hud-python 0.4.57__py3-none-any.whl → 0.4.58__py3-none-any.whl

hud/cli/__init__.py

@@ -253,10 +253,23 @@ def debug(
         else:
             # Assume it's an image name
             image = first_param
-            from .utils.docker import build_run_command
+            from .utils.docker import create_docker_run_command
+
+            # For image mode, check if there's a .env file in current directory
+            # and use it if available (similar to hud dev behavior)
+            cwd = Path.cwd()
+            if (cwd / ".env").exists():
+                # Use create_docker_run_command to load .env from current directory
+                command = create_docker_run_command(
+                    image,
+                    docker_args=docker_args,
+                    env_dir=cwd,  # Load .env from current directory
+                )
+            else:
+                # No .env file, use basic command without env loading
+                from .utils.docker import build_run_command
 
-            # Image-only mode: do not auto-inject local .env
-            command = build_run_command(image, docker_args)
+                command = build_run_command(image, docker_args)
     else:
         console.print(
             "[red]Error: Must specify a directory, Docker image, --config, or --cursor[/red]"
@@ -741,14 +754,14 @@ def remove(
 
 @app.command()
 def init(
-    name: str = typer.Argument(None, help="Environment name (default: current directory name)"),
+    name: str = typer.Argument(None, help="Environment name (default: chosen preset name)"),
     preset: str | None = typer.Option(
         None,
         "--preset",
         "-p",
         help="Preset to use: blank, deep-research, browser, rubrics. If omitted, you'll choose interactively.",  # noqa: E501
     ),
-    directory: str = typer.Option(".", "--dir", "-d", help="Target directory"),
+    directory: str = typer.Option(".", "--dir", "-d", help="Parent directory for the environment"),
     force: bool = typer.Option(False, "--force", "-f", help="Overwrite existing files"),
 ) -> None:
     """🚀 Initialize a new HUD environment with minimal boilerplate.
@@ -760,8 +773,8 @@ def init(
     - Required setup/evaluate tools
 
     Examples:
-        hud init                    # Use current directory name
-        hud init my-env             # Create in ./my-env/
+        hud init                    # Choose preset interactively, create ./preset-name/
+        hud init my-env             # Create new directory ./my-env/
         hud init my-env --dir /tmp  # Create in /tmp/my-env/
     """
     create_environment(name, directory, force, preset)

hud/cli/dev.py

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import asyncio
+import contextlib
 import importlib
 import importlib.util
 import logging
@@ -13,6 +14,8 @@ import threading
 from pathlib import Path
 from typing import Any
 
+import typer
+
 from hud.utils.hud_console import HUDConsole
 
 hud_console = HUDConsole()
@@ -26,6 +29,7 @@ def show_dev_server_info(
     interactive: bool,
     env_dir: Path | None = None,
     new: bool = False,
+    docker_mode: bool = False,
 ) -> str:
     """Show consistent server info for both Python and Docker modes.
 
@@ -54,7 +58,15 @@ def show_dev_server_info(
     if transport == "http":
         hud_console.section_title("Quick Links")
         hud_console.info(f"{hud_console.sym.ITEM} Docs: http://localhost:{port}/docs")
-        hud_console.info(f"{hud_console.sym.ITEM} Cursor: {cursor_deeplink}")
+        hud_console.info(f"{hud_console.sym.ITEM} Cursor:")
+        # Display the Cursor link on its own line to prevent wrapping
+        hud_console.link(cursor_deeplink)
+
+        # Show eval endpoint if in Docker mode
+        if docker_mode:
+            hud_console.info(
+                f"{hud_console.sym.ITEM} Eval API: http://localhost:{port}/eval (POST)"
+            )
 
         # Check for VNC (browser environment)
         if env_dir and (env_dir / "environment" / "server.py").exists():
@@ -510,6 +522,9 @@ def run_docker_dev_server(
     new: bool = False,
 ) -> None:
     """Run MCP server in Docker with volume mounts, expose via local HTTP proxy."""
+    import atexit
+    import signal
+
     import typer
     import yaml
 
@@ -522,6 +537,69 @@ def run_docker_dev_server(
 
     cwd = Path.cwd()
 
+    # Container name will be set later and used for cleanup
+    container_name: str | None = None
+    cleanup_done = False
+
+    def cleanup_container() -> None:
+        """Clean up Docker container on exit."""
+        nonlocal cleanup_done
+        if cleanup_done or not container_name:
+            return
+
+        cleanup_done = True
+        hud_console.debug(f"Cleaning up container: {container_name}")
+
+        # Check if container is still running
+        try:
+            result = subprocess.run(  # noqa: S603
+                ["docker", "ps", "-q", "-f", f"name={container_name}"],  # noqa: S607
+                stdout=subprocess.PIPE,
+                stderr=subprocess.DEVNULL,
+                text=True,
+                timeout=5,
+            )
+            if not result.stdout.strip():
+                # Container is not running, just try to remove it
+                subprocess.run(  # noqa: S603
+                    ["docker", "rm", "-f", container_name],  # noqa: S607
+                    stdout=subprocess.DEVNULL,
+                    stderr=subprocess.DEVNULL,
+                    timeout=5,
+                )
+                return
+        except Exception:  # noqa: S110
+            pass
+
+        try:
+            # First try to stop gracefully
+            subprocess.run(  # noqa: S603
+                ["docker", "stop", container_name],  # noqa: S607
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL,
+                timeout=10,
+            )
+            hud_console.debug(f"Container {container_name} stopped successfully")
+        except subprocess.TimeoutExpired:
+            # Force kill if stop times out
+            hud_console.debug(f"Container {container_name} stop timeout, forcing kill")
+            with contextlib.suppress(Exception):
+                subprocess.run(  # noqa: S603
+                    ["docker", "kill", container_name],  # noqa: S607
+                    stdout=subprocess.DEVNULL,
+                    stderr=subprocess.DEVNULL,
+                    timeout=5,
+                )
+
+    # Set up signal handlers for cleanup
+    def signal_handler(signum: int, frame: Any) -> None:
+        cleanup_container()
+        sys.exit(0)
+
+    signal.signal(signal.SIGTERM, signal_handler)
+    if sys.platform != "win32":
+        signal.signal(signal.SIGHUP, signal_handler)
+
     # Find environment directory (current or parent with hud.lock.yaml)
     env_dir = cwd
     lock_path = env_dir / "hud.lock.yaml"
@@ -562,10 +640,14 @@ def run_docker_dev_server(
     base_name = image_name.replace(":", "-").replace("/", "-")
     container_name = f"{base_name}-dev-{pid}"
 
