droidrun 0.3.8__tar.gz → 0.3.9__tar.gz

{droidrun-0.3.8 → droidrun-0.3.9}/.gitignore

@@ -23,7 +23,4 @@ patch_apis.py
 .git
 .arize-phoenix
 
-
-uv.lock
-
 todo.txt

droidrun-0.3.9/.python-version

@@ -0,0 +1 @@
+3.13.7

{droidrun-0.3.8 → droidrun-0.3.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: droidrun
-Version: 0.3.8
+Version: 0.3.9
 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
@@ -16,6 +16,8 @@ Classifier: License :: OSI Approved :: MIT License
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Communications :: Chat
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
@@ -28,13 +30,14 @@ Classifier: Topic :: Utilities
 Requires-Python: >=3.11
 Requires-Dist: adbutils>=2.10.2
 Requires-Dist: apkutils==2.0.0
-Requires-Dist: llama-index==0.13.6
-Requires-Dist: posthog>=6.7.4
-Requires-Dist: pydantic>=2.11.7
+Requires-Dist: llama-index-llms-google-genai>=0.6.2
+Requires-Dist: llama-index==0.14.4
+Requires-Dist: posthog>=6.7.6
+Requires-Dist: pydantic>=2.11.10
 Requires-Dist: rich>=14.1.0
 Provides-Extra: anthropic
 Requires-Dist: anthropic>=0.67.0; extra == 'anthropic'
-Requires-Dist: llama-index-llms-anthropic>=0.8.6; extra == 'anthropic'
+Requires-Dist: llama-index-llms-anthropic<0.9.0,>=0.8.6; extra == 'anthropic'
 Provides-Extra: deepseek
 Requires-Dist: llama-index-llms-deepseek>=0.2.1; extra == 'deepseek'
 Provides-Extra: dev
@@ -88,7 +91,7 @@ DroidRun is a powerful framework for controlling Android and iOS devices through
 ## 📦 Installation
 
 ```bash
-pip install droidrun[google,anthropic,openai,deepseek,ollama,dev]
+pip install 'droidrun[google,anthropic,openai,deepseek,ollama,dev]'
 ```
 
 ## 🚀 Quickstart

{droidrun-0.3.8 → droidrun-0.3.9}/README.md

@@ -33,7 +33,7 @@ DroidRun is a powerful framework for controlling Android and iOS devices through
 ## 📦 Installation
 
 ```bash
-pip install droidrun[google,anthropic,openai,deepseek,ollama,dev]
+pip install 'droidrun[google,anthropic,openai,deepseek,ollama,dev]'
 ```
 
 ## 🚀 Quickstart

{droidrun-0.3.8 → droidrun-0.3.9}/docs/docs.json

@@ -14,7 +14,7 @@
         "tab": "Framework",
         "versions": [
           {
-            "version": "0.3.6",
+            "version": "0.3.8",
             "groups": [
               {
                 "group": "Introduction",

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/guides/cli.mdx

@@ -13,7 +13,7 @@ DroidRun lets you control Android devices using natural language and LLM agents.
 <Steps>
     <Step title="Install DroidRun and its dependencies. Choose which ever provider you'd like to use.">
        ```sh 
-       pip install droidrun[google,anthropic,openai,deepseek,ollama,dev]
+       pip install 'droidrun[google,anthropic,openai,deepseek,ollama,dev]'
        ```
     </Step>
     <Step title="Ensure your Android device is connected and the DroidRun Portal is installed">

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/guides/gemini.mdx

@@ -32,7 +32,7 @@ Make sure you've set up and enabled the Droidrun Portal.
 ## 2. Install Required Python Packages
 
 ```sh
-pip install droidrun[google]
+pip install 'droidrun[google]'
 ```
 
 ## 3. Example: Using Droidrun with Gemini LLM

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/guides/ollama.mdx

@@ -47,7 +47,7 @@ ollama pull llama4
 Make sure you have the required Python packages:
 
 ```sh
-pip install droidrun[ollama]
+pip install 'droidrun[ollama]'
 ```
 
 ## 3. Example: Using Droidrun with Ollama LLM

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/guides/openailike.mdx

@@ -36,7 +36,7 @@ Get your API key for the provider you've choosen
 ## 2. Install the required Python packages:
 
 ```sh
-pip install droidrun[openai]
+pip install 'droidrun[openai]'
 ```
 
