blob: bf5e76f5fa7b1a98211af70745136728a716e0d1 [file] [log] [blame]
Avi Drissman8ba1bad2022-09-13 19:22:361// Copyright 2022 The Chromium Authors
Ramin Halavati26dcce22022-02-23 13:11:142// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
Abigail Klein7a63c572024-02-28 20:45:095#include "services/screen_ai/screen_ai_service_impl.h"
Ramin Halavati26dcce22022-02-23 13:11:146
Ramin Halavati26679592025-03-13 07:24:577#include <algorithm>
Ramin Halavati55fdf4d2022-11-07 06:35:188#include <memory>
Ramin Halavati63d9c252024-07-29 06:59:179#include <optional>
Ramin Halavatifd712d8e2022-10-19 05:11:2710#include <utility>
Ramin Halavatib044504d2022-10-24 06:36:3411#include <vector>
Ramin Halavatifd712d8e2022-10-19 05:11:2712
13#include "base/check.h"
Ramin Halavatiad969ef2024-05-21 06:59:4714#include "base/check_is_test.h"
Tom Sepeze9b4dd552024-08-14 22:44:2915#include "base/compiler_specific.h"
Ramin Halavati65f96fb2025-02-24 16:45:1816#include "base/cpu.h"
Ramin Halavati55fdf4d2022-11-07 06:35:1817#include "base/functional/bind.h"
18#include "base/location.h"
Ramin Halavatifafff152022-12-05 09:39:3819#include "base/logging.h"
Kyungjun Lee9d755a02022-11-08 17:57:1120#include "base/metrics/histogram_functions.h"
Ramin Halavatida3ec442025-05-06 04:14:0321#include "base/numerics/safe_conversions.h"
Ramin Halavati5ff855c2022-04-27 16:30:4822#include "base/process/process.h"
Ramin Halavati65f96fb2025-02-24 16:45:1823#include "base/strings/stringprintf.h"
Ramin Halavati26679592025-03-13 07:24:5724#include "base/system/sys_info.h"
Sean Maher5b9af51f2022-11-21 15:32:4725#include "base/task/single_thread_task_runner.h"
Ramin Halavati55fdf4d2022-11-07 06:35:1826#include "base/task/thread_pool.h"
Ramin Halavatiaa67ee52025-04-25 05:15:5327#include "base/threading/sequence_bound.h"
Ramin Halavati65f96fb2025-02-24 16:45:1828#include "components/crash/core/common/crash_key.h"
Abigail Klein7a63c572024-02-28 20:45:0929#include "services/screen_ai/buildflags/buildflags.h"
Ramin Halavati4c47eb82024-12-03 19:06:4230#include "services/screen_ai/proto/chrome_screen_ai.pb.h"
Abigail Klein7a63c572024-02-28 20:45:0931#include "services/screen_ai/proto/main_content_extractor_proto_convertor.h"
32#include "services/screen_ai/proto/visual_annotator_proto_convertor.h"
Ramin Halavati4c47eb82024-12-03 19:06:4233#include "services/screen_ai/public/cpp/metrics.h"
Abigail Klein7a63c572024-02-28 20:45:0934#include "services/screen_ai/public/cpp/utilities.h"
Ramin Halavatieddadb62022-05-04 17:29:4935#include "ui/accessibility/accessibility_features.h"
Ramin Halavati4f904d0a2024-06-25 15:32:3236#include "ui/accessibility/ax_node.h"
37#include "ui/accessibility/ax_tree.h"
Ramin Halavati84231072022-08-17 08:02:2038#include "ui/accessibility/ax_tree_id.h"
Nektarios Paisiosf73d6972022-06-04 11:32:2539#include "ui/gfx/geometry/rect_f.h"
Ramin Halavati367352a2022-04-14 06:00:2340
Ramin Halavatia9b50a102024-02-07 18:37:2341#if BUILDFLAG(USE_FAKE_SCREEN_AI)
Abigail Klein7a63c572024-02-28 20:45:0942#include "services/screen_ai/screen_ai_library_wrapper_fake.h"
Ramin Halavati767b8fc2024-02-02 06:23:0843#else
Abigail Klein7a63c572024-02-28 20:45:0944#include "services/screen_ai/screen_ai_library_wrapper_impl.h"
Ramin Halavati767b8fc2024-02-02 06:23:0845#endif
46
Ramin Halavati69bf7752022-04-04 09:58:1547namespace screen_ai {
Ramin Halavati26dcce22022-02-23 13:11:1448
Ramin Halavati6d4fc2b2022-06-16 15:33:3549namespace {
50
Ramin Halavati2acf2362024-08-02 14:22:5351// How often it would be checked that the service is idle and can be shutdown.
Ramin Halavatibe45dde2025-05-01 04:19:1052// LINT.IfChange(kIdleCheckingDelay)
Ramin Halavati1520b702025-04-23 15:24:1853constexpr base::TimeDelta kIdleCheckingDelay = base::Seconds(3);
Ramin Halavatibe45dde2025-05-01 04:19:1054// LINT.ThenChange(//chrome/browser/screen_ai/optical_character_recognizer_browsertest.cc:kServiceIdleCheckingDelay)
Ramin Halavati6ff56412024-08-13 07:15:5455
Ramin Halavati4825419e2025-04-02 05:49:4656// How long to wait for a request to the library be responded, before assuming
57// that the library is not responsive.
Ramin Halavatiaa67ee52025-04-25 05:15:5358constexpr base::TimeDelta kMaxWaitForResponseTime = base::Seconds(10);
Ramin Halavati4825419e2025-04-02 05:49:4659
Ramin Halavatiad969ef2024-05-21 06:59:4760// These values are persisted to logs. Entries should not be renumbered and
61// numeric values should never be reused.
Ramin Halavati29cc1152024-10-27 17:30:2862// See `screen_ai_service.mojom` for more info.