+    # Register cleanup function with atexit
+    atexit.register(cleanup_container)
+
     # Build docker run command with volume mounts and folder-mode envs
     from .utils.docker import create_docker_run_command
 
     base_args = [
+        "--rm",  # Automatically remove container when it stops
         "--name",
         container_name,
         "-v",
@@ -643,6 +725,7 @@ def run_docker_dev_server(
             interactive=interactive,
             env_dir=env_dir,
             new=new,
+            docker_mode=True,
         )
         hud_console.dim_info(
             "",
@@ -679,6 +762,11 @@ def run_docker_dev_server(
             os.environ["_HUD_DEV_DOCKER_CONTAINER"] = container_name
             hud_console.debug(f"Docker container: {container_name}")
 
+        # Store the docker mcp_config for the eval endpoint
+        import json
+
+        os.environ["_HUD_DEV_DOCKER_MCP_CONFIG"] = json.dumps(mcp_config)
+
         # Create FastMCP proxy using the ProxyClient
         fastmcp_proxy = FastMCP.as_proxy(proxy_client)
 
@@ -713,7 +801,15 @@ def run_docker_dev_server(
         asyncio.run(run_proxy())
     except KeyboardInterrupt:
         hud_console.info("\n\nStopping...")
+        cleanup_container()
         raise typer.Exit(0) from None
+    except Exception:
+        # Ensure cleanup happens on any exception
+        cleanup_container()
+        raise
+    finally:
+        # Final cleanup attempt
+        cleanup_container()
 
 
 def run_mcp_dev_server(
@@ -732,6 +828,20 @@ def run_mcp_dev_server(
     docker_args = docker_args or []
     cwd = Path.cwd()
 
+    # Find an available port if not using stdio transport
+    if not stdio:
+        from hud.cli.utils.logging import find_free_port
+
+        actual_port = find_free_port(port)
+        if actual_port is None:
+            hud_console.error(f"No available ports found starting from {port}")
+            raise typer.Exit(1)
+
+        if actual_port != port:
+            hud_console.info(f"Port {port} is in use, using port {actual_port} instead")
+
+        port = actual_port
+
     # Auto-detect Docker mode if Dockerfile present and no module specified
     if not docker and module is None and should_use_docker_mode(cwd):
         hud_console.note("Detected Dockerfile - using Docker mode with volume mounts")

hud/cli/eval.py

@@ -53,7 +53,7 @@ def get_available_models() -> list[dict[str, str | None]]:
     try:
         from hud.cli.rl import rl_api
 
-        hud_console.info("Fetching your models from https://hud.so/models")
+        hud_console.info("Fetching your models from https://hud.ai/models")
         models = rl_api.list_models()
 
         # Filter for ready models only and sort by recency
@@ -771,7 +771,7 @@ def eval_command(
     # Check for HUD_API_KEY if using HUD services
     if not settings.api_key:
         hud_console.warning("HUD_API_KEY not set. Some features may be limited.")
-        hud_console.info("Get your API key at: https://hud.so")
+        hud_console.info("Get your API key at: https://hud.ai")
         hud_console.info("Set it in your environment or run: hud set HUD_API_KEY=your-key-here")
 
     # Parse allowed tools

hud/cli/flows/dev.py

@@ -47,11 +47,11 @@ async def create_dynamic_trace(
 
     try:
         resp = await make_request("POST", url=url, json=payload, api_key=api_key)
-        # New API returns an id; construct the URL as https://hud.so/trace/{id}
+        # New API returns an id; construct the URL as https://hud.ai/trace/{id}
         trace_id = resp.get("id")
 
         if isinstance(trace_id, str) and trace_id:
-            return trace_id, f"https://hud.so/trace/{trace_id}"
+            return trace_id, f"https://hud.ai/trace/{trace_id}"
         return None, None
     except Exception as e:
         # Do not interrupt dev flow
@@ -114,7 +114,9 @@ def show_dev_ui(
     label = "Base image" if is_docker else "Server"
     hud_console.info("")
     hud_console.info(f"{hud_console.sym.ITEM} {label}: {server_name}")
-    hud_console.info(f"{hud_console.sym.ITEM} Cursor: {cursor_deeplink}")
+    hud_console.info(f"{hud_console.sym.ITEM} Cursor:")
+    # Display the Cursor link on its own line to prevent wrapping
+    hud_console.link(cursor_deeplink)
     hud_console.info("")
     hud_console.info(f"{hud_console.sym.SUCCESS} Hot-reload enabled")
     if is_docker:

hud/cli/init.py

@@ -182,17 +182,17 @@ def create_environment(
 
     hud_console = HUDConsole()
 
-    # Determine environment name/target directory
-    if name is None:
-        current_dir = Path.cwd()
-        name = current_dir.name
-        target_dir = current_dir
-        hud_console.info(f"Using current directory name: {name}")
-    else:
-        target_dir = Path(directory) / name
-
     # Choose preset
     preset_normalized = (preset or "").strip().lower() if preset else _prompt_for_preset()
+
+    # If no name is provided, use the preset name as the environment name
+    if name is None:
+        name = preset_normalized
+        hud_console.info(f"Using preset name as environment name: {name}")
+
+    # Always create a new directory based on the name
+    target_dir = Path.cwd() / name if directory == "." else Path(directory) / name
+
     if preset_normalized not in PRESET_MAP:
         hud_console.warning(
             f"Unknown preset '{preset_normalized}', defaulting to 'blank' "
@@ -263,14 +263,10 @@ def create_environment(
         hud_console.status_item(entry, "added")
 
     hud_console.section_title("Next steps")
-    if target_dir == Path.cwd():
-        hud_console.info("1. Start development server (with MCP inspector):")
-        hud_console.command_example("hud dev --inspector")
-    else:
-        hud_console.info("1. Enter the directory:")
-        hud_console.command_example(f"cd {target_dir}")
-        hud_console.info("\n2. Start development server (with MCP inspector):")
-        hud_console.command_example("hud dev --inspector")
-
+    # Since we now almost always create a new directory, show cd command
+    hud_console.info("1. Enter the directory:")
+    hud_console.command_example(f"cd {target_dir.name}")
+    hud_console.info("\n2. Start development server (with MCP inspector):")
+    hud_console.command_example("hud dev --inspector")
     hud_console.info("\n3. Review the README in this preset for specific instructions.")
     hud_console.info("\n4. Customize as needed.")