 ## 3. Example: Using Droidrun with OpenAI-like LLM

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/overview.mdx

@@ -29,7 +29,7 @@ DroidRun empowers you to automate sophisticated mobile workflows through intelli
 <Card icon="mobile" title="Physical Device" href="/v3/quickstart" arrow>
  - Connect your own physical Android device for direct automation
 </Card>
-<Card icon="cloud" title="Cloud Environment" href="https://droidrun.ai/cloud">
+<Card icon="cloud" title="Cloud Environment" href="https://cloud.droidrun.ai">
  - Access our managed cloud environment for instant mobile app automation without any setup.
 </Card>
 </CardGroup>

{droidrun-0.3.8 → droidrun-0.3.9}/docs/v3/quickstart.mdx

@@ -28,7 +28,7 @@ Before installing DroidRun, ensure you have:
 ### Install from PyPI
 Choose which ever provider you'd like to use. If you encounter dependency resolution issues use [`uv`](https://docs.astral.sh/uv/getting-started/installation/).
 ```bash
-pip install droidrun[google,anthropic,openai,deepseek,ollama,dev]
+pip install 'droidrun[google,anthropic,openai,deepseek,ollama,dev]'
 ```
 
 ### Setup the Portal APK

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/codeact/codeact_agent.py

@@ -18,7 +18,8 @@ from droidrun.agent.codeact.events import (
     TaskThinkingEvent,
     EpisodicMemoryEvent,
 )
-from droidrun.agent.common.events import ScreenshotEvent, RecordUIStateEvent
+from droidrun.agent.common.constants import LLM_HISTORY_LIMIT
+from droidrun.agent.common.events import RecordUIStateEvent, ScreenshotEvent
 from droidrun.agent.usage import get_usage_from_response
 from droidrun.agent.utils import chat_utils
 from droidrun.agent.utils.executer import SimpleCodeExecutor
@@ -109,7 +110,8 @@ class CodeActAgent(Workflow):
         """Prepare chat history from user input."""
         logger.info("💬 Preparing chat for task execution...")
 
-        self.chat_memory: Memory = await ctx.get(
+
+        self.chat_memory: Memory = await ctx.store.get(
             "chat_memory", default=Memory.from_defaults()
         )
 
@@ -135,7 +137,7 @@ class CodeActAgent(Workflow):
 
         await self.chat_memory.aput(self.user_message)
 
-        await ctx.set("chat_memory", self.chat_memory)
+        await ctx.store.set("chat_memory", self.chat_memory)
         input_messages = self.chat_memory.get_all()
         return TaskInputEvent(input=input_messages)
 
@@ -162,7 +164,7 @@ class CodeActAgent(Workflow):
         model = self.llm.class_name()
         
         if "remember" in self.tool_list and self.remembered_info:
-            await ctx.set("remembered_info", self.remembered_info)
+            await ctx.store.set("remembered_info", self.remembered_info)
             chat_history = await chat_utils.add_memory_block(self.remembered_info, chat_history)
 
         for context in self.required_context:
@@ -171,7 +173,7 @@ class CodeActAgent(Workflow):
                 screenshot = (self.tools.take_screenshot())[1]
                 ctx.write_event_to_stream(ScreenshotEvent(screenshot=screenshot))
 
-                await ctx.set("screenshot", screenshot)
+                await ctx.store.set("screenshot", screenshot)
                 if model == "DeepSeek":
                     logger.warning(
                         "[yellow]DeepSeek doesnt support images. Disabling screenshots[/]"
@@ -182,7 +184,7 @@ class CodeActAgent(Workflow):
             if context == "ui_state":
                 try:
                     state = self.tools.get_state()
-                    await ctx.set("ui_state", state["a11y_tree"])
+                    await ctx.store.set("ui_state", state["a11y_tree"])
                     ctx.write_event_to_stream(RecordUIStateEvent(ui_state=state["a11y_tree"]))
                     chat_history = await chat_utils.add_ui_text_block(
                         state["a11y_tree"], chat_history
@@ -320,7 +322,7 @@ class CodeActAgent(Workflow):
     async def finalize(self, ev: TaskEndEvent, ctx: Context) -> StopEvent:
         """Finalize the workflow."""
         self.tools.finished = False
-        await ctx.set("chat_memory", self.chat_memory)
+        await ctx.store.set("chat_memory", self.chat_memory)
         
