Merge branch 'browserbase:main' into a11y-include-input-attributes

Author: seanjh

.changeset/fortunate-friendly-avocet.md

@@ -1,5 +1,15 @@
 # Stagehand Python Changelog
 
+## 0.5.5
+
+### Patch Changes
+
+[#215](https://github.com/browserbase/stagehand-python/pull/215) [`cb35254`](https://github.com/browserbase/stagehand-python/commit/cb35254) Thanks @derekmeegan! - Fix ability to pass raw JSON to Extract schema
+[#225](https://github.com/browserbase/stagehand-python/pull/225) [`b23e005`](https://github.com/browserbase/stagehand-python/commit/b23e005) Thanks @derekmeegan! - add local cua example, remove root model from types
+[#218](https://github.com/browserbase/stagehand-python/pull/218) [`1a919ad`](https://github.com/browserbase/stagehand-python/commit/1a919ad) Thanks @derekmeegan! - Pass api_timeout param to Stagehand API correctly
+[#223](https://github.com/browserbase/stagehand-python/pull/223) [`de7d883`](https://github.com/browserbase/stagehand-python/commit/de7d883) Thanks @derekmeegan! - Fix search, navigate, go back, and go forward for gemini cua agent
+[#221](https://github.com/browserbase/stagehand-python/pull/221) [`da570a1`](https://github.com/browserbase/stagehand-python/commit/da570a1) Thanks @miguelg719! - Add support for Haiku 4.5 CUA
+
 ## 0.5.4
 
 ### Patch Changes

.changeset/smiling-snobbish-cockle.md

@@ -36,13 +36,11 @@ async def main():
     # Build a unified configuration object for Stagehand
     config = StagehandConfig(
         env="BROWSERBASE",
-        # env="LOCAL",
         api_key=os.getenv("BROWSERBASE_API_KEY"),
         project_id=os.getenv("BROWSERBASE_PROJECT_ID"),
-        model_name="gpt-4o",
-        self_heal=True,
         system_prompt="You are a browser automation assistant that helps users navigate websites effectively.",
         model_client_options={"apiKey": os.getenv("MODEL_API_KEY")},
+        self_heal=True,
         verbose=2,
     )
 
@@ -51,12 +49,11 @@ async def main():
 
     # Initialize - this creates a new session automatically.
     console.print("\n🚀 [info]Initializing Stagehand...[/]")
-    await stagehand.init()
-    if stagehand.env == "BROWSERBASE":    
-        console.print(f"\n[yellow]Created new session:[/] {stagehand.session_id}")
-        console.print(
-            f"🌐 [white]View your live browser:[/] [url]https://www.browserbase.com/sessions/{stagehand.session_id}[/]"
-        )
+    await stagehand.init() 
+    console.print(f"\n[yellow]Created new session:[/] {stagehand.session_id}")
+    console.print(
+        f"🌐 [white]View your live browser:[/] [url]https://www.browserbase.com/sessions/{stagehand.session_id}[/]"
+    )
 
     console.print("\n▶️ [highlight] Navigating[/] to Google")
     await stagehand.page.goto("https://google.com/")

CHANGELOG.md

@@ -0,0 +1,103 @@
+import asyncio
+import logging
+import os
+
+from dotenv import load_dotenv
+from rich.console import Console
+from rich.panel import Panel
+from rich.theme import Theme
+
+from stagehand import Stagehand, StagehandConfig, configure_logging
+
+# Create a custom theme for consistent styling
+custom_theme = Theme(
+    {
+        "info": "cyan",
+        "success": "green",
+        "warning": "yellow",
+        "error": "red bold",
+        "highlight": "magenta",
+        "url": "blue underline",
+    }
+)
+
+# Create a Rich console instance with our theme
+console = Console(theme=custom_theme)
+
+load_dotenv()
+
+# Configure logging with the utility function
+configure_logging(
+    level=logging.INFO,  # Set to INFO for regular logs, DEBUG for detailed
+    quiet_dependencies=True,  # Reduce noise from dependencies
+)
+
+async def main():
+    # Build a unified configuration object for Stagehand
+    config = StagehandConfig(
+        env="LOCAL",
+        system_prompt="You are a browser automation assistant that helps users navigate websites effectively.",
+        model_client_options={"apiKey": os.getenv("MODEL_API_KEY")},
+        self_heal=True,
+        verbose=2,
+    )
+
+    # Create a Stagehand client using the configuration object.
+    stagehand = Stagehand(config)
+
+    # Initialize - this creates a new session automatically.
+    console.print("\n🚀 [info]Initializing Stagehand...[/]")
+    await stagehand.init()
+
+    console.print("\n▶️ [highlight] Navigating[/] to Google")
+    await stagehand.page.goto("https://google.com/")
+    console.print("✅ [success]Navigated to Google[/]")
+    
+    console.print("\n▶️ [highlight] Using Agent to perform a task[/]: playing a game of 2048")
+    agent = stagehand.agent(
+        model="gemini-2.5-computer-use-preview-10-2025",
+        instructions="You are a helpful web navigation assistant that helps users find information. You are currently on the following page: google.com. Do not ask follow up questions, the user will trust your judgement.",
+        options={"apiKey": os.getenv("GEMINI_API_KEY")}
+    )
+    agent_result = await agent.execute(
+        instruction="Play a game of 2048",
+        max_steps=20,
+        auto_screenshot=True,
+    )
+
+    console.print(agent_result)
+
+    console.print("📊 [info]Agent execution result:[/]")
+    console.print(f"🎯 Completed: [bold]{'Yes' if agent_result.completed else 'No'}[/]")
+    if agent_result.message:
+        console.print(f"💬 Message: [italic]{agent_result.message}[/]")
+    
+    if agent_result.actions:
+        console.print(f"🔄 Actions performed: [bold]{len(agent_result.actions)}[/]")
+        for i, action in enumerate(agent_result.actions):
+            action_type = action.type
+
+            console.print(f"  Action {i+1}: {action_type if action_type else 'Unknown'}")
+    
+    # For debugging, you can also print the full JSON
+    console.print("[dim]Full response JSON:[/]")
+    console.print_json(f"{agent_result.model_dump_json()}")
+
+    # Close the session
+    console.print("\n⏹️  [warning]Closing session...[/]")
+    await stagehand.close()
+    console.print("✅ [success]Session closed successfully![/]")
+    console.rule("[bold]End of Example[/]")
+
+
+if __name__ == "__main__":
+    # Add a fancy header
+    console.print(
+        "\n",
+        Panel(
+            "[light_gray]Stagehand 🤘 Agent Example[/]",
+            border_style="green",
+            padding=(1, 10),
+        ),
+    )
+    asyncio.run(main()) 

examples/agent_example.py

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "stagehand"
-version = "0.5.4"
+version = "0.5.5"
 description = "Python SDK for Stagehand"
 readme = "README.md"
 classifiers = [ "Programming Language :: Python :: 3", "License :: OSI Approved :: MIT License", "Operating System :: OS Independent",]

examples/agent_example_local.py

@@ -17,17 +17,19 @@
 from .openai_cua import OpenAICUAClient
 
 MODEL_TO_CLIENT_CLASS_MAP: dict[str, type[AgentClient]] = {
-    "computer-use-preview-03-11": OpenAICUAClient,
+    "computer-use-preview-2025-03-11": OpenAICUAClient,
     "claude-3-5-sonnet-latest": AnthropicCUAClient,
     "claude-3-7-sonnet-latest": AnthropicCUAClient,
+    "claude-haiku-4-5-20251001": AnthropicCUAClient,
     "claude-sonnet-4-20250514": AnthropicCUAClient,
     "claude-sonnet-4-5-20250929": AnthropicCUAClient,
     "gemini-2.5-computer-use-preview-10-2025": GoogleCUAClient,
 }
 MODEL_TO_PROVIDER_MAP: dict[str, AgentProvider] = {
-    "computer-use-preview-03-11": AgentProvider.OPENAI,
+    "computer-use-preview-2025-03-11": AgentProvider.OPENAI,
     "claude-3-5-sonnet-20240620": AgentProvider.ANTHROPIC,
     "claude-3-7-sonnet-20250219": AgentProvider.ANTHROPIC,
+    "claude-haiku-4-5-20251001": AgentProvider.ANTHROPIC,
     "claude-sonnet-4-20250514": AgentProvider.ANTHROPIC,
     "claude-sonnet-4-5-20250929": AgentProvider.ANTHROPIC,
     "gemini-2.5-computer-use-preview-10-2025": AgentProvider.GOOGLE,
@@ -168,13 +170,10 @@ async def execute(
                 f"Agent execution finished. Success: {agent_result.completed}. Message: {agent_result.message}",
                 category="agent",
             )
-            # To clean up pydantic model output
-            actions_repr = [action.root for action in agent_result.actions]
             self.logger.debug(
-                f"Agent actions: {actions_repr}",
+                f"Agent actions: {agent_result.actions}",
                 category="agent",
             )
-            agent_result.actions = actions_repr
             return agent_result
         else:
             agent_config_payload = self.config.model_dump(