run, every affected slot has two files:
+//   <test>_<seq>_<old-hash>.json  -- valid 200 response, wrong hash
+//   <test>_<seq>_<new-hash>.json  -- right hash, but a 401 body
+//
+// This script transplants the 200 response into the new-hash file and deletes
+// the old-hash file, leaving exactly one file per slot with the right hash
+// and the right response.
+//
+// Usage:
+//   pnpm --filter @blocknote/xl-ai test     # populates the new-hash files
+//   pnpm --filter @blocknote/xl-ai rename-msw-snapshots
+//   pnpm --filter @blocknote/xl-ai test     # all green
+import {
+  readFileSync,
+  readdirSync,
+  statSync,
+  unlinkSync,
+  writeFileSync,
+} from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const PKG_ROOT = path.resolve(__dirname, "..");
+const SEARCH_ROOT = path.join(PKG_ROOT, "src");
+
+const FILE_RE = /^(.+)_(\d+)_([a-f0-9]+)\.json$/;
+
+function* walk(dir) {
+  for (const entry of readdirSync(dir, { withFileTypes: true })) {
+    const p = path.join(dir, entry.name);
+    if (entry.isDirectory()) {
+      yield* walk(p);
+    } else if (entry.isFile() && entry.name.endsWith(".json")) {
+      yield p;
+    }
+  }
+}
+
+const filesByDir = new Map();
+for (const file of walk(SEARCH_ROOT)) {
+  if (!file.includes(`${path.sep}__msw_snapshots__${path.sep}`)) continue;
+  const dir = path.dirname(file);
+  const list = filesByDir.get(dir) ?? [];
+  list.push(path.basename(file));
+  filesByDir.set(dir, list);
+}
+
+let migrated = 0;
+let skipped = 0;
+const skipNotes = [];
+
+for (const [dir, files] of filesByDir) {
+  const groups = new Map();
+  for (const file of files) {
+    const match = FILE_RE.exec(file);
+    if (!match) continue;
+    const slot = `${match[1]}_${match[2]}`;
+    const list = groups.get(slot) ?? [];
+    list.push(file);
+    groups.set(slot, list);
+  }
+
+  for (const [slot, group] of groups) {
+    if (group.length < 2) continue;
+
+    const entries = group.map((file) => {
+      const fp = path.join(dir, file);
+      const data = JSON.parse(readFileSync(fp, "utf8"));
+      return {
+        file,
+        path: fp,
+        data,
+        status: data?.response?.status,
+        mtime: statSync(fp).mtimeMs,
+      };
+    });
+
+    const good = entries.filter((e) => e.status === 200);
+    const bad = entries.filter((e) => e.status !== 200);
+
+    if (good.length !== 1 || bad.length === 0) {
+      skipped++;
+      skipNotes.push(
+        `  ${path.relative(PKG_ROOT, dir)}/${slot}: ${good.length} good + ${bad.length} bad`,
+      );
+      continue;
+    }
+
+    // Use the most recently written bad file as the destination — its hash
+    // matches the current request body.
+    bad.sort((a, b) => b.mtime - a.mtime);
+    const target = bad[0];
+
+    target.data.response = good[0].data.response;
+    writeFileSync(target.path, JSON.stringify(target.data, null, 2));
+    unlinkSync(good[0].path);
+    for (const extra of bad.slice(1)) unlinkSync(extra.path);
+
+    migrated++;
+    console.log(
+      `migrated ${path.relative(PKG_ROOT, dir)}/${slot} -> ${target.file}`,
+    );
+  }
+}
+
+console.log(`\nDone. ${migrated} migrated, ${skipped} skipped.`);
+if (skipped > 0) {
+  console.log("\nSkipped slots (need manual attention):");
+  for (const note of skipNotes) console.log(note);
+}
+if (migrated === 0 && skipped === 0) {
+  console.log(
+    "\nNo mismatched snapshot pairs found. If you expected some, run\n" +
+      "`pnpm --filter @blocknote/xl-ai test` first to let msw-snapshot record\n" +
+      "the new-hash files alongside the existing old-hash ones.",
+  );
+}

packages/xl-ai/src/api/formats/html-blocks/htmlBlocks.test.ts

@@ -1,5 +1,4 @@
-import { getCurrentTest } from "@vitest/runner";
-import { getSortedEntries, snapshot, toHashString } from "msw-snapshot";
+import { snapshot } from "msw-snapshot";
 import { setupServer } from "msw/node";
 import path from "path";
 import { afterAll, afterEach, beforeAll, describe, it } from "vitest";
@@ -8,6 +7,7 @@ import { testAIModels } from "../../../testUtil/testAIModels.js";
 import { BlockNoteEditor } from "@blocknote/core";
 import { StreamToolExecutor } from "../../../streamTool/StreamToolExecutor.js";
 import { ClientSideTransport } from "../../../streamTool/vercelAiSdk/clientside/ClientSideTransport.js";
+import { createSnapshotPathFn } from "../tests/snapshotPath.js";
 import { generateSharedTestCases } from "../tests/sharedTestCases.js";
 import { htmlBlockLLMFormat } from "./htmlBlocks.js";
 