Ramin Halavatib3fe2fe2025-03-20 16:53:0363// LINT.IfChange(OcrClientType)
Ramin Halavatiad969ef2024-05-21 06:59:4764enum class OcrClientTypeForMetrics {
65 kTest = 0,
66 kPdfViewer = 1,
67 kLocalSearch = 2,
68 kCameraApp = 3,
Ramin Halavati29cc1152024-10-27 17:30:2869 kNotUsed = 4, // Can be used for a new client.
Ramin Halavatiad969ef2024-05-21 06:59:4770 kMediaApp = 5,
Michelle Chen5192d522024-10-22 23:38:2271 kScreenshotTextDetection,
72 kMaxValue = kScreenshotTextDetection
Ramin Halavatiad969ef2024-05-21 06:59:4773};
Ramin Halavatib3fe2fe2025-03-20 16:53:0374// LINT.ThenChange(//tools/metrics/histograms/metadata/accessibility/enums.xml:OcrClientType)
75
76// These values are persisted to logs. Entries should not be renumbered and
77// numeric values should never be reused.
78// See `screen_ai_service.mojom` for more info.
79// LINT.IfChange(MainContentExtractionClientType)
80enum class MainContentExtractionClientTypeForMetrics {
81 kTest = 0,
82 kReadingMode = 1,
83 kMainNode = 2,
84 kMahi = 3,
85 kMaxValue = kMahi
86};
87// LINT.ThenChange(//tools/metrics/histograms/metadata/accessibility/enums.xml:MainContentExtractionClientType)
Ramin Halavatiad969ef2024-05-21 06:59:4788
89OcrClientTypeForMetrics GetClientType(mojom::OcrClientType client_type) {
90 switch (client_type) {
91 case mojom::OcrClientType::kTest:
92 CHECK_IS_TEST();
93 return OcrClientTypeForMetrics::kTest;
94 case mojom::OcrClientType::kPdfViewer:
95 return OcrClientTypeForMetrics::kPdfViewer;
96 case mojom::OcrClientType::kLocalSearch:
97 return OcrClientTypeForMetrics::kLocalSearch;
98 case mojom::OcrClientType::kCameraApp:
99 return OcrClientTypeForMetrics::kCameraApp;
Ramin Halavatiad969ef2024-05-21 06:59:47100 case mojom::OcrClientType::kMediaApp:
101 return OcrClientTypeForMetrics::kMediaApp;
Michelle Chen5192d522024-10-22 23:38:22102 case mojom::OcrClientType::kScreenshotTextDetection:
103 return OcrClientTypeForMetrics::kScreenshotTextDetection;
Ramin Halavatiad969ef2024-05-21 06:59:47104 }
105}
106
Ramin Halavatib3fe2fe2025-03-20 16:53:03107MainContentExtractionClientTypeForMetrics GetClientType(
108 mojom::MceClientType client_type) {
109 switch (client_type) {
110 case mojom::MceClientType::kTest:
111 CHECK_IS_TEST();
112 return MainContentExtractionClientTypeForMetrics::kTest;
113 case mojom::MceClientType::kReadingMode:
114 return MainContentExtractionClientTypeForMetrics::kReadingMode;
115 case mojom::MceClientType::kMainNode:
116 return MainContentExtractionClientTypeForMetrics::kMainNode;
117 case mojom::MceClientType::kMahi:
118 return MainContentExtractionClientTypeForMetrics::kMahi;
119 }
120}
121
Ramin Halavati87cfa8e62025-06-02 06:48:10122#if BUILDFLAG(IS_CHROMEOS)
Ramin Halavati89a36c92023-05-05 15:01:03123ui::AXTreeUpdate ConvertVisualAnnotationToTreeUpdate(
Ramin Halavatidcb09a172024-07-15 08:28:55124 std::optional<chrome_screen_ai::VisualAnnotation>& annotation_proto,
Ramin Halavati89a36c92023-05-05 15:01:03125 const gfx::Rect& image_rect) {
126 if (!annotation_proto) {
127 VLOG(0) << "Screen AI library could not process snapshot or no OCR data.";
128 return ui::AXTreeUpdate();
129 }
130
131 return VisualAnnotationToAXTreeUpdate(*annotation_proto, image_rect);
132}
Ramin Halavati87cfa8e62025-06-02 06:48:10133#endif // BUILDFLAG(IS_CHROMEOS)
Ramin Halavati89a36c92023-05-05 15:01:03134
Abigail Klein18673702024-03-05 20:59:06135ui::AXNodeID ComputeMainNode(
136 const ui::AXTree* tree,
137 const std::vector<ui::AXNodeID>& content_node_ids) {
138 ui::AXNode* front = tree->GetFromId(content_node_ids.front());
139 ui::AXNode* back = tree->GetFromId(content_node_ids.back());
140 ui::AXNode* main = front->GetLowestCommonAncestor(*back);
141 return main->id();
142}
143
Ramin Halavati65f96fb2025-02-24 16:45:18144#if !BUILDFLAG(USE_FAKE_SCREEN_AI)
145void SetCPUInstructionSetCrashKey() {
146#if defined(ARCH_CPU_X86_FAMILY)
147 base::CPU();
148 // Report cpu micro architecture in case of crash.
149 static crash_reporter::CrashKeyString<3> cpu_info("intel_micro_architecture");
150 cpu_info.Set(
151 base::StringPrintf("%i", base::CPU().GetIntelMicroArchitecture()));
152#endif
153}
154#endif
155
Ramin Halavati26679592025-03-13 07:24:57156// Return a maximum 11 character string with the signature of available and
157// total memory, both in MB and capped to 99999.