hud/cli/push.py

@@ -152,7 +152,7 @@ def push_environment(
         hud_console.error("No HUD API key found")
         hud_console.warning("A HUD API key is required to push environments.")
         hud_console.info("\nTo get started:")
-        hud_console.info("1. Get your API key at: https://hud.so/settings")
+        hud_console.info("1. Get your API key at: https://hud.ai/settings")
         hud_console.info("Set it in your environment or run: hud set HUD_API_KEY=your-key-here")
         hud_console.command_example("hud push", "Try again")
         hud_console.info("")
@@ -440,7 +440,7 @@ def push_environment(
         elif response.status_code == 401:
             hud_console.error("Authentication failed")
             hud_console.info("Check your HUD_API_KEY is valid")
-            hud_console.info("Get a new key at: https://hud.so/settings")
+            hud_console.info("Get a new key at: https://hud.ai/settings")
             hud_console.info("Set it in your environment or run: hud set HUD_API_KEY=your-key-here")
         elif response.status_code == 403:
             hud_console.error("Permission denied")

hud/cli/rl/__init__.py

@@ -25,7 +25,7 @@ def rl_command(
     ),
     model: str | None = typer.Argument(
         None,
-        help="Model to train from https://hud.so/models (default: interactive selection)",
+        help="Model to train from https://hud.ai/models (default: interactive selection)",
     ),
     config_file: Path | None = typer.Option(  # noqa: B008
         None,

hud/cli/rl/celebrate.py

@@ -133,7 +133,7 @@ def show_confetti(console: Console, seconds: float = 2.5) -> None:
     """
     # Show celebratory message first
     console.print(
-        "[bold green]🎉 Starting training! See your model on https://hud.so/models[/bold green]"
+        "[bold green]🎉 Starting training! See your model on https://hud.ai/models[/bold green]"
     )
     time.sleep(0.3)  # Brief pause to see the message
 

hud/cli/rl/remote_runner.py

@@ -55,7 +55,7 @@ def ensure_vllm_deployed(
     hud_console.info("Waiting for vLLM server to be ready...")
     start_time = time.time()
     with hud_console.progress() as progress:
-        progress.update("Checking deployment status (see live status on https://hud.so/models)")
+        progress.update("Checking deployment status (see live status on https://hud.ai/models)")
         while True:
             if time.time() - start_time > timeout:
                 hud_console.error("Timeout waiting for vLLM deployment")
@@ -139,7 +139,7 @@ def run_remote_training(
     hud_console.section_title("Model Selection")
 
     # Fetch existing models
-    hud_console.info("Fetching your models from https://hud.so/models")
+    hud_console.info("Fetching your models from https://hud.ai/models")
 
     try:
         models = rl_api.list_models()
@@ -312,7 +312,7 @@ def run_remote_training(
         # gpu_table.add_column("Price/hr", style="yellow")
 
         # for gpu, info in GPU_PRICING.items():
-        #     gpu_table.add_row(gpu, info["memory"], "see pricing on hud.so")
+        #     gpu_table.add_row(gpu, info["memory"], "see pricing on hud.ai")
 
         # console.print(gpu_table)
 

hud/clients/base.py

@@ -140,7 +140,7 @@ class BaseHUDClient(AgentMCPClient):
                     raise HudAuthenticationError(
                         f'Sending authorization "{headers.get("Authorization", "")}", which may'
                         " be incomplete. Ensure HUD_API_KEY environment variable is set or send it"
-                        " as a header. You can get an API key at https://hud.so"
+                        " as a header. You can get an API key at https://hud.ai"
                     )
             # Subclasses implement connection
             await self._connect(self._mcp_config)

hud/clients/fastmcp.py

@@ -95,7 +95,7 @@ class FastMCPHUDClient(BaseHUDClient):
                             raise RuntimeError(
                                 "Authentication failed for HUD API. "
                                 "Please ensure your HUD_API_KEY environment variable is set correctly."  # noqa: E501
-                                "You can get an API key at https://hud.so"
+                                "You can get an API key at https://hud.ai"
                             ) from e
                     # Generic 401 error
                     raise RuntimeError(

hud/otel/config.py

@@ -113,7 +113,7 @@ def configure_telemetry(
         # Error if no exporters are configured
         raise ValueError(
             "No telemetry backend configured. Either:\n"
-            "1. Set HUD_API_KEY environment variable for HUD telemetry (https://hud.so)\n"
+            "1. Set HUD_API_KEY environment variable for HUD telemetry (https://hud.ai)\n"
             "2. Use enable_otlp=True with configure_telemetry() for alternative backends (e.g., Jaeger)\n"  # noqa: E501
         )
     elif not settings.telemetry_enabled:

hud/otel/context.py

@@ -408,7 +408,7 @@ def _print_trace_url(task_run_id: str) -> None:
     if not (settings.telemetry_enabled and settings.api_key):
         return
 
-    url = f"https://hud.so/trace/{task_run_id}"
+    url = f"https://hud.ai/trace/{task_run_id}"
     header = "🚀 See your agent live at:"
 
     # ANSI color codes
@@ -447,7 +447,7 @@ def _print_trace_complete_url(task_run_id: str, error_occurred: bool = False) ->
     if not (settings.telemetry_enabled and settings.api_key):
         return
 
-    url = f"https://hud.so/trace/{task_run_id}"
+    url = f"https://hud.ai/trace/{task_run_id}"
 
     # ANSI color codes
     GREEN = "\033[92m"

hud/server/server.py

@@ -16,7 +16,9 @@ from fastmcp.server.server import FastMCP, Transport
 from starlette.requests import Request
 from starlette.responses import JSONResponse, Response
 
+from hud.cli.eval import run_full_dataset
 from hud.server.low_level import LowLevelServerWithInit
+from hud.types import Task
 
 if TYPE_CHECKING:
     from collections.abc import AsyncGenerator, Callable
@@ -486,6 +488,84 @@ class MCPServer(FastMCP):
             self._prompt_manager._prompts[new_key] = prompt
         # await self.import_server(hidden_router, prefix=None, **kwargs)
 
+    def _get_docker_logs(
+        self,
+        tail: int = 100,
+        since: str | None = None,
+        until: str | None = None,
+        timestamps: bool = False,
+    ) -> dict[str, Any]:
+        """Helper function to get Docker container logs.
+
+        Args:
+            tail: Number of lines to show from the end of the logs
+            since: Show logs since timestamp or relative time
+            until: Show logs before a timestamp or relative time
+            timestamps: Show timestamps in log output
+
+        Returns:
+            Dictionary with logs data or error information
+        """
+        import subprocess
+
+        container_name = os.environ.get("_HUD_DEV_DOCKER_CONTAINER")
+        if not container_name:
+            return {"items": [], "container_name": None, "error": "No container name found"}
+
+        # Build docker logs command
+        cmd = ["docker", "logs", "--tail", str(tail)]
+
+        if since:
+            cmd.extend(["--since", since])
+        if until:
+            cmd.extend(["--until", until])
+        if timestamps:
+            cmd.append("--timestamps")
+
+        cmd.append(container_name)
+
+        try:
+            # Run docker logs to get output
+            result = subprocess.run(  # noqa: S603
+                cmd,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.STDOUT,
+                text=True,
+                encoding="utf-8",
+                errors="replace",
+                timeout=5,
+            )
+
+            # Parse logs into items
+            items = []
+            lines = result.stdout.strip().split("\n") if result.stdout else []
+
+            for i, line in enumerate(lines):
+                if line.strip():
+                    items.append(
+                        {
+                            "id": i,
+                            "stream": "mixed",
+                            "log": line,
+                            "container_name": container_name,
+                        }
+                    )
+
+            return {
+                "items": items,
+                "container_name": container_name,
+                "total_lines": len(items),
+            }
+
+        except subprocess.TimeoutExpired:
+            return {"error": "Docker logs timeout", "container_name": container_name, "items": []}
+        except Exception as e:
+            return {
+                "error": f"Failed to get logs: {e!s}",
+                "container_name": container_name,
+                "items": [],
+            }
+
     def _register_hud_helpers(self) -> None:
         """Register development helper endpoints.
 
@@ -494,6 +574,7 @@ class MCPServer(FastMCP):
         - POST /api/tools/{name} - REST wrappers for MCP tools
         - GET /openapi.json - OpenAPI spec for REST endpoints
         - GET /logs - Development log endpoint (when provided by dev runtime)
+        - hud-logs tool - MCP tool for fetching logs (when in Docker mode)
         """
 
         # Register REST wrapper for each tool
@@ -544,7 +625,7 @@ class MCPServer(FastMCP):
             endpoint = create_tool_endpoint(tool_key)
             self.custom_route(f"/api/tools/{tool_key}", methods=["POST"])(endpoint)
 
-        # Development log endpoint - only if dev runtime set a provider
+        # Development endpoints - only if dev runtime set a provider
         provider = os.environ.get("_HUD_DEV_LOGS_PROVIDER")
         if provider == "enabled":
 
@@ -556,50 +637,182 @@ class MCPServer(FastMCP):
                   - limit: max number of lines to return (default 100)
                   - tail: number of lines from end to return (default 100)
                 """
-                import subprocess
-
-                # Get container name from environment
-                container_name = os.environ.get("_HUD_DEV_DOCKER_CONTAINER")
-                if not container_name:
-                    return JSONResponse({"items": [], "next": None})
-
                 # Get query params
                 params = request.query_params
-                tail = params.get("tail", "100")
+                tail = int(params.get("tail", "100"))
+
+                # Use helper function to get logs
+                result = self._get_docker_logs(tail=tail)
+
+                # Add 'next' field for compatibility with existing API
+                if "error" in result:
+                    return JSONResponse(result, status_code=500)
+                else:
+                    items = result.get("items", [])
+                    return JSONResponse(
+                        {
+                            "items": items,
+                            "next": len(items) - 1 if items else None,
+                        }
+                    )
+
+            # Import existing types from the codebase
+            from pydantic import BaseModel
+
+            from hud.types import AgentType
+
+            class EvalRequest(BaseModel):
+                """Request model for /eval endpoint."""
+
+                tasks: list[dict[str, Any]] = []
+                agent: str = "claude"
+                model: str | None = None
+                max_steps: int = 10
+                verbose: bool = False
+                group_size: int = 1
+                name: str | None = None
+
+            @self.custom_route("/eval", methods=["POST"])
+            async def run_eval(request: Request) -> Response:
+                """Run evaluation on tasks using the current Docker environment."""
+                import asyncio
+                import json
 
                 try:
-                    # Run docker logs to get recent output
-                    result = subprocess.run(  # noqa: S603, ASYNC221
-                        ["docker", "logs", "--tail", tail, container_name],  # noqa: S607
-                        stdout=subprocess.PIPE,
-                        stderr=subprocess.STDOUT,
-                        text=True,
-                        encoding="utf-8",
-                        errors="replace",
-                        timeout=5,
-                    )
+                    body = await request.body()
+                    data = json.loads(body)
+
+                    # Validate request using Pydantic model
+                    try:
+                        eval_request = EvalRequest(**data)
+                    except Exception as e:
+                        return JSONResponse({"error": f"Invalid request: {e!s}"}, status_code=400)
+
+                    # Get the Docker MCP config from environment
+                    docker_mcp_config = os.environ.get("_HUD_DEV_DOCKER_MCP_CONFIG")
+                    if not docker_mcp_config:
+                        return JSONResponse(
+                            {"error": "Docker MCP config not available"}, status_code=500
+                        )
 
-                    # Parse logs into items
-                    items = []
-                    lines = result.stdout.strip().split("\n") if result.stdout else []
-
-                    for i, line in enumerate(lines):
-                        if line.strip():
-                            items.append(
-                                {
-                                    "id": i,
-                                    "stream": "mixed",
-                                    "log": line,
-                                    "container_name": container_name,
-                                }
+                    docker_config = json.loads(docker_mcp_config)
+
+                    # Simplify Docker config for evaluation
+                    if "docker" in docker_config and "args" in docker_config["docker"]:
+                        original_args = docker_config["docker"]["args"]
+                        filtered_args = []
+                        i = 0
+
+                        while i < len(original_args):
+                            arg = original_args[i]
+
+                            # Skip volume mounts and their values
+                            if arg in ["-v", "--volume"]:
+                                i += 2  # Skip the flag and its value
+                                continue
+
+                            # Skip combined volume mount args
+                            if arg.startswith(("-v", "--volume=")):
+                                i += 1
+                                continue
+
+                            # Skip explicit container name to avoid collisions
+                            if arg == "--name" and i + 1 < len(original_args):
+                                i += 2  # Skip the --name and its value
+                                continue
+
+                            # Skip dev-specific environment variables
+                            if arg == "-e" and i + 1 < len(original_args):
+                                next_arg = original_args[i + 1]
+                                if next_arg in [
+                                    "PYTHONPATH=/app",
+                                    "HUD_DEV=1",
+                                    "PYTHONUNBUFFERED=1",
+                                ]:
+                                    i += 2  # Skip the -e and its value
+                                    continue
+
+                            filtered_args.append(arg)
+                            i += 1
+
+                        # Update the docker args with filtered version
+                        docker_config["docker"]["args"] = filtered_args
+
+                    try:
+                        agent_type = AgentType(eval_request.agent.lower())
+                    except ValueError:
+                        valid_agents = [
+                            a.value for a in AgentType if a != AgentType.INTEGRATION_TEST
+                        ]
+                        return JSONResponse(
+                            {
+                                "error": f"Invalid agent type: {eval_request.agent}",
+                                "valid_agents": valid_agents,
+                            },
+                            status_code=400,
+                        )
+
+                    # Add MCP config to each task and validate basic structure
+                    tasks = []
+                    for task_data in eval_request.tasks:
+                        task_data["mcp_config"] = docker_config
+                        tasks.append(Task.model_validate(task_data).model_dump())
+
+                    # Save tasks to temporary file
+                    import tempfile
+
+                    with tempfile.NamedTemporaryFile(
+                        mode="w", prefix="hud-eval-", suffix=".json", delete=False
+                    ) as f:
+                        json.dump(tasks, f)
+                        task_file = f.name
+
+                    # Fire and forget - launch evaluation in background
+                    async def run_eval_background() -> None:
+                        try:
+                            await run_full_dataset(
+                                task_file,
+                                agent_type=agent_type,
+                                model=eval_request.model,
+                                max_steps=eval_request.max_steps,
+                                verbose=eval_request.verbose,
+                                group_size=eval_request.group_size,
                             )
+                        except Exception as e:
+                            raise e
+                        finally:
+                            # Clean up temp file
+                            import os
+
+                            if os.path.exists(task_file):
+                                os.unlink(task_file)
+
+                    # Start the evaluation in the background (fire and forget)
+                    asyncio.create_task(run_eval_background())  # noqa: RUF006
+
+                    # Return immediately
+                    response_data = {
+                        "status": "started",
+                        "message": f"Evaluation launched with {len(tasks)} task(s)",
+                        "agent": eval_request.agent,
+                        "model": eval_request.model,
+                        "max_steps": eval_request.max_steps,
+                        "verbose": eval_request.verbose,
+                    }
+
+                    # Include group_size if > 1
+                    if eval_request.group_size > 1:
+                        response_data["group_size"] = eval_request.group_size
+                        response_data["total_episodes"] = len(tasks) * eval_request.group_size
 
