droidrun 0.5.1.dev1__tar.gz → 0.5.1.dev3__tar.gz

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: droidrun
-Version: 0.5.1.dev1
+Version: 0.5.1.dev3
 Summary: A framework for controlling Android devices through LLM agents
 Project-URL: Homepage, https://github.com/droidrun/droidrun
 Project-URL: Bug Tracker, https://github.com/droidrun/droidrun/issues

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/droidrun/agent/droid/droid_agent.py

@@ -1085,7 +1085,7 @@ class DroidAgent(Workflow):
                 if self.config.logging.debug:
                     logger.error(traceback.format_exc())
 
-        # Capture final screenshot (independent of trajectory persistence)
+        # Capture final screenshot and UI state (independent of trajectory persistence)
         vision_any = (
             self.config.agent.manager.vision
             or self.config.agent.executor.vision
@@ -1111,6 +1111,15 @@ class DroidAgent(Workflow):
             except Exception as e:
                 logger.warning(f"Failed to capture final screenshot: {e}")
 
+            try:
+                ui_state = await self.state_provider.get_state()
+                ctx.write_event_to_stream(
+                    RecordUIStateEvent(ui_state=ui_state.elements)
+                )
+                logger.debug("📋 Final UI state captured")
+            except Exception as e:
+                logger.warning(f"Failed to capture final UI state: {e}")
+
         # Save trajectory to disk
         if self.config.logging.save_trajectory != "none":
             # Populate macro data from RecordingDriver log

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/droidrun/tools/driver/base.py

@@ -52,7 +52,9 @@ class DeviceDriver:
         """Swipe from (x1, y1) to (x2, y2)."""
         raise NotImplementedError
 
-    async def input_text(self, text: str, clear: bool = False) -> bool:
+    async def input_text(
+        self, text: str, clear: bool = False, stealth: bool = False, wpm: int = 0,
+    ) -> bool:
         """Type *text* into the currently focused field.
 
         Returns ``True`` on success, ``False`` on failure.

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/droidrun/tools/driver/cloud.py

@@ -115,10 +115,21 @@ class CloudDriver(DeviceDriver):
             )
         )
 
-    async def input_text(self, text: str, clear: bool = False) -> bool:
+    async def input_text(
+        self, text: str, clear: bool = False, stealth: bool = False, wpm: int = 0,
+    ) -> bool:
+        extra_body: dict = {}
+        if stealth:
+            extra_body["stealth"] = True
+        if wpm:
+            extra_body["wpm"] = wpm
         await self._call(
             self._client.devices.keyboard.write(
-                self.device_id, text=text, clear=clear, **self._display_kw
+                self.device_id,
+                text=text,
+                clear=clear,
+                extra_body=extra_body or None,
+                **self._display_kw,
             )
         )
         return True

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "droidrun"
-version = "0.5.1.dev1"
+version = "0.5.1.dev3"
 description = "A framework for controlling Android devices through LLM agents"
 authors = [{ name = "Niels Schmidt", email = "niels.schmidt@droidrun.ai" }]
 dependencies = [

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/uv.lock

@@ -760,7 +760,7 @@ wheels = [
 
 [[package]]
 name = "droidrun"
-version = "0.5.1.dev1"
+version = "0.5.1.dev3"
 source = { editable = "." }
 dependencies = [
     { name = "aiofiles" },

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/droidrun/agent/manager/manager_agent.py

@@ -387,6 +387,30 @@ class ManagerAgent(Workflow):
         """Gather context and prepare manager prompt."""
         logger.debug("💬 Preparing manager context...")
 