158std::string GetMemoryStatusForCrashKey() {
159 int total_memory = base::SysInfo::AmountOfPhysicalMemoryMB();
160 int available_memory = static_cast<int>(
161 base::SysInfo::AmountOfAvailablePhysicalMemory() / (1024 * 1024));
162
163 // Cap the number of digits for crash report.
164 total_memory = std::min(total_memory, 99999);
165 available_memory = std::min(available_memory, 99999);
166 return base::StringPrintf("%i,%i", available_memory, total_memory);
167}
168
Ramin Halavatiaa67ee52025-04-25 05:15:53169class HangTimer : public base::OneShotTimer {
170 public:
171 explicit HangTimer(bool is_ocr) : is_ocr_(is_ocr) {}
172
173 void StartTimer() {
174 Start(FROM_HERE, kMaxWaitForResponseTime,
175 base::BindOnce(
176 [](bool request_is_ocr) {
177 base::UmaHistogramBoolean(
Ramin Halavatib5db88472025-05-23 14:34:07178 "Accessibility.ScreenAI.Service.NotResponsive.IsOCR",
Ramin Halavatiaa67ee52025-04-25 05:15:53179 request_is_ocr);
180 base::Process::TerminateCurrentProcessImmediately(0);
181 },
182 is_ocr_));
183 }
184
185 private:
186 bool is_ocr_;
187};
188
Ramin Halavati26ae6b72022-11-10 06:45:05189} // namespace
190
Ramin Halavati39bab0d2024-01-30 06:05:01191// The library accepts simple pointers to model data retrieval functions, hence
192// callback functions with linked object are not safe to pass.
Ramin Halavati63d9c252024-07-29 06:59:17193// This global variable keeps the pointer the only instance of this class.
194ModelDataHolder* g_model_data_holder_instance = nullptr;
Ramin Halavati8ced6392023-09-25 07:27:50195
Ramin Halavati63d9c252024-07-29 06:59:17196// Keeps the handles of model files, and replies to calls for copying their
197// content.
198class ModelDataHolder {
Ramin Halavati8ced6392023-09-25 07:27:50199 public:
Ramin Halavati63d9c252024-07-29 06:59:17200 ModelDataHolder() {
201 CHECK(!g_model_data_holder_instance);
202 g_model_data_holder_instance = this;
203 }
Ramin Halavati8ced6392023-09-25 07:27:50204
Ramin Halavati63d9c252024-07-29 06:59:17205 ModelDataHolder(const ModelDataHolder&) = delete;
206 ModelDataHolder& operator=(const ModelDataHolder&) = delete;
207
208 ~ModelDataHolder() {
209 CHECK_EQ(g_model_data_holder_instance, this);
210 g_model_data_holder_instance = nullptr;
Ramin Halavati8ced6392023-09-25 07:27:50211 }
212
213 // Returns 0 if file is not found.
214 static uint32_t GetDataSize(const char* relative_file_path) {
Ramin Halavati63d9c252024-07-29 06:59:17215 CHECK(g_model_data_holder_instance);
216 base::File* model_file =
217 g_model_data_holder_instance->GetModelFile(relative_file_path);
218 return model_file ? model_file->GetLength() : 0;
Ramin Halavati8ced6392023-09-25 07:27:50219 }
220
Ramin Halavati63d9c252024-07-29 06:59:17221 // Copies content of the file in `relative_file_path` to `buffer`. Expects
222 // that `buffer_size` would be enough for the entire file content.
Ramin Halavati8ced6392023-09-25 07:27:50223 static void CopyData(const char* relative_file_path,
224 uint32_t buffer_size,
225 char* buffer) {
Ramin Halavati63d9c252024-07-29 06:59:17226 CHECK(g_model_data_holder_instance);
227 base::File* model_file =
228 g_model_data_holder_instance->GetModelFile(relative_file_path);
229 CHECK(model_file);
230
231 int64_t length = model_file->GetLength();
232 CHECK_GE(buffer_size, length);
Tom Sepeze9b4dd552024-08-14 22:44:29233 CHECK_EQ(UNSAFE_TODO(model_file->Read(0, buffer, length)), length);
Ramin Halavati8ced6392023-09-25 07:27:50234 }
235
Ramin Halavati63d9c252024-07-29 06:59:17236 void AddModelFiles(base::flat_map<base::FilePath, base::File> model_files) {
237 for (auto& model_file : model_files) {
238 model_files_[model_file.first.MaybeAsASCII()] =
239 std::move(model_file.second);
Ramin Halavati39bab0d2024-01-30 06:05:01240 }
241 }
Ramin Halavati8ced6392023-09-25 07:27:50242
Ramin Halavati63d9c252024-07-29 06:59:17243 // Returns the file handle for `relative_file_path` if it exists.