         # Add final state observation to episodic memory
         if self.vision:
@@ -347,14 +349,15 @@ class CodeActAgent(Workflow):
         self, ctx: Context, chat_history: List[ChatMessage]
     ) -> ChatResponse | None:
         logger.debug("🔍 Getting LLM response...")
-        messages_to_send = [self.system_prompt] + chat_history
+        limited_history = self._limit_history(chat_history)
+        messages_to_send = [self.system_prompt] + limited_history
         messages_to_send = [chat_utils.message_copy(msg) for msg in messages_to_send]
         try:
             response = await self.llm.achat(messages=messages_to_send)
             logger.debug("🔍 Received LLM response.")
 
             filtered_chat_history = []
-            for msg in chat_history:
+            for msg in limited_history:
                 filtered_msg = chat_utils.message_copy(msg)
                 if hasattr(filtered_msg, "blocks") and filtered_msg.blocks:
                     filtered_msg.blocks = [
@@ -379,9 +382,10 @@ class CodeActAgent(Workflow):
                 chat_history=chat_history_str,
                 response=response_str,
                 timestamp=time.time(),
-                screenshot=(await ctx.get("screenshot", None))
+                screenshot=(await ctx.store.get("screenshot", None))
             )
 
+
             self.episodic_memory.steps.append(step)
 
             assert hasattr(
@@ -403,12 +407,46 @@ class CodeActAgent(Workflow):
                     time.sleep(40)
                 logger.debug("🔍 Retrying call to LLM...")
                 response = await self.llm.achat(messages=messages_to_send)
+            elif (
+                self.llm.class_name() == "Anthropic_LLM"
+                and "overloaded_error" in str(e)
+            ):
+                # Use exponential backoff for Anthropic errors
+                if not hasattr(self, '_anthropic_retry_count'):
+                    self._anthropic_retry_count = 0
+                self._anthropic_retry_count += 1
+                seconds = min(2 ** self._anthropic_retry_count, 60)  # Cap at 60 seconds
+                logger.error(f"Anthropic overload error. Retrying in {seconds} seconds... (attempt {self._anthropic_retry_count})")
+                time.sleep(seconds)
+                logger.debug("🔍 Retrying call to LLM...")
+                response = await self.llm.achat(messages=messages_to_send)
+                self._anthropic_retry_count = 0  # Reset on success
             else:
                 logger.error(f"Could not get an answer from LLM: {repr(e)}")
                 raise e
         logger.debug("  - Received response from LLM.")
         return response
 
+    def _limit_history(
+        self, chat_history: List[ChatMessage]
+    ) -> List[ChatMessage]:
+        if LLM_HISTORY_LIMIT <= 0:
+            return chat_history
+
+        max_messages = LLM_HISTORY_LIMIT * 2
+        if len(chat_history) <= max_messages:
+            return chat_history
+
+        preserved_head: List[ChatMessage] = []
+        if chat_history and chat_history[0].role == "user":
+            preserved_head = [chat_history[0]]
+
+        tail = chat_history[-max_messages:]
+        if preserved_head and preserved_head[0] in tail:
+            preserved_head = []
+
+        return preserved_head + tail
+
     async def _add_final_state_observation(self, ctx: Context) -> None:
         """Add the current UI state and screenshot as the final observation step."""
         try:

droidrun-0.3.9/droidrun/agent/common/constants.py

@@ -0,0 +1,2 @@
+"""Max number of recent conversation steps to include in LLM prompt"""
+LLM_HISTORY_LIMIT = 20

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/common/events.py

@@ -48,4 +48,4 @@ class StartAppEvent(MacroEvent):
     activity: str = None
 
 class RecordUIStateEvent(Event):
-    ui_state: list[Dict[str, Any]]
+    ui_state: list[Dict[str, Any]]

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/droid/droid_agent.py

@@ -165,7 +165,6 @@ class DroidAgent(Workflow):
                 timeout=timeout,
                 debug=debug,
             )
-            self.add_workflows(planner_agent=self.planner_agent)
             self.max_codeact_steps = 5
 
             if self.reflection:
@@ -335,7 +334,6 @@ class DroidAgent(Workflow):
         self,
         ctx: Context,
         ev: ReasoningLogicEvent,
-        planner_agent: Workflow = MockWorkflow(),
     ) -> FinalizeEvent | CodeActExecuteEvent:
         try:
             if self.step_counter >= self.max_steps:
@@ -352,7 +350,7 @@ class DroidAgent(Workflow):
             self.step_counter += 1
 
             if ev.reflection:
-                handler = planner_agent.run(
+                handler = self.planner_agent.run(
                     remembered_info=self.tools_instance.memory, reflection=ev.reflection
                 )
             else:
@@ -365,7 +363,7 @@ class DroidAgent(Workflow):
 
                 logger.debug(f"Planning step {self.step_counter}/{self.max_steps}")
 