+        # Capture screenshot if needed
+        screenshot = None
+        if self.vision or self._stream_screenshots or self.save_trajectory != "none":
+            try:
+                screenshot = await self.action_ctx.driver.screenshot()
+
+                if screenshot:
+                    ctx.write_event_to_stream(ScreenshotEvent(screenshot=screenshot))
+                    parent_span = trace.get_current_span()
+                    record_langfuse_screenshot(
+                        screenshot,
+                        parent_span=parent_span,
+                        screenshots_enabled=bool(
+                            self.tracing_config
+                            and self.tracing_config.langfuse_screenshots
+                        ),
+                        vision_enabled=self.vision,
+                    )
+                    logger.debug("📸 Screenshot captured for Manager")
+            except DeviceDisconnectedError:
+                raise
+            except Exception as e:
+                logger.warning(f"Failed to capture screenshot: {e}")
+
         # Get and format device state
         ui_state = await self.state_provider.get_state()
         self.action_ctx.ui = ui_state
@@ -422,30 +446,6 @@ class ManagerAgent(Workflow):
         else:
             self.shared_state.app_card = ""
 
-        # Capture screenshot if needed
-        screenshot = None
-        if self.vision or self._stream_screenshots or self.save_trajectory != "none":
-            try:
-                screenshot = await self.action_ctx.driver.screenshot()
-
-                if screenshot:
-                    ctx.write_event_to_stream(ScreenshotEvent(screenshot=screenshot))
-                    parent_span = trace.get_current_span()
-                    record_langfuse_screenshot(
-                        screenshot,
-                        parent_span=parent_span,
-                        screenshots_enabled=bool(
-                            self.tracing_config
-                            and self.tracing_config.langfuse_screenshots
-                        ),
-                        vision_enabled=self.vision,
-                    )
-                    logger.debug("📸 Screenshot captured for Manager")
-            except DeviceDisconnectedError:
-                raise
-            except Exception as e:
-                logger.warning(f"Failed to capture screenshot: {e}")
-
         # Detect text manipulation mode
         focused_text_clean = self.shared_state.focused_text.replace("'", "").strip()
         has_text_to_modify = focused_text_clean != ""

{droidrun-0.5.1.dev1 → droidrun-0.5.1.dev3}/droidrun/agent/manager/stateless_manager_agent.py

@@ -175,24 +175,6 @@ class StatelessManagerAgent(Workflow):
     async def prepare_context(
         self, ctx: Context, ev: StartEvent
     ) -> ManagerContextEvent:
-        ui_state = await self.state_provider.get_state()
-        self.action_ctx.ui = ui_state
-
-        self.shared_state.previous_formatted_device_state = (
-            self.shared_state.formatted_device_state
-        )
-        self.shared_state.formatted_device_state = ui_state.formatted_text
-        self.shared_state.focused_text = ui_state.focused_text
-        self.shared_state.a11y_tree = ui_state.elements
-        self.shared_state.phone_state = ui_state.phone_state
-
-        self.shared_state.update_current_app(
-            package_name=ui_state.phone_state.get("packageName", "Unknown"),
-            activity_name=ui_state.phone_state.get("currentApp", "Unknown"),
-        )
-
-        ctx.write_event_to_stream(RecordUIStateEvent(ui_state=ui_state.elements))
-
         screenshot = None
         if self.vision or self.save_trajectory != "none":
             try:
@@ -215,6 +197,24 @@ class StatelessManagerAgent(Workflow):
             except Exception as e:
                 logger.warning(f"Failed to capture screenshot: {e}")
 
+        ui_state = await self.state_provider.get_state()
+        self.action_ctx.ui = ui_state
+
+        self.shared_state.previous_formatted_device_state = (
+            self.shared_state.formatted_device_state
+        )
+        self.shared_state.formatted_device_state = ui_state.formatted_text
+        self.shared_state.focused_text = ui_state.focused_text
+        self.shared_state.a11y_tree = ui_state.elements
+        self.shared_state.phone_state = ui_state.phone_state
+
+        self.shared_state.update_current_app(
+            package_name=ui_state.phone_state.get("packageName", "Unknown"),
+            activity_name=ui_state.phone_state.get("currentApp", "Unknown"),
+        )
+
+        ctx.write_event_to_stream(RecordUIStateEvent(ui_state=ui_state.elements))
+
         focused_text_clean = self.shared_state.focused_text.replace("'", "").strip()
         has_text_to_modify = focused_text_clean != ""