244 base::File* GetModelFile(const char* relative_file_path) {
245 if (!base::Contains(model_files_, relative_file_path)) {
246 return nullptr;
247 }
248 return &model_files_[relative_file_path];
249 }
250
Ramin Halavati8ced6392023-09-25 07:27:50251 private:
Ramin Halavati63d9c252024-07-29 06:59:17252 std::map<std::string, base::File> model_files_;
Ramin Halavati8ced6392023-09-25 07:27:50253};
254
Ramin Halavati26ae6b72022-11-10 06:45:05255ScreenAIService::ScreenAIService(
Ramin Halavatic1e4fa92023-05-17 17:22:06256 mojo::PendingReceiver<mojom::ScreenAIServiceFactory> receiver)
257 : factory_receiver_(this, std::move(receiver)),
258 ocr_receiver_(this),
Ramin Halavatiad969ef2024-05-21 06:59:47259 main_content_extraction_receiver_(this) {
Ramin Halavati6ff56412024-08-13 07:15:54260 screen2x_main_content_extractors_.set_disconnect_handler(
Ramin Halavati1520b702025-04-23 15:24:18261 base::BindRepeating(&ScreenAIService::MceReceiverDisconnected,
Ramin Halavati6ff56412024-08-13 07:15:54262 weak_ptr_factory_.GetWeakPtr()));
263 screen_ai_annotators_.set_disconnect_handler(
264 base::BindRepeating(&ScreenAIService::OcrReceiverDisconnected,
265 weak_ptr_factory_.GetWeakPtr()));
Ramin Halavati63d9c252024-07-29 06:59:17266 model_data_holder_ = std::make_unique<ModelDataHolder>();
Ramin Halavatiaa67ee52025-04-25 05:15:53267
268 background_task_runner_ = base::ThreadPool::CreateSequencedTaskRunner(
269 {base::TaskPriority::BEST_EFFORT,
270 base::TaskShutdownBehavior::SKIP_ON_SHUTDOWN});
Ramin Halavatiad969ef2024-05-21 06:59:47271}
Ramin Halavati26ae6b72022-11-10 06:45:05272
273ScreenAIService::~ScreenAIService() = default;
274
Ramin Halavati8ced6392023-09-25 07:27:50275void ScreenAIService::LoadLibrary(const base::FilePath& library_path) {
Abigail Klein7a63c572024-02-28 20:45:09276 // The ScopedBlockingCall in LoadLibrary guarantees that this is not run on
277 // the UI thread.
Ramin Halavatia9b50a102024-02-07 18:37:23278#if BUILDFLAG(USE_FAKE_SCREEN_AI)
Ramin Halavati767b8fc2024-02-02 06:23:08279 library_ = std::make_unique<ScreenAILibraryWrapperFake>();
280#else
281 library_ = std::make_unique<ScreenAILibraryWrapperImpl>();
Ramin Halavati65f96fb2025-02-24 16:45:18282
283 // TODO(crbug.com/381256355): Remove when the library is SSE3 compatible.
284 SetCPUInstructionSetCrashKey();
Ramin Halavati767b8fc2024-02-02 06:23:08285#endif
Ramin Halavati8ced6392023-09-25 07:27:50286
287 bool load_sucessful = library_->Load(library_path);
288 base::UmaHistogramBoolean("Accessibility.ScreenAI.Library.Initialized",
289 load_sucessful);
290
291 if (!load_sucessful) {
292 library_.reset();
293 return;
294 }
295
296 uint32_t version_major;
297 uint32_t version_minor;
298 library_->GetLibraryVersion(version_major, version_minor);
299 VLOG(2) << "Screen AI library version: " << version_major << "."
300 << version_minor;
301
Georg Neis06e387772024-12-25 07:32:22302#if BUILDFLAG(IS_CHROMEOS)
Ramin Halavati8ced6392023-09-25 07:27:50303 library_->SetLogger();
304#endif
305
306 if (features::IsScreenAIDebugModeEnabled()) {
307 library_->EnableDebugMode();
308 }
309
Ramin Halavati63d9c252024-07-29 06:59:17310 library_->SetFileContentFunctions(&ModelDataHolder::GetDataSize,
311 &ModelDataHolder::CopyData);
Ramin Halavati8ced6392023-09-25 07:27:50312}
313
Ramin Halavatic1e4fa92023-05-17 17:22:06314void ScreenAIService::InitializeMainContentExtraction(
Ramin Halavatid139a0f2023-02-28 13:25:28315 const base::FilePath& library_path,
Ramin Halavatifbc62022023-11-10 11:05:07316 base::flat_map<base::FilePath, base::File> model_files,
Ramin Halavatic1e4fa92023-05-17 17:22:06317 mojo::PendingReceiver<mojom::MainContentExtractionService>
318 main_content_extractor_service_receiver,
319 InitializeMainContentExtractionCallback callback) {
320 if (!library_) {
Ramin Halavati8ced6392023-09-25 07:27:50321 LoadLibrary(library_path);
Ramin Halavatic1e4fa92023-05-17 17:22:06322 }
323
324 if (!library_) {
325 std::move(callback).Run(false);
326 base::Process::TerminateCurrentProcessImmediately(-1);
327 }
328
Ramin Halavati63d9c252024-07-29 06:59:17329 model_data_holder_->AddModelFiles(std::move(model_files));
Ramin Halavatib044504d2022-10-24 06:36:34330
Ramin Halavati8ced6392023-09-25 07:27:50331 bool init_successful = library_->InitMainContentExtraction();
Ramin Halavati1d57c2d2023-05-24 05:18:24332 base::UmaHistogramBoolean(
333 "Accessibility.ScreenAI.MainContentExtraction.Initialized",
334 init_successful);
335 if (!init_successful) {
Ramin Halavatic1e4fa92023-05-17 17:22:06336 std::move(callback).Run(false);
Ramin Halavati89a36c92023-05-05 15:01:03337 return;
Ramin Halavatid139a0f2023-02-28 13:25:28338 }
Ramin Halavati89a36c92023-05-05 15:01:03339
Ramin Halavatic1e4fa92023-05-17 17:22:06340 // This interface should be created only once.