-                handler = planner_agent.run(
+                handler = self.planner_agent.run(
                     remembered_info=self.tools_instance.memory, reflection=None
                 )
 

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/planner/planner_agent.py

@@ -18,7 +18,8 @@ from droidrun.agent.utils.executer import SimpleCodeExecutor
 from droidrun.agent.utils import chat_utils
 from droidrun.agent.context.task_manager import TaskManager
 from droidrun.tools import Tools
-from droidrun.agent.common.events import ScreenshotEvent, RecordUIStateEvent
+from droidrun.agent.common.constants import LLM_HISTORY_LIMIT
+from droidrun.agent.common.events import RecordUIStateEvent, ScreenshotEvent
 from droidrun.agent.planner.events import (
     PlanInputEvent,
     PlanCreatedEvent,
@@ -97,7 +98,7 @@ class PlannerAgent(Workflow):
     async def prepare_chat(self, ctx: Context, ev: StartEvent) -> PlanInputEvent:
         logger.info("💬 Preparing planning session...")
 
-        self.chat_memory: Memory = await ctx.get(
+        self.chat_memory: Memory = await ctx.store.get(
             "chat_memory", default=Memory.from_defaults()
         )
         await self.chat_memory.aput(self.user_message)
@@ -134,19 +135,19 @@ class PlannerAgent(Workflow):
         if self.vision:
             screenshot = (self.tools_instance.take_screenshot())[1]
             ctx.write_event_to_stream(ScreenshotEvent(screenshot=screenshot))
-            await ctx.set("screenshot", screenshot)
+            await ctx.store.set("screenshot", screenshot)
 
         try:
             state = self.tools_instance.get_state()
-            await ctx.set("ui_state", state["a11y_tree"])
-            await ctx.set("phone_state", state["phone_state"])
+            await ctx.store.set("ui_state", state["a11y_tree"])
+            await ctx.store.set("phone_state", state["phone_state"])
             ctx.write_event_to_stream(RecordUIStateEvent(ui_state=state["a11y_tree"]))
         except Exception as e:
             logger.warning(f"⚠️ Error retrieving state from the connected device. Is the Accessibility Service enabled?")
 
 
-        await ctx.set("remembered_info", self.remembered_info)
-        await ctx.set("reflection", self.reflection)
+        await ctx.store.set("remembered_info", self.remembered_info)
+        await ctx.store.set("reflection", self.reflection)
 
         response = await self._get_llm_response(ctx, chat_history)
         try:
@@ -237,7 +238,7 @@ wrap your code inside this:
     @step
     async def finalize(self, ev: PlanCreatedEvent, ctx: Context) -> StopEvent:
         """Finalize the workflow."""
-        await ctx.set("chat_memory", self.chat_memory)
+        await ctx.store.set("chat_memory", self.chat_memory)
 
         result = {}
         result.update(
@@ -263,7 +264,7 @@ wrap your code inside this:
                     )
                 else:
                     chat_history = await chat_utils.add_screenshot_image_block(
-                        await ctx.get("screenshot"), chat_history
+                        await ctx.store.get("screenshot"), chat_history
                     )                   
 
 
@@ -275,18 +276,19 @@ wrap your code inside this:
                 chat_history,
             )
 
-            remembered_info = await ctx.get("remembered_info", default=None)
+            remembered_info = await ctx.store.get("remembered_info", default=None)
             if remembered_info:
                 chat_history = await chat_utils.add_memory_block(remembered_info, chat_history)
 
-            reflection = await ctx.get("reflection", None)
+            reflection = await ctx.store.get("reflection", None)
             if reflection:
                 chat_history = await chat_utils.add_reflection_summary(reflection, chat_history)
 
-            chat_history = await chat_utils.add_phone_state_block(await ctx.get("phone_state"), chat_history)
-            chat_history = await chat_utils.add_ui_text_block(await ctx.get("ui_state"), chat_history)
+            chat_history = await chat_utils.add_phone_state_block(await ctx.store.get("phone_state"), chat_history)
+            chat_history = await chat_utils.add_ui_text_block(await ctx.store.get("ui_state"), chat_history)
 