@@ -17,54 +17,14 @@ const BASE_FILE_PATH = path.resolve(
   path.basename(__filename),
 );
 
-const fetchCountMap: Record<string, number> = {};
-
-async function createRequestHash(req: Request) {
-  const url = new URL(req.url);
-  return [
-    // url.host,
-    // url.pathname,
-    toHashString([
-      req.method,
-      url.origin,
-      url.pathname,
-      getSortedEntries(url.searchParams),
-      getSortedEntries(req.headers),
-      // getSortedEntries(req.cookies),
-      new TextDecoder("utf-8").decode(await req.arrayBuffer()),
-    ]),
-  ].join("/");
-}
-
 // Main test suite with snapshot middleware
 describe("Models", () => {
   // Define server with snapshot middleware for the main tests
   const server = setupServer(
     snapshot({
       updateSnapshots: "missing",
-      // onSnapshotUpdated: "all",
-      // ignoreSnapshots: true,
-      async createSnapshotPath(info) {
-        // use a unique path for each model
-        const t = getCurrentTest()!;
-        const mswPath = path.join(
-          t.suite!.name, // same directory as the test snapshot
-          "__msw_snapshots__",
-          t.suite!.suite!.name, // model / streaming params
-          t.name,
-        );
-        // in case there are multiple requests in a test, we need to use a separate snapshot for each request
-        fetchCountMap[mswPath] = (fetchCountMap[mswPath] || 0) + 1;
-        const hash = await createRequestHash(info.request);
-        return mswPath + `_${fetchCountMap[mswPath]}_${hash}.json`;
-      },
+      createSnapshotPath: createSnapshotPathFn(BASE_FILE_PATH),
       basePath: BASE_FILE_PATH,
-      // onFetchFromSnapshot(info, snapshot) {
-      // console.log("onFetchFromSnapshot", info, snapshot);
-      // },
-      // onFetchFromServer(info, snapshot) {
-      //   console.log("onFetchFromServer", info, snapshot);
-      // },
     }),
   );
 

packages/xl-ai/src/api/formats/json/json.test.ts

@@ -1,9 +1,9 @@
 import { afterAll, afterEach, beforeAll, describe } from "vitest";
 
-import { getCurrentTest } from "@vitest/runner";
-import { getSortedEntries, snapshot, toHashString } from "msw-snapshot";
+import { snapshot } from "msw-snapshot";
 import { setupServer } from "msw/node";
 import path from "path";
+import { createSnapshotPathFn } from "../tests/snapshotPath.js";
 import { generateSharedTestCases } from "../tests/sharedTestCases.js";
 
 import { ClientSideTransport } from "../../../streamTool/vercelAiSdk/clientside/ClientSideTransport.js";
@@ -16,54 +16,14 @@ const BASE_FILE_PATH = path.resolve(
   path.basename(__filename),
 );
 
-const fetchCountMap: Record<string, number> = {};
-
-async function createRequestHash(req: Request) {
-  const url = new URL(req.url);
-  return [
-    // url.host,
-    // url.pathname,
-    toHashString([
-      req.method,
-      url.origin,
-      url.pathname,
-      getSortedEntries(url.searchParams),
-      getSortedEntries(req.headers),
-      // getSortedEntries(req.cookies),
-      new TextDecoder("utf-8").decode(await req.arrayBuffer()),
-    ]),
-  ].join("/");
-}
-
 // Main test suite with snapshot middleware
 describe.skip("Models", () => {
   // Define server with snapshot middleware for the main tests
   const server = setupServer(
     snapshot({
       updateSnapshots: "missing",
-      // updateSnapshots: "all",
-      // ignoreSnapshots: true,
-      async createSnapshotPath(info) {
-        // use a unique path for each model
-        const t = getCurrentTest()!;
-        const mswPath = path.join(
-          t.suite!.name, // same directory as the test snapshot
-          "__msw_snapshots__",
-          t.suite!.suite!.name, // model / streaming params
-          t.name,
-        );
-        // in case there are multiple requests in a test, we need to use a separate snapshot for each request
-        fetchCountMap[mswPath] = (fetchCountMap[mswPath] || 0) + 1;
-        const hash = await createRequestHash(info.request);
-        return mswPath + `_${fetchCountMap[mswPath]}_${hash}.json`;
-      },
+      createSnapshotPath: createSnapshotPathFn(BASE_FILE_PATH),
       basePath: BASE_FILE_PATH,
-      // onFetchFromSnapshot(info, snapshot) {
-      // console.log("onFetchFromSnapshot", info, snapshot);
-      // },
-      // onFetchFromServer(info, snapshot) {
-      //   console.log("onFetchFromServer", info, snapshot);
-      // },
     }),
   );