341 CHECK(!main_content_extraction_receiver_.is_bound());
342
343 main_content_extraction_receiver_.Bind(
344 std::move(main_content_extractor_service_receiver));
345
346 std::move(callback).Run(true);
Ramin Halavati1520b702025-04-23 15:24:18347 mce_last_used_ = base::TimeTicks::Now();
Ramin Halavaticf155012025-05-05 15:50:01348 StartShutDownOnIdleTimer();
Ramin Halavatic1e4fa92023-05-17 17:22:06349}
350
351void ScreenAIService::InitializeOCR(
352 const base::FilePath& library_path,
Ramin Halavatifbc62022023-11-10 11:05:07353 base::flat_map<base::FilePath, base::File> model_files,
Ramin Halavatic1e4fa92023-05-17 17:22:06354 mojo::PendingReceiver<mojom::OCRService> ocr_service_receiver,
355 InitializeOCRCallback callback) {
Ramin Halavati26679592025-03-13 07:24:57356 static crash_reporter::CrashKeyString<12> memory_ocr_init(
357 "screen_ai_mem_ocr_init");
358 memory_ocr_init.Set(GetMemoryStatusForCrashKey());
Ramin Halavatic1e4fa92023-05-17 17:22:06359 if (!library_) {
Ramin Halavati8ced6392023-09-25 07:27:50360 LoadLibrary(library_path);
Ramin Halavatic1e4fa92023-05-17 17:22:06361 }
362
363 if (!library_) {
364 std::move(callback).Run(false);
365 base::Process::TerminateCurrentProcessImmediately(-1);
366 }
367
Ramin Halavati63d9c252024-07-29 06:59:17368 model_data_holder_->AddModelFiles(std::move(model_files));
369
Ramin Halavati0e2ae5742023-11-10 05:57:54370 bool init_successful = library_->InitOCR();
Ramin Halavati35d129f2023-06-22 16:37:09371 base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Initialized",
Ramin Halavati1d57c2d2023-05-24 05:18:24372 init_successful);
Ramin Halavati480584332023-08-02 06:38:40373
Ramin Halavati1d57c2d2023-05-24 05:18:24374 if (!init_successful) {
Ramin Halavatic1e4fa92023-05-17 17:22:06375 std::move(callback).Run(false);
376 return;
377 }
378
Ramin Halavatida3ec442025-05-06 04:14:03379 max_ocr_dimension_ = library_->GetMaxImageDimension();
380 CHECK(max_ocr_dimension_);
381
Ramin Halavatic1e4fa92023-05-17 17:22:06382 // This interface should be created only once.
383 CHECK(!ocr_receiver_.is_bound());
384
385 ocr_receiver_.Bind(std::move(ocr_service_receiver));
386
387 std::move(callback).Run(true);
Ramin Halavati2acf2362024-08-02 14:22:53388 ocr_last_used_ = base::TimeTicks::Now();
Ramin Halavaticf155012025-05-05 15:50:01389 StartShutDownOnIdleTimer();
Ramin Halavatib044504d2022-10-24 06:36:34390}
Ramin Halavati26dcce22022-02-23 13:11:14391
Ramin Halavatif26d7d42025-03-12 16:04:46392void ScreenAIService::BindShutdownHandler(
393 mojo::PendingRemote<mojom::ScreenAIServiceShutdownHandler>
394 shutdown_handler) {
395 DCHECK(!screen_ai_shutdown_handler_.is_bound());
396 screen_ai_shutdown_handler_.Bind(std::move(shutdown_handler));
397}
398
Ramin Halavati26dcce22022-02-23 13:11:14399void ScreenAIService::BindAnnotator(
400 mojo::PendingReceiver<mojom::ScreenAIAnnotator> annotator) {
401 screen_ai_annotators_.Add(this, std::move(annotator));
402}
403
Ramin Halavatieddadb62022-05-04 17:29:49404void ScreenAIService::BindMainContentExtractor(
405 mojo::PendingReceiver<mojom::Screen2xMainContentExtractor>
406 main_content_extractor) {
Ramin Halavati6ff56412024-08-13 07:15:54407 screen2x_main_content_extractors_.Add(this,
408 std::move(main_content_extractor));
Ramin Halavatieddadb62022-05-04 17:29:49409}
410
Arthur Sonzognic571efb2024-01-26 20:26:18411std::optional<chrome_screen_ai::VisualAnnotation>
Ramin Halavati7716b032024-08-19 14:56:20412ScreenAIService::PerformOcrAndRecordMetrics(const SkBitmap& image) {
Ramin Halavati26679592025-03-13 07:24:57413 static crash_reporter::CrashKeyString<12> memory_perform_ocr(
414 "screen_ai_mem_ocr_perform");
415 memory_perform_ocr.Set(GetMemoryStatusForCrashKey());
416
Ramin Halavati7716b032024-08-19 14:56:20417 CHECK(base::Contains(ocr_client_types_,
418 screen_ai_annotators_.current_receiver()));
Ramin Halavati4805c882025-03-03 18:28:48419 OcrClientTypeForMetrics client_type = GetClientType(
420 ocr_client_types_.find(screen_ai_annotators_.current_receiver())->second);
Ramin Halavatiad969ef2024-05-21 06:59:47421 base::UmaHistogramEnumeration("Accessibility.ScreenAI.OCR.ClientType",
Ramin Halavati4805c882025-03-03 18:28:48422 client_type);
Ramin Halavatiad969ef2024-05-21 06:59:47423
Ramin Halavatiac1ee2922025-05-21 17:17:17424 base::TimeTicks start_time = base::TimeTicks::Now();
Ramin Halavatiaa67ee52025-04-25 05:15:53425 base::SequenceBound<HangTimer> hang_timer(background_task_runner_,
426 /*is_ocr=*/true);
427 hang_timer.AsyncCall(&HangTimer::StartTimer);
Ramin Halavati7252cd02023-05-10 07:21:33428 auto result = library_->PerformOcr(image);