-                    return JSONResponse({"items": items, "next": len(items) - 1 if items else None})
+                    return JSONResponse(response_data)
 
-                except subprocess.TimeoutExpired:
-                    return JSONResponse({"error": "Docker logs timeout"}, status_code=500)
+                except json.JSONDecodeError:
+                    return JSONResponse({"error": "Invalid JSON in request body"}, status_code=400)
                 except Exception as e:
-                    return JSONResponse({"error": f"Failed to get logs: {e!s}"}, status_code=500)
+                    return JSONResponse(
+                        {"error": f"Failed to run evaluation: {e!s}"}, status_code=500
+                    )
 
         @self.custom_route("/openapi.json", methods=["GET"])
         async def openapi_spec(request: Request) -> Response:
@@ -656,6 +869,40 @@ class MCPServer(FastMCP):
 
             return JSONResponse(spec)
 
+        # Register hud-logs tool when in Docker dev mode
+        container_name = os.environ.get("_HUD_DEV_DOCKER_CONTAINER")
+        if container_name:
+
+            @self.tool("hud-logs")
+            async def get_docker_logs(
+                tail: int = 100,
+                since: str | None = None,
+                until: str | None = None,
+                timestamps: bool = False,
+            ) -> dict[str, Any]:
+                """Get logs from the Docker container running the HUD environment.
+
+                Args:
+                    tail: Number of lines to show from the end of the logs (default: 100)
+                    since: Show logs since timestamp (e.g. 2013-01-02T13:23:37Z) or relative (42m)
+                    until: Show logs before timestamp (e.g. 2013-01-02T13:23:37Z) or relative (42m)
+                    timestamps: Show timestamps in log output
+
+                Returns:
+                    Dictionary with:
+                    - items: List of log entries
+                    - container_name: Name of the container
+                    - total_lines: Total number of log lines returned
+                    - error: Error message if logs could not be retrieved
+                """
+                # Use helper function to get logs
+                return self._get_docker_logs(
+                    tail=tail,
+                    since=since,
+                    until=until,
+                    timestamps=timestamps,
+                )
+
         @self.custom_route("/docs", methods=["GET"])
         async def docs_page(request: Request) -> Response:
             """Interactive documentation page."""

hud/shared/hints.py

@@ -38,7 +38,7 @@ HUD_API_KEY_MISSING = Hint(
     message="Missing or invalid HUD_API_KEY.",
     tips=[
         "Set HUD_API_KEY in your environment or run: hud set HUD_API_KEY=your-key-here",
-        "Get a key at https://hud.so",
+        "Get a key at https://hud.ai",
         "Check for whitespace or truncation",
     ],
     docs_url=None,
@@ -68,7 +68,7 @@ PRO_PLAN_REQUIRED = Hint(
     tips=[
         "Upgrade your plan to continue",
     ],
-    docs_url="https://hud.so/project/billing",
+    docs_url="https://hud.ai/project/billing",
     command_examples=None,
     code="PRO_PLAN_REQUIRED",
     context=["billing", "plan"],
@@ -80,7 +80,7 @@ CREDITS_EXHAUSTED = Hint(
     tips=[
         "Top up credits or upgrade your plan",
     ],
-    docs_url="https://hud.so/project/billing",
+    docs_url="https://hud.ai/project/billing",
     command_examples=None,
     code="CREDITS_EXHAUSTED",
     context=["billing", "credits"],

hud/telemetry/job.py

@@ -170,7 +170,7 @@ def _print_job_url(job_id: str, job_name: str) -> None:
     if not (settings.telemetry_enabled and settings.api_key):
         return
 
-    url = f"https://hud.so/jobs/{job_id}"
+    url = f"https://hud.ai/jobs/{job_id}"
     header = f"🚀 Job '{job_name}' started:"
 
     # ANSI color codes
@@ -209,7 +209,7 @@ def _print_job_complete_url(job_id: str, job_name: str, error_occurred: bool = F
     if not (settings.telemetry_enabled and settings.api_key):
         return
 
-    url = f"https://hud.so/jobs/{job_id}"
+    url = f"https://hud.ai/jobs/{job_id}"
 
     # ANSI color codes
     GREEN = "\033[92m"

hud/tools/playwright.py

@@ -225,7 +225,14 @@ class PlaywrightTool(BaseTool):
             if self._browser_context is None:
                 raise RuntimeError("Browser context failed to initialize")
 
-            self.page = await self._browser_context.new_page()
+            # Reuse existing page if available (for CDP connections), otherwise create new one
+            pages = self._browser_context.pages
+            if pages:
+                self.page = pages[0]
+                logger.info("Reusing existing browser page")
+            else:
+                self.page = await self._browser_context.new_page()
+                logger.info("Created new browser page")
             logger.info("Playwright browser launched successfully")
 
     async def navigate(

hud/types.py

@@ -230,7 +230,7 @@ class AgentResponse(BaseModel):
     tool_calls: list[MCPToolCall] = Field(default_factory=list)
     done: bool = Field(default=False)
 
