simplex 1.2.1__tar.gz → 1.2.3__tar.gz

{simplex-1.2.1 → simplex-1.2.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: simplex
-Version: 1.2.1
+Version: 1.2.3
 Summary: Official Python SDK for Simplex API
 Home-page: https://github.com/shreyka/simplex-python
 Author: Simplex Labs, Inc.

{simplex-1.2.1 → simplex-1.2.3}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="simplex",
-    version="1.2.1",
+    version="1.2.3",
     packages=find_packages(),
     install_requires=[
         "openai>=1.0.0",

{simplex-1.2.1 → simplex-1.2.3}/simplex/simplex.py

@@ -26,12 +26,11 @@ class Simplex:
             self.driver = self.browser.new_page()
         else:
             self.playwright = None
-            if len(self.browser.contexts) > 0 and len(self.browser.contexts[0].pages) > 0:
-                self.driver = self.browser.contexts[0].pages[0]
-            else:
-                self.driver = self.browser.new_page()
+            self.driver = self.browser.contexts[0].pages[0]
+            self.driver.set_default_navigation_timeout(0)
+            self.driver.set_default_timeout(0)
 
-    def find_element(self, element_description: str, state: Image.Image | None = None, annotate: bool = True) -> List[int]:
+    def extract_bbox(self, element_description: str, state: Image.Image | None = None, annotate: bool = True) -> List[int]:
         """
         Find an element in the screenshot using the element description
 
@@ -42,6 +41,7 @@ class Simplex:
         Returns:
             bounding_box (tuple): [x1, y1, x2, y2] bounding box of the found element
         """
+        print(f"[SIMPLEX] Finding element \"{element_description}\"...")
         if state is None:
             state = self.take_stable_screenshot()
 
@@ -58,7 +58,7 @@ class Simplex:
             'element_description': (None, element_description),
             'api_key': (None, self.api_key)
         }        
-
+        print("[SIMPLEX] Sending screenshot to server...")
         # Make the request
         response = requests.post(
             endpoint,
@@ -67,7 +67,6 @@ class Simplex:
 
 
         # Print the results
-        print(f"Status Code: {response.status_code}")
         if response.status_code == 200:
             res = response.json()
             bbox = [int(res['x1']), int(res['y1']), int(res['x2']), int(res['y2'])]
@@ -87,7 +86,7 @@ class Simplex:
                         const overlay = document.createElement('div');
                         overlay.id = 'simplex-bbox-overlay';
                         overlay.style.position = 'fixed';
-                        overlay.style.border = '2px dashed rgba(74, 144, 226, 1)';
+                        overlay.style.border = '2px dashed rgba(0, 255, 0, 1)';
                         overlay.style.background = 'rgba(74, 144, 226, 0.1)';
                         overlay.style.animation = 'marching-ants 0.5s linear infinite';
                         overlay.style.left = bbox[0] + 'px';
@@ -124,7 +123,7 @@ class Simplex:
                 self.driver.wait_for_selector('#simplex-bbox-overlay')
             return bbox 
         else:
-            print("Error:", response.text)
+            print("[SIMPLEX] Error:", response.text)
 
     def step_to_action(self, step_description: str, state: Image.Image | None = None) -> List[List[str]]:
         """
@@ -168,7 +167,7 @@ class Simplex:
             actions = [[action.strip() for action in action_pair] for action_pair in actions]
             return actions
         else:
-            print(f"Error: {response.status_code}")
+            print(f"[SIMPLEX] Error: {response.status_code}")
             print(response.text)
             return []
         
@@ -182,8 +181,40 @@ class Simplex:
         """
         if new_tab:
             self.driver = self.browser.new_page()
+        self.driver.wait_for_load_state()
+        print(f"[SIMPLEX] Navigating to URL {url}...")
         self.driver.goto(url)
 