Ramin Halavatiaa67ee52025-04-25 05:15:53429 hang_timer.AsyncCall(&base::OneShotTimer::Stop);
Ramin Halavatiac1ee2922025-05-21 17:17:17430 base::TimeDelta elapsed_time = base::TimeTicks::Now() - start_time;
Ramin Halavati4825419e2025-04-02 05:49:46431
Ramin Halavati2650ebe2023-11-27 19:56:35432 int lines_count = result ? result->lines_size() : 0;
433 VLOG(1) << "OCR returned " << lines_count << " lines in " << elapsed_time;
Ramin Halavati7252cd02023-05-10 07:21:33434
Ramin Halavati18aba872024-08-23 14:33:05435 if (!result) {
436 base::UmaHistogramEnumeration(
Ramin Halavati4805c882025-03-03 18:28:48437 "Accessibility.ScreenAI.OCR.Failed.ClientType", client_type);
Ramin Halavati18aba872024-08-23 14:33:05438 }
Ramin Halavati49c2b5c2025-04-25 05:43:01439
Ramin Halavatida3ec442025-05-06 04:14:03440 int max_dimension = base::checked_cast<int>(max_ocr_dimension_);
441 if (image.width() > max_dimension || image.height() > max_dimension) {
Ramin Halavati4805c882025-03-03 18:28:48442 base::UmaHistogramEnumeration(
Ramin Halavati49c2b5c2025-04-25 05:43:01443 "Accessibility.ScreenAI.OCR.Downsampled.ClientType", client_type);
Ramin Halavati4805c882025-03-03 18:28:48444 }
445
Ramin Halavati81b6d922024-06-07 14:20:34446 base::UmaHistogramBoolean("Accessibility.ScreenAI.OCR.Successful",
447 result.has_value());
Ramin Halavati2650ebe2023-11-27 19:56:35448 base::UmaHistogramCounts100("Accessibility.ScreenAI.OCR.LinesCount",
449 lines_count);
Ramin Halavatida3ec442025-05-06 04:14:03450 if (image.width() < max_dimension && image.height() < max_dimension) {
Ramin Halavati49c2b5c2025-04-25 05:43:01451 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.NotDownsampled",
Ramin Halavati3ae33ca2024-01-01 12:58:39452 elapsed_time);
453 } else {
Ramin Halavati49c2b5c2025-04-25 05:43:01454 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Latency.Downsampled",
Ramin Halavati3ae33ca2024-01-01 12:58:39455 elapsed_time);
456 }
Ramin Halavati2650ebe2023-11-27 19:56:35457
Ramin Halavati7716b032024-08-19 14:56:20458 // MediaApp provides OCR for ChromeOS PDF viewer.
Ramin Halavati4805c882025-03-03 18:28:48459 if (client_type == OcrClientTypeForMetrics::kPdfViewer ||
460 client_type == OcrClientTypeForMetrics::kMediaApp) {
Ramin Halavati2650ebe2023-11-27 19:56:35461 base::UmaHistogramCounts100("Accessibility.ScreenAI.OCR.LinesCount.PDF",
462 lines_count);
463 base::UmaHistogramTimes("Accessibility.ScreenAI.OCR.Time.PDF",
464 elapsed_time);
Ramin Halavati696b8d5c2025-02-28 16:41:58465 base::UmaHistogramCounts10M(
466 lines_count ? "Accessibility.ScreenAI.OCR.ImageSize.PDF.WithText"
467 : "Accessibility.ScreenAI.OCR.ImageSize.PDF.NoText",
Ramin Halavati49c2b5c2025-04-25 05:43:01468 image.width() * image.height());
Ramin Halavati4c47eb82024-12-03 19:06:42469
Ramin Halavati01267d12025-01-21 18:11:30470 if (result.has_value()) {
471 std::optional<uint64_t> most_detected_language =
472 GetMostDetectedLanguageInOcrData(*result);
473 if (most_detected_language.has_value()) {
474 base::UmaHistogramSparse(
475 "Accessibility.ScreenAI.OCR.MostDetectedLanguage.PDF",
476 most_detected_language.value());
477 }
Ramin Halavati4c47eb82024-12-03 19:06:42478 }
Ramin Halavati2650ebe2023-11-27 19:56:35479 }
480
Ramin Halavatiac1ee2922025-05-21 17:17:17481 ocr_last_used_ = base::TimeTicks::Now();
Ramin Halavati7252cd02023-05-10 07:21:33482 return result;
483}
484
Ramin Halavatiad969ef2024-05-21 06:59:47485void ScreenAIService::SetClientType(mojom::OcrClientType client_type) {
486 ocr_client_types_[screen_ai_annotators_.current_receiver()] = client_type;
487}
488
Ramin Halavatid0d6231b2025-03-20 05:36:43489void ScreenAIService::SetClientType(mojom::MceClientType client_type) {
490 mce_client_types_[screen2x_main_content_extractors_.current_receiver()] =
491 client_type;
492}
493
Ramin Halavati1520b702025-04-23 15:24:18494void ScreenAIService::OcrReceiverDisconnected() {
495 auto entry = ocr_client_types_.find(screen_ai_annotators_.current_receiver());
496 if (entry != ocr_client_types_.end()) {
497 ocr_client_types_.erase(entry);
498 }
499 // Modify last used time to ensure the service does not shutdown while a
500 // client is disconnecting.
501 ocr_last_used_ = base::TimeTicks::Now();
502}
503
504void ScreenAIService::MceReceiverDisconnected() {
505 auto entry = mce_client_types_.find(
506 screen2x_main_content_extractors_.current_receiver());
507 if (entry != mce_client_types_.end()) {
508 mce_client_types_.erase(entry);
509 }
510 // Modify last used time to ensure the service does not shutdown while a
511 // client is disconnecting.