-    # --- TELEMETRY [hud.so] ---
+    # --- TELEMETRY [hud.ai] ---
     # Responses
     content: str | None = Field(default=None)
     reasoning: str | None = Field(default=None)

hud/utils/tests/test_version.py

@@ -5,4 +5,4 @@ def test_import():
     """Test that the package can be imported."""
     import hud
 
-    assert hud.__version__ == "0.4.57"
+    assert hud.__version__ == "0.4.58"

hud/version.py

@@ -4,4 +4,4 @@ Version information for the HUD SDK.
 
 from __future__ import annotations
 
-__version__ = "0.4.57"
+__version__ = "0.4.58"

{hud_python-0.4.57.dist-info → hud_python-0.4.58.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hud-python
-Version: 0.4.57
+Version: 0.4.58
 Summary: SDK for the HUD platform.
 Project-URL: Homepage, https://github.com/hud-evals/hud-python
 Project-URL: Bug Tracker, https://github.com/hud-evals/hud-python/issues

{hud_python-0.4.57.dist-info → hud_python-0.4.58.dist-info}/RECORD

@@ -1,8 +1,8 @@
 hud/__init__.py,sha256=0LQ9PyuU6yZx7Fxu8YJXKC1i3TTHjg3UrInThh759QE,653
 hud/__main__.py,sha256=YR8Dq8OhINOsVfQ55PmRXXg4fEK84Rt_-rMtJ5rvhWo,145
 hud/settings.py,sha256=9SPEZcsZGqplw_LuV8RvJisDRdlsbQgxBqG8ifmjGNc,4117
-hud/types.py,sha256=jeC1npyPlno5CYNWBWH06Kxw7VMJld9YLEFBFNtjwoI,11608
-hud/version.py,sha256=5oonlS-3cemOdbkDddf3Eg6Wr5IkBtjvigjh9Y3A664,105
+hud/types.py,sha256=XA15zkH25Kt8IQAjT3xfEPew6oCKsnkERKrrKrYmYbM,11608
+hud/version.py,sha256=7lqYv2fdoclIKiUgO55ISA4NzOcYp9DBF4rdJcsoHYk,105
 hud/agents/__init__.py,sha256=UoIkljWdbq4bM0LD-mSaw6w826EqdEjOk7r6glNYwYQ,286
 hud/agents/base.py,sha256=nRRnK_kA_PZDT0fepMMTz3QDTHj0jpqvbTqdgRPKSlg,32514
 hud/agents/claude.py,sha256=Lf7p_h5H4sd2w16ZNc2ru3xuxNP7IDeFayrcdePOLSE,16030
@@ -22,32 +22,32 @@ hud/agents/tests/test_claude.py,sha256=YsI2TV87Sex2BJsLaO2l99WEL3PL3OhQDlTwUHn04
 hud/agents/tests/test_client.py,sha256=uikgh6yhjPPX2RBU4XJQMz1mNox9uXjuwsP8t93id18,13337
 hud/agents/tests/test_grounded_openai_agent.py,sha256=VK8lUvHIjWicMX00VKPE-FZyjiJqTEhb80MuRRa9fVc,5437
 hud/agents/tests/test_openai.py,sha256=hPkL_ANbkrB_HUwvd4wBONuWomCarhLJQcFRpmwSbEk,8016
-hud/cli/__init__.py,sha256=kfCsHOG94qIkWG_qipPJWEDqgNDG_qgqew4nf3pr8CE,43083
+hud/cli/__init__.py,sha256=sb0EyBBGEBKbfu2Ph3Wns42plYmA1Hejne5lLZ1hn4A,43747
 hud/cli/__main__.py,sha256=fDH7XITyuDITwSDIVwRso06aouADO0CzTHKqp5TOwJE,143
 hud/cli/analyze.py,sha256=4u5oYfJMquOjT9PzzRTYVcTZDxDi0ilNP_g532_hpOU,14716
 hud/cli/build.py,sha256=EV6PsJ08p3mZZkfeUFjZ687ithxKGlUZ66yAF9lvtaI,28780
 hud/cli/clone.py,sha256=AwVDIuhr8mHb1oT2Af2HrD25SiTdwATpE6zd93vzLgA,6099
 hud/cli/debug.py,sha256=jtFW8J5F_3rhq1Hf1_SkJ7aLS3wjnyIs_LsC8k5cnzc,14200
-hud/cli/dev.py,sha256=dGA8p5DvKFUgcYA9awXbbIfJ8OV04ew_jd0fElfJRoU,26255
-hud/cli/eval.py,sha256=1guW_pNQqFLpF1RylZltBKdXt6iFd_jBNV3Rxy36QRE,29024
+hud/cli/dev.py,sha256=Cg9BLUs4mcMR6UQw43vW68Tx548BPpzlawlC8NgR2Fk,30018
+hud/cli/eval.py,sha256=8LC8kS4RNuvgEGGf_sP7lz0LOgD8mVYOtEEQMXN59Bw,29024
 hud/cli/get.py,sha256=sksKrdzBGZa7ZuSoQkc0haj-CvOGVSSikoVXeaUd3N4,6274
-hud/cli/init.py,sha256=D1NS4PBHxNqsl9hz2VsMTj6HSB_teJgCqsyY77T5lLs,10042
+hud/cli/init.py,sha256=bkpby7vBDqRH3GUDunNA5PIkbQ-Jnpas5pEeN1ltK7k,9976
 hud/cli/list_func.py,sha256=EVi2Vc3Lb3glBNJxFx4MPnZknZ4xmuJz1OFg_dc8a_E,7177
 hud/cli/pull.py,sha256=XGEZ8n60tbzLQP_8d9h7XYmzyCW0e2-Rkr3_tLG7jvw,12449
-hud/cli/push.py,sha256=rWJIqHebvp3DchK-00L6G0olD3-klsobLutRW4PP_ts,19488
+hud/cli/push.py,sha256=y-1xYup2M396wlJ15d_N2jqthofmlWWiqCbtdYfZyGc,19488
 hud/cli/remove.py,sha256=8vGQyXDqgtjz85_vtusoIG8zurH4RHz6z8UMevQRYM4,6861
 hud/cli/flows/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-hud/cli/flows/dev.py,sha256=7-Dki2C8MbdcFcHewKhmRfkfjVdo6E4CGmItKVMaUSI,4760
+hud/cli/flows/dev.py,sha256=aLF_hzra0MIin2_2jWi3--yIZdHm5wi2o2PoQlpNm6M,4846
 hud/cli/flows/tasks.py,sha256=SAQzS53g6vBPMvuEaMRsmkYOJMwvy28RQ-miVmVnP_8,18028
-hud/cli/rl/__init__.py,sha256=pGx4WGaL-yGdogJNzgEE7BtjFtT4I9CTI_UdCm49h98,5376
-hud/cli/rl/celebrate.py,sha256=trGEJn3xebexlHwFVKPJKhRujVVV8sy7TQTJvRd2p9A,5947
+hud/cli/rl/__init__.py,sha256=Tn_dYNrpk0UlKiPx0VZjGlGWlRUYxGj5PThp5PqTWbc,5376
+hud/cli/rl/celebrate.py,sha256=vFex4RtHTgbhdsRpF4qS2vo4gd_W6LRd0ksO1EIv6J4,5947
 hud/cli/rl/config.py,sha256=A-4WWwAS68GRKx1cP_DJ-NZD_96cFNnGwx0P3pQT1ps,3271
 hud/cli/rl/display.py,sha256=hqJVGmO9csYinladhZwjF-GMvppYWngxDHajTyIJ_gM,5214
 hud/cli/rl/gpu.py,sha256=peXS-NdUF5RyuSs0aZoCzGLboneBUpCy8f9f99WMrG0,2009
 hud/cli/rl/gpu_utils.py,sha256=0nFRrmJZzLOHh_0bjMhIsBj94PAuu95vwxLd_sa4Q5g,11202
 hud/cli/rl/local_runner.py,sha256=_4_Bs8UXPz5Is6PyVg_DnrG0zL_Mr-6DJ5_6jC9o0qA,23199
 hud/cli/rl/presets.py,sha256=DzOO82xL5QyzdVtlX-Do1CODMvDz9ILMPapjU92jcZg,3051
-hud/cli/rl/remote_runner.py,sha256=fKmOVKSBUWfakunfe9-HAllpUJDxfRNZwL00fPw-QTI,17837
+hud/cli/rl/remote_runner.py,sha256=R8XXVrpO4l87nA8hlCcURmPfGzrszNLPuoi7-ZKPrhs,17837
 hud/cli/rl/rl_api.py,sha256=fvRMWQXhTSLM2zQaWWDas_u47RreH8erLgtXRKnQGeA,4350
 hud/cli/rl/viewer.py,sha256=ExQs1IX3T8x_9aBzc4JojZ779jmFvFTh7EjOYIHzYsU,4441
 hud/cli/rl/vllm.py,sha256=xTRvex9mmdatTnNaqwZFDBwUsX70bJP5pAGTpbWjmvM,6153
@@ -113,8 +113,8 @@ hud/cli/utils/tests/test_source_hash.py,sha256=GGFklT9xrVPMEpPwgiUzr94An66R-nhNn
 hud/cli/utils/tests/test_tasks.py,sha256=FT4smFGUHaJpZGIqj-yevBH8l9ARCu03RoedGxiLc2M,2624
 hud/clients/README.md,sha256=XNE3mch95ozDgVqfwCGcrhlHY9CwT1GKfNANNboowto,3826
 hud/clients/__init__.py,sha256=N5M_gZv4nP7dLRwpAiaqqaxyaLieGW6397FszeG7JGw,364
-hud/clients/base.py,sha256=QPFAjWKxczJWQEcsMGULQ9zj_xEMjMK52fWlTZ8rv9c,15274
-hud/clients/fastmcp.py,sha256=_dk1Iczvv_hKhti2z64KLE3cIY93W3vm3cvS8Og3C6w,9209
+hud/clients/base.py,sha256=3U3AVLxKdtH6fw2B4V_CXxLv-J93jzlGsdc3ArxDq2I,15274
+hud/clients/fastmcp.py,sha256=7yNk9SqOw-iBKPFJlt3hLAhR2Cl_QwVEln4OwXRd1Ww,9209
 hud/clients/mcp_use.py,sha256=Zq_NBFPAMb81qoQa_adCbdTm-yUFjDFCMFG7m8tOlr4,14781
 hud/clients/tests/__init__.py,sha256=sKOtJFFa4mDIXh1U6O8ZUHjigE8CiRMQ2PzJTIBZuVE,33
 hud/clients/tests/test_client_integration.py,sha256=kohU6jfCNfwSnAushHeB1_CmDlRfQc7VBL0GEdJYSeI,4198
@@ -141,8 +141,8 @@ hud/native/tests/test_comparator.py,sha256=pDch3r3xDi2o5YXF_bkoLfIdHcCjse3foAaqy
 hud/native/tests/test_native_init.py,sha256=Z-2dinbQYEkrbCcfBrBOLGdpXtWWOtkfPzp7ZKri68Y,2839
 hud/otel/__init__.py,sha256=ii17ayoWiS5vAhA7UAmZ8TkmP52gs2pWyHsD46-uYbE,1003
 hud/otel/collector.py,sha256=jLZymZ8r7xt2VDuWexfbnT7PY1-0aiyLMgjBy8KDY1M,4497
-hud/otel/config.py,sha256=rqw4gaQCqJNEFWO_AfyqwgZVia706qJP0UveNSgDVSg,7037
-hud/otel/context.py,sha256=pk73k0-ISpRdAuwp-2RDxvC85USggKlvzYUyxkf6I84,20403
+hud/otel/config.py,sha256=eHim5S4QdtrLKq53hJSiq02ezaYevN7Js2QbZyjHdVo,7037
+hud/otel/context.py,sha256=Lic5T4PFdyeM9xfQ1yqzRoK3jhmT4CsRR78mMmRP8co,20403
 hud/otel/exporters.py,sha256=k0yfUppMbcJ3IfNH3cc_u1hdR54qnnkT7lQjxO3I06g,21470
 hud/otel/instrumentation.py,sha256=fsFG9W89RdewFDxWKN9Ft4GUb7WbIKpfucTc16WxaZU,5093
 hud/otel/processors.py,sha256=-gGRbwifplcExDQBLfx_9tqWreDImULJNcENgO9q7VU,4700
@@ -170,7 +170,7 @@ hud/server/__init__.py,sha256=ZTxwhR7tMtSE14i1sONTz5UaMXURW1AYoFZMbWGBviU,134
 hud/server/context.py,sha256=6bCdSzv1FGyItu9472HbbYef279H7QuMGJDR8EtYg5Y,3210
 hud/server/low_level.py,sha256=XYs2pOJ9kN4OcJ6ahDmXM5mWkzq5wJLpKFInUYrWEok,4701
 hud/server/router.py,sha256=T9hLe08-W8SGvO14RfwY3j2zfcvB3lKWS5sR9h_4awM,5343
-hud/server/server.py,sha256=zj2FGv1SO8mR_3pjrBzhzN8LT1KpCls7_LCYwSjJ8Nw,30944
+hud/server/server.py,sha256=G_tnVuEBhC5gDNPfXChsPimo8L8ArMjQ9v3c1CSg848,40863
 hud/server/helper/__init__.py,sha256=ZxO8VP3RZEBBp-q65VixuhzQgqEPSVzW0hEY9J9QqDA,116
 hud/server/tests/__init__.py,sha256=eEYYkxX5Hz9woXVOBJ2H2_CQoEih0vH6nRt3sH2Z8v8,49