+    def click(self, element_description: str, annotate: bool = True) -> None:
+        """
+        Click on an element
+        """
+        self.execute_action(["CLICK", element_description], annotate=annotate)
+
+    def type(self, text: str) -> None:
+        """
+        Type text into an element
+        """
+        self.execute_action(["TYPE", text])
+
+    def press_enter(self, annotate: bool = True) -> None:
+        """
+        Press enter
+        """
+        self.execute_action(["ENTER", ""], annotate=annotate)
+
+    def scroll(self, scroll_amount: int, annotate: bool = True) -> None:
+        """
+        Scroll the page
+        """
+        self.execute_action(["SCROLL", scroll_amount], annotate=annotate)
+
+    def wait(self, wait_time: int, annotate: bool = True) -> None:
+        """
+        Wait for a given amount of time
+        """
+        self.execute_action(["WAIT", wait_time], annotate=annotate)
+
     def execute_action(self, action: List[List[str]], state: Image.Image | None = None, annotate: bool = True) -> None:
         """
         Execute an action with playwright driver
@@ -192,34 +223,30 @@ class Simplex:
             action (List[List[str]]): List of actions to perform
         """
         action_type, description = action
-        if state is None:
-            state = self.take_stable_screenshot()
-
         try:
             if action_type == "CLICK":
-                bbox = self.find_element(description, state, annotate=annotate)
+                bbox = self.extract_bbox(description, state, annotate=annotate)
                 center_x, center_y = center_bbox(bbox)
                 self.driver.mouse.click(center_x, center_y)
-
-            elif action_type == "HOVER":
-                bbox = self.find_element(description, state, annotate=annotate)
-                center_x, center_y = center_bbox(bbox)
-                self.driver.mouse.move(center_x, center_y)
-
+                print(f"[SIMPLEX] Clicked on element \"{description}\"")
             elif action_type == "TYPE":
                 self.driver.keyboard.type(description)
+                print(f"[SIMPLEX] Typed \"{description}\"")
 
             elif action_type == "ENTER":
                 self.driver.keyboard.press("Enter")
+                print(f"[SIMPLEX] Pressed enter")
 
             elif action_type == "SCROLL":
                 self.driver.mouse.wheel(0, int(description))
+                print(f"[SIMPLEX] Scrolled {description} pixels")
 
             elif action_type == "WAIT":
                 self.driver.wait_for_timeout(int(description))
+                print(f"[SIMPLEX] Waited {description} seconds")
 
         except Exception as e:
-            print(f"Error executing action: {e}")
+            print(f"[SIMPLEX] Error executing action: {e}")
             return None
         
     def do(self, step_description: str, annotate: bool = True) -> None:
@@ -230,6 +257,57 @@ class Simplex:
         actions = self.step_to_action(step_description, state)
         for action in actions:
             self.execute_action(action, annotate=annotate)
+    
+    def extract_text(self, element_description: str, state: Image.Image | None = None) -> List[str] | None:
+        """
+        Extract an element text from the page
+        """
+        print(f"[SIMPLEX] Finding element \"{element_description}\"...")
+        if state is None:
+            state = self.take_stable_screenshot()
+
+        endpoint = f"{BASE_URL}/extract-text"
+        
+        # Convert PIL Image to bytes
+        img_byte_arr = io.BytesIO()
+        state.save(img_byte_arr, format='PNG')
+        img_byte_arr = img_byte_arr.getvalue()
+        
+        # Prepare multipart form data
+        files = {
+            'image_data': ('screenshot.png', img_byte_arr, 'image/png'),
+            'element_description': (None, element_description),
+            'api_key': (None, self.api_key)
+        }        
+        print("[SIMPLEX] Sending screenshot to server...")
+
+        # Make the request
+        response = requests.post(
+            endpoint,
+            files=files
+        )
+        
+        # Print the results
+        if response.status_code == 200:
+            res = response.json()
+            text = res['text']
+            return text
+        else:
+            print("[SIMPLEX] Error:", response.text)
+            return None
+        
+    def extract_image(self, element_description: str, state: Image.Image | None = None) -> Image.Image | None:
+        """
+        Extract an element image from the page
+        """
+        if state is None:
+            state = self.take_stable_screenshot()
+            
+        bbox = self.extract_bbox(element_description, state)
+        cropped_state = state.crop((bbox[0], bbox[1], bbox[2], bbox[3]))
+        return cropped_state
+        
+        
             
     def take_stable_screenshot(self) -> Image.Image:
         """
@@ -238,6 +316,7 @@ class Simplex:
         Returns:
             PIL.Image.Image: Screenshot of the current page
         """
+        print("[SIMPLEX] Taking screenshot of the page...")
         self.driver.wait_for_load_state('networkidle')
         return screenshot_to_image(self.driver.screenshot())
             

{simplex-1.2.1 → simplex-1.2.3}/simplex.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: simplex
-Version: 1.2.1
+Version: 1.2.3
 Summary: Official Python SDK for Simplex API
 Home-page: https://github.com/shreyka/simplex-python
 Author: Simplex Labs, Inc.