512 mce_last_used_ = base::TimeTicks::Now();
513}
514
Ramin Halavati062042f2025-05-02 05:47:57515void ScreenAIService::GetMaxImageDimension(
516 GetMaxImageDimensionCallback callback) {
Ramin Halavatida3ec442025-05-06 04:14:03517 CHECK(max_ocr_dimension_);
518 std::move(callback).Run(max_ocr_dimension_);
Ramin Halavati062042f2025-05-02 05:47:57519}
520
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07521void ScreenAIService::PerformOcrAndReturnAnnotation(
522 const SkBitmap& image,
523 PerformOcrAndReturnAnnotationCallback callback) {
Arthur Sonzognic571efb2024-01-26 20:26:18524 std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto =
Ramin Halavati7716b032024-08-19 14:56:20525 PerformOcrAndRecordMetrics(image);
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07526
Ramin Halavati89a36c92023-05-05 15:01:03527 if (annotation_proto) {
528 std::move(callback).Run(ConvertProtoToVisualAnnotation(*annotation_proto));
529 return;
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07530 }
Ramin Halavati89a36c92023-05-05 15:01:03531
532 std::move(callback).Run(mojom::VisualAnnotation::New());
Dmitry Grebenyukc49b95ee2023-05-02 05:52:07533}
534
Ramin Halavati87cfa8e62025-06-02 06:48:10535#if BUILDFLAG(IS_CHROMEOS)
Kyungjun Lee90461522023-04-25 06:50:09536void ScreenAIService::PerformOcrAndReturnAXTreeUpdate(
Ramin Halavatib446c022023-03-29 15:36:36537 const SkBitmap& image,
Kyungjun Lee90461522023-04-25 06:50:09538 PerformOcrAndReturnAXTreeUpdateCallback callback) {
Arthur Sonzognic571efb2024-01-26 20:26:18539 std::optional<chrome_screen_ai::VisualAnnotation> annotation_proto =
Ramin Halavati7716b032024-08-19 14:56:20540 PerformOcrAndRecordMetrics(image);
Ramin Halavati89a36c92023-05-05 15:01:03541 ui::AXTreeUpdate update = ConvertVisualAnnotationToTreeUpdate(
542 annotation_proto, gfx::Rect(image.width(), image.height()));
Kyungjun Lee90461522023-04-25 06:50:09543
Ramin Halavati89a36c92023-05-05 15:01:03544 // The original caller is always replied to, and an empty AXTreeUpdate tells
545 // that the annotation function was not successful.
546 std::move(callback).Run(update);
Ramin Halavatib446c022023-03-29 15:36:36547}
Ramin Halavati87cfa8e62025-06-02 06:48:10548#endif // BUILDFLAG(IS_CHROMEOS)
Ramin Halavatib446c022023-03-29 15:36:36549
550void ScreenAIService::ExtractMainContent(const ui::AXTreeUpdate& snapshot,
Ramin Halavatib446c022023-03-29 15:36:36551 ExtractMainContentCallback callback) {
Abigail Klein18673702024-03-05 20:59:06552 ui::AXTree tree;
553 std::optional<std::vector<int32_t>> content_node_ids;
Ramin Halavatib3fe2fe2025-03-20 16:53:03554 bool success = ExtractMainContentInternalAndRecordMetrics(snapshot, tree,
555 content_node_ids);
Abigail Klein18673702024-03-05 20:59:06556
557 if (success) {
558 std::move(callback).Run(*content_node_ids);
559 } else {
560 std::move(callback).Run(std::vector<int32_t>());
561 }
562}
563
564void ScreenAIService::ExtractMainNode(const ui::AXTreeUpdate& snapshot,
565 ExtractMainNodeCallback callback) {
566 ui::AXTree tree;
567 std::optional<std::vector<int32_t>> content_node_ids;
Ramin Halavatib3fe2fe2025-03-20 16:53:03568 bool success = ExtractMainContentInternalAndRecordMetrics(snapshot, tree,
569 content_node_ids);
Abigail Klein18673702024-03-05 20:59:06570
571 if (success) {
572 ui::AXNodeID main_node_id = ComputeMainNode(&tree, *content_node_ids);
573 std::move(callback).Run(main_node_id);
574 } else {
575 std::move(callback).Run(ui::kInvalidAXNodeID);
576 }
577}
578
Mark Schillaci79c82812025-03-07 21:46:37579void ScreenAIService::IdentifyMainNode(const ui::AXTreeUpdate& snapshot,
580 IdentifyMainNodeCallback callback) {
581 ui::AXTree tree;
582 std::optional<std::vector<int32_t>> content_node_ids;
Ramin Halavatib3fe2fe2025-03-20 16:53:03583 bool success = ExtractMainContentInternalAndRecordMetrics(snapshot, tree,
584 content_node_ids);
Mark Schillaci79c82812025-03-07 21:46:37585
586 if (success) {
587 ui::AXNodeID main_node_id = ComputeMainNode(&tree, *content_node_ids);
588 std::move(callback).Run(tree.GetAXTreeID(), main_node_id);
589 } else {
590 std::move(callback).Run(ui::AXTreeIDUnknown(), ui::kInvalidAXNodeID);
591 }
592}
593
Ramin Halavatib3fe2fe2025-03-20 16:53:03594bool ScreenAIService::ExtractMainContentInternalAndRecordMetrics(
Abigail Klein18673702024-03-05 20:59:06595 const ui::AXTreeUpdate& snapshot,
596 ui::AXTree& tree,
597 std::optional<std::vector<int32_t>>& content_node_ids) {
Ramin Halavatid0d6231b2025-03-20 05:36:43598 CHECK(base::Contains(mce_client_types_,
599 screen2x_main_content_extractors_.current_receiver()));
Ramin Halavati1520b702025-04-23 15:24:18600 mce_last_used_ = base::TimeTicks::Now();
Ramin Halavatib3fe2fe2025-03-20 16:53:03601 MainContentExtractionClientTypeForMetrics client_type = GetClientType(
602 mce_client_types_[screen2x_main_content_extractors_.current_receiver()]);
603
604 static crash_reporter::CrashKeyString<2> cpu_info(
605 "main_content_extraction_client");
606 cpu_info.Set(base::StringPrintf("%i", static_cast<int>(client_type)));
607
Ramin Halavatib446c022023-03-29 15:36:36608 // Early return if input is empty.