-            messages_to_send = [self.system_message] + chat_history
+            limited_history = self._limit_history(chat_history)
+            messages_to_send = [self.system_message] + limited_history
             messages_to_send = [
                 chat_utils.message_copy(msg) for msg in messages_to_send
             ]
@@ -302,3 +304,23 @@ wrap your code inside this:
         except Exception as e:
             logger.error(f"Could not get an answer from LLM: {repr(e)}")
             raise e
+
+    def _limit_history(
+        self, chat_history: List[ChatMessage]
+    ) -> List[ChatMessage]:
+        if LLM_HISTORY_LIMIT <= 0:
+            return chat_history
+
+        max_messages = LLM_HISTORY_LIMIT * 2
+        if len(chat_history) <= max_messages:
+            return chat_history
+
+        preserved_head: List[ChatMessage] = []
+        if chat_history and chat_history[0].role == "user":
+            preserved_head = [chat_history[0]]
+
+        tail = chat_history[-max_messages:]
+        if preserved_head and preserved_head[0] in tail:
+            preserved_head = []
+
+        return preserved_head + tail

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/usage.py

@@ -12,7 +12,9 @@ logger = logging.getLogger("droidrun")
 SUPPORTED_PROVIDERS = [
     "Gemini",
     "GoogleGenAI",
+    "GenAI",
     "OpenAI",
+    "openai_llm",
     "Anthropic",
     "Ollama",
     "DeepSeek",
@@ -32,14 +34,14 @@ def get_usage_from_response(provider: str, chat_rsp: ChatResponse) -> UsageResul
 
     print(f"rsp: {rsp.__class__.__name__}")
 
-    if provider == "Gemini" or provider == "GoogleGenAI":
+    if provider == "Gemini" or provider == "GoogleGenAI" or provider == "GenAI":
         return UsageResult(
             request_tokens=rsp["usage_metadata"]["prompt_token_count"],
             response_tokens=rsp["usage_metadata"]["candidates_token_count"],
             total_tokens=rsp["usage_metadata"]["total_token_count"],
             requests=1,
         )
-    elif provider == "OpenAI":
+    elif provider == "OpenAI" or provider == "openai_llm":
         from openai.types import CompletionUsage as OpenAIUsage
 
         usage: OpenAIUsage = rsp.usage

{droidrun-0.3.8 → droidrun-0.3.9}/droidrun/agent/utils/executer.py

@@ -98,7 +98,7 @@ class SimpleCodeExecutor:
             str: Output from the execution, including print statements.
         """
         # Update UI elements before execution
-        self.globals['ui_state'] = await ctx.get("ui_state", None)
+        self.globals['ui_state'] = await ctx.store.get("ui_state", None)
         self.globals['step_screenshots'] = []
         self.globals['step_ui_states'] = []
         

{droidrun-0.3.8 → droidrun-0.3.9}/pyproject.toml

@@ -1,15 +1,16 @@
 [project]
 name = "droidrun"
-version = "0.3.8"
+version = "0.3.9"
 description = "A framework for controlling Android devices through LLM agents"
 authors = [{ name = "Niels Schmidt", email = "niels.schmidt@droidrun.ai" }]
 dependencies = [
     "adbutils>=2.10.2",
     # dependency of adbutils[apk]
     "apkutils==2.0.0",
-    "llama-index==0.13.6",
-    "posthog>=6.7.4",
-    "pydantic>=2.11.7",
+    "llama-index==0.14.4",
+    "llama-index-llms-google-genai>=0.6.2",
+    "posthog>=6.7.6",
+    "pydantic>=2.11.10",
     "rich>=14.1.0",
 ]
 requires-python = ">=3.11"
@@ -24,6 +25,8 @@ classifiers = [
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
     "Topic :: Software Development :: Libraries :: Python Modules",
     "Topic :: Software Development :: Testing",
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
@@ -43,7 +46,7 @@ Documentation = "https://docs.droidrun.ai/"
 [project.optional-dependencies]
 anthropic = [
     "anthropic>=0.67.0",
-    "llama-index-llms-anthropic>=0.8.6",
+    "llama-index-llms-anthropic>=0.8.6,<0.9.0",
 ]
 openai = [
     "openai>=1.99.1",