 hud/server/tests/test_add_tool.py,sha256=9Y59LJpow3BQ31Jg7fowhV7nAeyqude9Tap9tEs_vBE,1863
@@ -183,7 +183,7 @@ hud/server/tests/test_server_extra.py,sha256=blIO0ne4Rq7R40y1-O-P8ANeRHnv2iKFWqS
 hud/server/tests/test_sigterm_runner.py,sha256=HTM_0DAxA2exGYj7LK4udxMGXHZhY9LDZaKkHhQMu_Y,2610
 hud/shared/__init__.py,sha256=IPxPCqtPLguryN-nBq78Sakypw2bRiE2iHv3SXG8YRk,139
 hud/shared/exceptions.py,sha256=9IIqwlGuaNmcshEiFnph1jceZpWKpCnWgoH2OH-NL7I,13260
-hud/shared/hints.py,sha256=1UViFu6CwDhaKn66SUvJ-mdFnTVifG6u6Y1n7jufqCg,5717
+hud/shared/hints.py,sha256=aH9iIhGvC3Uo56z0H5H68Mh6HMMBMBqQyK8rvjv1Hc0,5717
 hud/shared/requests.py,sha256=FbvMU_dKSWu7OzNBKGAXlq-Q34zL1cULhEdGFa6rJyQ,9993
 hud/shared/tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 hud/shared/tests/test_exceptions.py,sha256=66k9Z1zQmYK4yPlG2NC7zggabpCQMFpGqRG3_lOWM8U,16376
@@ -192,7 +192,7 @@ hud/shared/tests/test_requests.py,sha256=nKFcSN1sjrOouVU2xik9lE5Wxapy3EWsO8iIXrM
 hud/telemetry/__init__.py,sha256=oYJ9fiaiwUGzfMH0qBixX74uSpAR_hyhl8Y-REjmNak,1396
 hud/telemetry/async_context.py,sha256=H3MBt1vi_ogxUB0U0zSkeO8EqjcU5Iv_04LVQWtTO-E,11550
 hud/telemetry/instrument.py,sha256=m3u6YK02PTk39Jr4L3se7l-cYyKx0maCaqf5Z5JqWNA,14096
-hud/telemetry/job.py,sha256=3W9KHLfBR13IYZvll496P9F6a9HUXCthFMY-MJLi7EA,13278
+hud/telemetry/job.py,sha256=fhIWvsGdp1RbwINVPRUC47i4zQj4xGfe3pjft4mu8No,13278
 hud/telemetry/replay.py,sha256=YW17s314s5Wy6Rl8MXHqg1FU8EF9_XcHBMJI0rrkyS4,2306
 hud/telemetry/trace.py,sha256=Q8ypSKKjjMOV6BdNH65iMnya8ifd2dcOqA9-38Td15Y,5254
 hud/telemetry/tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -205,7 +205,7 @@ hud/tools/__init__.py,sha256=i6lE0GxYcPnlLLd-55ryCCHo7o9anC4RfqkuYFXvzMQ,1009
 hud/tools/base.py,sha256=IxpnEHfxUtgz1k_w-09l65KKpDV3tQu4WdF3nZDsdwU,17854
 hud/tools/bash.py,sha256=1jl7cpB1ApGXn7Hy8zghJ2fXugEol6UeN0aYUSiM2EQ,5189
 hud/tools/edit.py,sha256=NYQL3coPIaG_-TP6DOpsVWFg1xcaMZwM5LtFye9WgNE,12644
-hud/tools/playwright.py,sha256=HY91JsC_6cvIa-9W14wIhjsAkAtkLPaXgUqMizTG-1c,14989
+hud/tools/playwright.py,sha256=7ggBkylR71taQtrPr5UNwdxtQrHNXMFzKcHW8qDOgzA,15330
 hud/tools/response.py,sha256=t6Oc8NM4u951A1XMCBaIkFyu3VNEQ8dcWURyTygfZmA,2228
 hud/tools/submit.py,sha256=hJG2G3Oex4fz_3CsAUVhOhAA56UvDMhquB29xCT-C3M,1973
 hud/tools/types.py,sha256=g-CWnUUDSxxIfUy54S1bpY1nfTzdYO1R_nPKYReABjQ,2734
@@ -266,10 +266,10 @@ hud/utils/tests/test_progress.py,sha256=QSF7Kpi03Ff_l3mAeqW9qs1nhK50j9vBiSobZq7T
 hud/utils/tests/test_tasks.py,sha256=Rbbm51vZxygyWlhjunFq4IfFPefVB3qevM9_CZSt5w4,5774
 hud/utils/tests/test_telemetry.py,sha256=5jl7bEx8C8b-FfFUko5pf4UY-mPOR-9HaeL98dGtVHM,2781
 hud/utils/tests/test_tool_shorthand.py,sha256=1p3j3D0G93OXHqnUXbvTs3G4A8awrPvwhPpLi6YPeOM,5458
-hud/utils/tests/test_version.py,sha256=hsHAkfILDrqA97RyX0l2JapaS40Gi1kJnbThs2QFZHg,160
+hud/utils/tests/test_version.py,sha256=3eFNjP_ugwRDZqUNNdWr6SX43UcvMw4739huirlrQ1o,160
 hud/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-hud_python-0.4.57.dist-info/METADATA,sha256=sHE5ewR3U5SRv-JbKoxZth7mPbXzqn50W6ZR3vl_g6g,22338
-hud_python-0.4.57.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-hud_python-0.4.57.dist-info/entry_points.txt,sha256=jJbodNFg1m0-CDofe5AHvB4zKBq7sSdP97-ohaQ3ae4,63
-hud_python-0.4.57.dist-info/licenses/LICENSE,sha256=yIzBheVUf86FC1bztAcr7RYWWNxyd3B-UJQ3uddg1HA,1078
-hud_python-0.4.57.dist-info/RECORD,,
+hud_python-0.4.58.dist-info/METADATA,sha256=XasQ6ALt1lQPCHGOmu7h_bgDmPc3TghvlPP7O9N7Jhs,22338
+hud_python-0.4.58.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+hud_python-0.4.58.dist-info/entry_points.txt,sha256=jJbodNFg1m0-CDofe5AHvB4zKBq7sSdP97-ohaQ3ae4,63
+hud_python-0.4.58.dist-info/licenses/LICENSE,sha256=yIzBheVUf86FC1bztAcr7RYWWNxyd3B-UJQ3uddg1HA,1078
+hud_python-0.4.58.dist-info/RECORD,,