609 if (snapshot.nodes.empty()) {
Ramin Halavatib3fe2fe2025-03-20 16:53:03610 base::UmaHistogramEnumeration(
611 "Accessibility.ScreenAI.MainContentExtraction.Error.SnapshotEmpty",
612 client_type);
Abigail Klein18673702024-03-05 20:59:06613 return false;
Ramin Halavatib33fc0c2022-05-06 09:32:22614 }
Ramin Halavatie2549a92022-08-02 07:43:19615
Abigail Klein18673702024-03-05 20:59:06616 // Deserialize the snapshot and reserialize it to a view hierarchy proto.
Ramin Halavatib3fe2fe2025-03-20 16:53:03617 if (!tree.Unserialize(snapshot)) {
618 base::UmaHistogramEnumeration(
619 "Accessibility.ScreenAI.MainContentExtraction.Error."
620 "SnapshotUnserialize",
621 client_type);
Ramin Halavati62e034d2024-08-02 08:58:22622 return false;
623 }
624
Ramin Halavatib3fe2fe2025-03-20 16:53:03625 std::optional<ViewHierarchyAndTreeSize> converted_snapshot =
626 SnapshotToViewHierarchy(tree);
627 if (!converted_snapshot) {
628 base::UmaHistogramEnumeration(
629 "Accessibility.ScreenAI.MainContentExtraction.Error.SnapshotProto",
630 client_type);
631 return false;
632 }
633
634 base::TimeTicks start_time = base::TimeTicks::Now();
Ramin Halavatiaa67ee52025-04-25 05:15:53635 base::SequenceBound<HangTimer> hang_timer(background_task_runner_,
636 /*is_ocr=*/false);
637 hang_timer.AsyncCall(&HangTimer::StartTimer);
Ramin Halavati62e034d2024-08-02 08:58:22638 content_node_ids =
639 library_->ExtractMainContent(converted_snapshot->serialized_proto);
Ramin Halavatiaa67ee52025-04-25 05:15:53640 hang_timer.AsyncCall(&HangTimer::Stop);
Ramin Halavatib3fe2fe2025-03-20 16:53:03641 base::TimeDelta elapsed_time = base::TimeTicks::Now() - start_time;
642
643 bool successful =
644 content_node_ids.has_value() && content_node_ids->size() > 0;
Ramin Halavati81b6d922024-06-07 14:20:34645 base::UmaHistogramBoolean(
Ramin Halavatib3fe2fe2025-03-20 16:53:03646 "Accessibility.ScreenAI.MainContentExtraction.Successful2", successful);
647
648 if (!content_node_ids.has_value()) {
649 base::UmaHistogramEnumeration(
650 "Accessibility.ScreenAI.MainContentExtraction.Error.ResultNull",
651 client_type);
652 } else if (content_node_ids->empty()) {
653 base::UmaHistogramEnumeration(
654 "Accessibility.ScreenAI.MainContentExtraction.Error.ResultEmpty",
655 client_type);
656 }
657
Ramin Halavatiac1ee2922025-05-21 17:17:17658 mce_last_used_ = base::TimeTicks::Now();
Ramin Halavatib3fe2fe2025-03-20 16:53:03659 if (successful) {
660 base::UmaHistogramTimes(
661 "Accessibility.ScreenAI.MainContentExtraction.Latency.Success",
662 elapsed_time);
Ramin Halavati5aad701d2023-05-09 17:23:29663 VLOG(2) << "Screen2x returned " << content_node_ids->size() << " node ids.";
Abigail Klein18673702024-03-05 20:59:06664 return true;
665 } else {
Ramin Halavatib3fe2fe2025-03-20 16:53:03666 base::UmaHistogramTimes(
667 "Accessibility.ScreenAI.MainContentExtraction.Latency.Failure",
668 elapsed_time);
Abigail Klein18673702024-03-05 20:59:06669 VLOG(0) << "Screen2x returned no results.";
670 return false;
Ramin Halavati5aad701d2023-05-09 17:23:29671 }
Abigail Klein18673702024-03-05 20:59:06672}
Ramin Halavati89a36c92023-05-05 15:01:03673
Abigail Klein18673702024-03-05 20:59:06674ui::AXNodeID ScreenAIService::ComputeMainNodeForTesting(
675 const ui::AXTree* tree,
676 const std::vector<ui::AXNodeID>& content_node_ids) {
677 return ComputeMainNode(tree, content_node_ids);
Abigail Kleindfdde352023-01-27 21:03:10678}
679
Ramin Halavaticf155012025-05-05 15:50:01680void ScreenAIService::StartShutDownOnIdleTimer() {
681 if (!idle_checking_timer_) {
682 idle_checking_timer_ = std::make_unique<base::RepeatingTimer>();
683 idle_checking_timer_->Start(FROM_HERE, kIdleCheckingDelay, this,
684 &ScreenAIService::ShutDownOnIdle);
685 }
686}
687
Ramin Halavatibe45dde2025-05-01 04:19:10688void ScreenAIService::ShutDownOnIdle() {
Ramin Halavati2acf2362024-08-02 14:22:53689 const base::TimeTicks kIdlenessThreshold =
690 base::TimeTicks::Now() - kIdleCheckingDelay;
Ramin Halavatibe45dde2025-05-01 04:19:10691 if (ocr_last_used_ < kIdlenessThreshold &&
692 mce_last_used_ < kIdlenessThreshold) {
Ramin Halavatif26d7d42025-03-12 16:04:46693 screen_ai_shutdown_handler_->ShuttingDownOnIdle();
Ramin Halavatidc0b2e512024-08-16 17:39:09694 base::Process::TerminateCurrentProcessImmediately(0);
Ramin Halavati2acf2362024-08-02 14:22:53695 }
Ramin Halavati3842e7e2024-07-31 04:39:21696}
697
Ramin Halavati26dcce22022-02-23 13:11:14698} // namespace screen_ai