Load the image-to-text model when opening the pdf viewer in Firefox (…

…bug 1908938)
mozilla · Jul 19, 2024 · c9fb206 · c9fb206
1 parent 5b0e15a
commit c9fb206
Show file tree

Hide file tree

Showing 5 changed files with 59 additions and 24 deletions.
diff --git a/src/display/editor/stamp.js b/src/display/editor/stamp.js
@@ -430,7 +430,8 @@ class StampEditor extends AnnotationEditor {
       return;
     }
     this.#hasMLBeenQueried = true;
-    if (!this._uiManager.isMLEnabledFor("altText") || this.hasAltText()) {
+    const isMLEnabled = await this._uiManager.isMLEnabledFor("altText");
+    if (!isMLEnabled || this.hasAltText()) {
       return;
     }
     const offscreen = new OffscreenCanvas(width, height);
@@ -447,7 +448,7 @@ class StampEditor extends AnnotationEditor {
       height
     );
     const response = await this._uiManager.mlGuess({
-      service: "image-to-text",
+      service: "moz-image-to-text",
       request: {
         data: ctx.getImageData(0, 0, width, height).data,
         width,

diff --git a/src/display/editor/tools.js b/src/display/editor/tools.js
@@ -855,8 +855,8 @@ class AnnotationEditorUIManager {
     return this.#mlManager?.guess(data) || null;
   }
 
-  isMLEnabledFor(name) {
-    return !!this.#mlManager?.isEnabledFor(name);
+  async isMLEnabledFor(name) {
+    return !!(await this.#mlManager?.isEnabledFor(name));
   }
 
   get useNewAltTextFlow() {

diff --git a/web/app.js b/web/app.js
@@ -155,6 +155,7 @@ const PDFViewerApplication = {
   isViewerEmbedded: window.parent !== window,
   url: "",
   baseUrl: "",
+  mlManager: null,
   _downloadUrl: "",
   _eventBusAbortController: null,
   _windowAbortController: null,
@@ -205,6 +206,11 @@ const PDFViewerApplication = {
       if (mode) {
         document.documentElement.classList.add(mode);
       }
+    } else {
+      // We want to load the image-to-text AI engine as soon as possible.
+      this.mlManager = new MLManager({
+        enableAltText: AppOptions.get("enableAltText"),
+      });
     }
 
     // Ensure that the `L10n`-instance has been initialized before creating
@@ -370,11 +376,14 @@ const PDFViewerApplication = {
 
     let eventBus;
     if (typeof PDFJSDev !== "undefined" && PDFJSDev.test("MOZCENTRAL")) {
-      eventBus = AppOptions.eventBus = new FirefoxEventBus(
-        AppOptions.get("allowedGlobalEvents"),
-        externalServices,
-        AppOptions.get("isInAutomation")
-      );
+      eventBus =
+        AppOptions.eventBus =
+        this.mlManager.eventBus =
+          new FirefoxEventBus(
+            AppOptions.get("allowedGlobalEvents"),
+            externalServices,
+            AppOptions.get("isInAutomation")
+          );
     } else {
       eventBus = new EventBus();
     }
@@ -731,15 +740,6 @@ const PDFViewerApplication = {
     return shadow(this, "externalServices", new ExternalServices());
   },
 
-  get mlManager() {
-    const enableAltText = AppOptions.get("enableAltText");
-    return shadow(
-      this,
-      "mlManager",
-      enableAltText === true ? new MLManager({ enableAltText }) : null
-    );
-  },
-
   get initialized() {
     return this._initializedCapability.settled;
   },

diff --git a/web/firefoxcom.js b/web/firefoxcom.js
@@ -308,19 +308,53 @@ class FirefoxScripting {
 }
 
 class MLManager {
-  #enabled = new Map();
+  #enabled = null;
 
-  constructor({ enableAltText }) {
-    this.#enabled.set("altText", enableAltText);
+  eventBus = null;
+
+  constructor(options) {
+    this.enable({ ...options, listenToProgress: false });
   }
 
-  isEnabledFor(name) {
-    return this.#enabled.get(name);
+  async isEnabledFor(name) {
+    return !!(await this.#enabled.get(name));
   }
 
   guess(data) {
     return FirefoxCom.requestAsync("mlGuess", data);
   }
+
+  enable({ enableAltText, listenToProgress }) {
+    if (enableAltText) {
+      this.#loadAltTextEngine(listenToProgress);
+    }
+  }
+
+  async #loadAltTextEngine(listenToProgress) {
+    if (this.#enabled.has("altText")) {
+      // We already have a promise for the "altText" service.
+      return;
+    }
+    if (listenToProgress) {
+      const callback = ({ detail }) => {
+        this.eventBus.dispatch("loadaiengineprogress", {
+          source: this,
+          detail,
+        });
+        if (detail.finished) {
+          window.removeEventListener("loadAIEngineProgress", callback);
+        }
+      };
+      window.addEventListener("loadAIEngineProgress", callback);
+    }
+    (this.#enabled ||= new Map()).set(
+      "altText",
+      FirefoxCom.requestAsync("loadAIEngine", {
+        service: "moz-image-to-text",
+        listenToProgress,
+      })
+    );
+  }
 }
 
 class ExternalServices extends BaseExternalServices {

diff --git a/web/genericcom.js b/web/genericcom.js
@@ -48,7 +48,7 @@ class ExternalServices extends BaseExternalServices {
 }
 
 class MLManager {
-  isEnabledFor(_name) {
+  async isEnabledFor(_name) {
     return false;
   }