{simplex-1.2.1 → simplex-1.2.3}/tests/test_local.py

@@ -9,6 +9,7 @@ from playwright.sync_api import sync_playwright
 from PIL import Image
 import time
 import os
+from browserbase import Browserbase
 from hyperbrowser import Hyperbrowser
 from hyperbrowser.models.session import CreateSessionParams
 
@@ -153,14 +154,32 @@ def test_find_element():
 def test_find_element_2():
 
     simplex = Simplex(api_key=os.getenv("SIMPLEX_API_KEY"))
-    simplex.goto("https://www.cgtrader.com/")
-    simplex.find_element("search bar")
+    # simplex.goto("https://www.cgtrader.com/")
+    state = Image.open("nvidia-stock.png")
+    bbox = simplex.extract_bbox("nvidia stock graph", state)
+    print(bbox)
+
+    state.crop(bbox)
+    state.save("cropped_state.png")
+
+
+
 
+
+import requests
 def test_hyperbrowser_integration():
     """Test Simplex integration with Hyperbrowser"""
+    import time
+    start_time = time.time()
+
+    bb = Browserbase(api_key=os.getenv("BROWSERBASE_API_KEY"))
+    session_start = time.time()
+    session = bb.sessions.create(project_id=os.getenv("BROWSERBASE_PROJECT_ID"))
+    print(session.id)
+    print(f"Session creation took {time.time() - session_start:.2f}s")
 
     # Initialize Hyperbrowser client
-    client = Hyperbrowser(api_key=os.getenv("HYPERBROWSER_API_KEY"))
+    # client = Hyperbrowser(api_key=os.getenv("HYPERBROWSER_API_KEY"))
     
     # Create session params with CAPTCHA solving enabled
     session_params = CreateSessionParams(
@@ -169,28 +188,45 @@ def test_hyperbrowser_integration():
     )
     
     # Create a new session with params
-    session = client.sessions.create(session_params)
-    ws_endpoint = session.ws_endpoint
+    # session = client.sessions.create(session_params)
+    ws_endpoint = session.connect_url
 
     try:
         with sync_playwright() as p:
             # Connect browser to Hyperbrowser session
+            browser_start = time.time()
             browser = p.chromium.connect_over_cdp(ws_endpoint)
-            
+            print(f"Browser connection took {time.time() - browser_start:.2f}s")
+            url = f"https://api.browserbase.com/v1/sessions/{session.id}/debug"
+            headers = {"X-BB-API-Key": os.getenv("BROWSERBASE_API_KEY")}
+            response = requests.get(url, headers=headers)
+            print(response.text)
+            print(response.json()["pages"][0]["debuggerUrl"])
+
             # Initialize Simplex with the Hyperbrowser-connected page
             simplex = Simplex(api_key=os.getenv("SIMPLEX_API_KEY"), browser=browser)
             
             # Test basic functionality
+            nav_start = time.time()
             simplex.goto("https://www.turbosquid.com/")
-            simplex.do("search for iphone")
+            print(f"Navigation took {time.time() - nav_start:.2f}s")
+            
+            # search_start = time.time()
+            # simplex.do("search for iphone")
+            # print(f"Search action took {time.time() - search_start:.2f}s")
             
             # Verify the search worked by finding the search results
-            bbox = simplex.find_element("search results")
+            find_start = time.time()
+            bbox = simplex.find_element("search results", annotate=True)
+            print(f"Finding results took {time.time() - find_start:.2f}s")
             assert bbox is not None, "Search results not found"
             
     finally:
         # Always stop the Hyperbrowser session
-        client.sessions.stop(session.id)
+        # client.sessions.stop(session.id)
+        browser.close()
+        print(f"Total test time: {time.time() - start_time:.2f}s")
+        
 
 
 def execute_action_test():
@@ -216,7 +252,28 @@ def execute_action_test():
         browser.close()
         playwright.stop()
 
+
+def test_extract_text():
+    simplex = Simplex(api_key=os.getenv("SIMPLEX_API_KEY"))
+    simplex.goto("https://www.cgtrader.com/3d-models/electronics/other/apple-watch-ultra-2022-hq-3d-model")
+    time.sleep(2)
+    start = time.time()
+    text = simplex.extract_text("type of license")
+    end = time.time()
+    print(f"Time taken: {end - start:.2f}s")
+    print(text)
+
+
+def test_extract_image():
+    simplex = Simplex(api_key=os.getenv("SIMPLEX_API_KEY"))
+    state = Image.open("nvidia-stock.png")
+    image = simplex.extract_image("stock trends and market summary", state)
+    image.save("cropped_image.png")
+
+    print(type(simplex.driver))
+
+
 if __name__ == "__main__":
-    execute_action_test()
+    test_extract_image()