allow full page viewport

Dockers/puppeteer-api/app/services/browser.py

@@ -188,6 +188,12 @@ async def interactions_service(
 
     async def interactions_operation(page):
         try:
+            # When the API caller sets viewport.height = -1, treat that as "capture full webpage".
+            # We implement this by resizing the viewport to the document's scroll height just
+            # before the final screenshot. (This keeps existing coordinate-click behavior intact.)
+            full_page_via_viewport = False
+            requested_viewport_width = None
+
             async def resolve_viewport_point(px: int, py: int) -> tuple[float, float]:
                 """
                 Playwright's `document.elementFromPoint(x, y)` uses viewport coordinates.
@@ -251,8 +257,15 @@ async def interactions_service(
                         vp_width = getattr(viewport, "width", None)
                         vp_height = getattr(viewport, "height", None)
 
+                    requested_viewport_width = vp_width
                     if vp_width and vp_height:
-                        await page.set_viewport_size({"width": vp_width, "height": vp_height})
+                        if int(vp_height) == -1:
+                            # Defer height calculation until after navigation/interactions.
+                            full_page_via_viewport = True
+                            vp_size = page.viewport_size or {"width": 1280, "height": 720}
+                            await page.set_viewport_size({"width": int(vp_width), "height": int(vp_size.get("height", 720))})
+                        else:
+                            await page.set_viewport_size({"width": int(vp_width), "height": int(vp_height)})
                 except Exception as e:
                     print(f"Error setting viewport: {e}")
 
@@ -405,7 +418,30 @@ async def interactions_service(
             if return_type == "html":
                 return await page.content()
 
-            # For coordinate-based workflows, return the viewport screenshot so pixel coords align.
+            if full_page_via_viewport:
+                try:
+                    # Chromium has practical limits on viewport height; cap to avoid errors.
+                    max_vp_h = 16384
+
+                    # Ensure we capture from the top of the document.
+                    await page.evaluate("() => window.scrollTo({ left: 0, top: 0, behavior: 'auto' })")
+                    await page.wait_for_timeout(250)
+
+                    full_h = await page.evaluate(
+                        "() => Math.max(document.body?.scrollHeight || 0, document.documentElement?.scrollHeight || 0)"
+                    )
+                    target_h = int(min(max(full_h, 1), max_vp_h))
+
+                    cur_vp = page.viewport_size or {"width": 1280, "height": 720}
+                    target_w = int(requested_viewport_width or cur_vp.get("width", 1280))
+
+                    await page.set_viewport_size({"width": target_w, "height": target_h})
+                    await page.wait_for_timeout(500)
+                except Exception as e:
+                    print(f"Error resizing viewport for full-page capture: {e}")
+
+            # For coordinate-based workflows, default is viewport screenshot so pixel coords align.
+            # If full_page_via_viewport is enabled, the viewport has been expanded to the full document height.
             return await page.screenshot(full_page=False, type="png")
 
         except Exception as e: