hud-python 0.4.71__tar.gz → 0.4.73__tar.gz

{hud_python-0.4.71 → hud_python-0.4.73}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hud-python
-Version: 0.4.71
+Version: 0.4.73
 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
@@ -131,6 +131,7 @@ OSS RL environment + evals toolkit. Wrap software as environments, run benchmark
 [![Discord](https://img.shields.io/discord/1327447144772407390?label=Discord&logo=discord&style=flat-square)](https://discord.gg/wkjtmHYYjm)
 [![X Follow](https://img.shields.io/twitter/follow/hud_evals?style=social)](https://x.com/intent/user?screen_name=hud_evals)
 [![Shop](https://img.shields.io/badge/_-white.svg?label=shop&logo=data:image/png;base64,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&style=social)](https://shop.hud.ai)
+[![Scarf](https://static.scarf.sh/a.png?x-pxid=6530ff33-4945-452b-81f9-626872593933)](https://scarf.sh)
 
 
 ### Are you an enterprise building agents?

{hud_python-0.4.71 → hud_python-0.4.73}/README.md

@@ -14,6 +14,7 @@ OSS RL environment + evals toolkit. Wrap software as environments, run benchmark
 [![Discord](https://img.shields.io/discord/1327447144772407390?label=Discord&logo=discord&style=flat-square)](https://discord.gg/wkjtmHYYjm)
 [![X Follow](https://img.shields.io/twitter/follow/hud_evals?style=social)](https://x.com/intent/user?screen_name=hud_evals)
 [![Shop](https://img.shields.io/badge/_-white.svg?label=shop&logo=data:image/png;base64,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&style=social)](https://shop.hud.ai)
+[![Scarf](https://static.scarf.sh/a.png?x-pxid=6530ff33-4945-452b-81f9-626872593933)](https://scarf.sh)
 
 
 ### Are you an enterprise building agents?

{hud_python-0.4.71 → hud_python-0.4.73}/hud/agents/gemini_cua.py

@@ -62,7 +62,7 @@ class GeminiCUAConfig(GeminiConfig):
     model_config = ConfigDict(arbitrary_types_allowed=True)
 
     model_name: str = "GeminiCUA"
-    checkpoint_name: str = "gemini-2.5-computer-use-preview-10-2025"
+    checkpoint_name: str = "gemini-2.5-computer-use-preview"
     excluded_predefined_functions: list[str] = Field(default_factory=list)
 
 

{hud_python-0.4.71 → hud_python-0.4.73}/hud/agents/misc/response_agent.py

@@ -37,14 +37,14 @@ class ResponseAgent:
 
     def __init__(
         self,
-        model: str = "openai/gpt-4o",
+        model: str = "gpt-4o",
         system_prompt: str | None = None,
     ) -> None:
         """
         Initialize the ResponseAgent.
 
         Args:
-            model: The model to use via HUD inference gateway (default: "openai/gpt-4o-mini").
+            model: The model to use via HUD inference gateway (default: "gpt-4o").
                    Supports any model available through inference.hud.ai.
             system_prompt: Optional custom system prompt for determining responses.
         """

{hud_python-0.4.71 → hud_python-0.4.73}/hud/agents/tests/test_gemini.py

@@ -317,12 +317,12 @@ class TestGeminiCUAAgent:
         agent = GeminiCUAAgent.create(
             mcp_client=mock_mcp_client_gemini_computer,
             model_client=mock_gemini_client,
-            checkpoint_name="gemini-2.5-computer-use-preview-10-2025",
+            checkpoint_name="gemini-2.5-computer-use-preview",
             validate_api_key=False,  # Skip validation in tests
         )
 
         assert agent.model_name == "GeminiCUA"
-        assert agent.config.checkpoint_name == "gemini-2.5-computer-use-preview-10-2025"
+        assert agent.config.checkpoint_name == "gemini-2.5-computer-use-preview"
         assert agent.gemini_client == mock_gemini_client
 
     @pytest.mark.asyncio

{hud_python-0.4.71 → hud_python-0.4.73}/hud/cli/__init__.py

@@ -560,10 +560,11 @@ def rft_run(
         ...,
         help="Path to tasks file (JSON/JSONL)",
     ),
-    provider: str = typer.Option(
-        "openai",
-        "--provider",
-        help="Provider to use (e.g., openai)",
+    model_id: str | None = typer.Option(
+        None,
+        "--model-id",
+        "-m",
+        help="Model ID to train (skip interactive selection)",
     ),
     reasoning_effort: str = typer.Option(
         "medium",
@@ -586,10 +587,10 @@ def rft_run(
     """Launch an RFT training job."""
     rft_command(
         tasks_file=tasks_file,
-        provider=provider,
         reasoning_effort=reasoning_effort,
         verbose=verbose,
         yes=yes,
+        model_id=model_id,
     )
 
 

{hud_python-0.4.71 → hud_python-0.4.73}/hud/cli/eval.py

@@ -54,7 +54,7 @@ _AGENT_PRESETS: list[AgentPreset] = [
     AgentPreset(
         "Gemini CUA (Gemini Computer Use)",
         AgentType.GEMINI_CUA,
-        "gemini-2.5-computer-use-preview-10-2025",
+        "gemini-2.5-computer-use-preview",
     ),
     # HUD Gateway presets (models via HUD Inference API)
     AgentPreset(
@@ -66,7 +66,7 @@ _AGENT_PRESETS: list[AgentPreset] = [
     AgentPreset(
         "GLM-4.5V (Z-AI)",
         AgentType.OPENAI_COMPATIBLE,
-        "glm-4.5v",
+        "z-ai/glm-4.5v",
         {"openai_compatible": {"base_url": settings.hud_gateway_url, "model_name": "GLM"}},
     ),
 ]

{hud_python-0.4.71 → hud_python-0.4.73}/hud/cli/rft.py

@@ -3,8 +3,10 @@ from __future__ import annotations
 import logging
 from typing import Any
 
+import httpx
 import typer
 from rich.console import Console
+from rich.table import Table
 
 from hud.settings import settings
 from hud.utils.hud_console import HUDConsole
@@ -39,12 +41,86 @@ def _patch_mcp_urls_to_staging(tasks: list[dict[str, Any]]) -> list[dict[str, An
     return [patch_value(task) for task in tasks]
 
 
+def _fetch_models() -> list[dict[str, Any]]:
+    """Fetch trainable models from the HUD API for the user's team."""
+    url = f"{settings.hud_api_url}/models/"
+    headers = {
+        "Authorization": f"Bearer {settings.api_key}",
+        "x-api-key": settings.api_key or "",
+    }
+    params = {"team_only": "true", "limit": 200}
+
+    try:
+        with httpx.Client(timeout=30.0) as client:
+            resp = client.get(url, headers=headers, params=params)
+            resp.raise_for_status()
+            data = resp.json()
+            return data.get("models", [])
+    except httpx.HTTPStatusError as e:
+        hud_console.error(f"Failed to fetch models: {e.response.status_code}")
+        if e.response.status_code == 401:
+            hud_console.hint("Check that your HUD_API_KEY is valid")
+        raise typer.Exit(1) from e
+    except httpx.RequestError as e:
+        hud_console.error(f"Connection error while fetching models: {e}")
+        raise typer.Exit(1) from e
+
+
+def _select_model(models: list[dict[str, Any]]) -> dict[str, Any]:
+    """Display models and let user select one for training."""
+    # Filter to only trainable models that are ready
+    trainable_models = [
+        m
+        for m in models
+        if m.get("is_trainable", False)
+        and m.get("status") == "ready"
+        and not m.get("public", False)
+        and m.get("model_name") is not None
+    ]
+
+    if not trainable_models:
+        hud_console.error("No trainable models found in your team.")
+        hud_console.hint("Fork a trainable model at https://api.hud.so/models to start training.")
+        raise typer.Exit(1)
+
+    # Display models in a table
+    hud_console.section_title("Available Trainable Models")
+    table = Table(show_header=True, header_style="bold")
+    table.add_column("#", style="dim", width=4)
+    table.add_column("Name", style="bold")
+    table.add_column("Status")
+    table.add_column("Provider")
+
+    for i, model in enumerate(trainable_models, 1):
+        provider_name = (
+            model.get("provider", {}).get("name", "unknown") if model.get("provider") else "unknown"
+        )
+        table.add_row(
+            str(i),
+            model.get("name", "unnamed"),
+            model.get("status", "unknown"),
+            provider_name,
+        )
+
+    hud_console.console.print(table)
+    hud_console.print("")
+
+    # Build choices for selection
+    choices = [
+        {"name": f"{m.get('name', 'unnamed')} ({m.get('base_model', 'unknown')})", "value": m}
+        for m in trainable_models
+    ]
+
+    selected: dict[str, Any] = hud_console.select("Select a model to train:", choices)  # type: ignore[assignment]
+    return selected
+
+
 def rft_command(
     tasks_file: str,
-    provider: str = "openai",
     reasoning_effort: str = "medium",
     verbose: bool = False,
     yes: bool = False,
+    model_id: str | None = None,
 ) -> None:
     """
     Run Reinforcement Fine-Tuning (RFT) via the HUD RL service.
@@ -57,6 +133,47 @@ def rft_command(
         hud_console.info("Run 'hud set HUD_API_KEY=...' or export it.")
         raise typer.Exit(1)
 
+    # Model selection
+    selected_model_id: str
+    if model_id:
+        # Use provided model_id directly
+        selected_model_id = model_id
+        hud_console.info(f"Using provided model ID: {selected_model_id}")
+    else:
+        # Fetch and let user select a model
+        hud_console.section_title("Fetching available models")
+        hud_console.info("Loading models from your team...")
+        models = _fetch_models()
+
+        if yes:
+            # Auto-select first trainable model in non-interactive mode
+            trainable_models = [
+                m
+                for m in models
+                if m.get("is_trainable", False)
+                and m.get("status") == "ready"
+                and not m.get("public", False)
+                and m.get("model_name") is not None
+            ]
+            if not trainable_models:
+                hud_console.error("No trainable models found in your team.")
+                hud_console.hint(
+                    "Fork a trainable model at https://api.hud.so/models to start training."
+                )
+                raise typer.Exit(1)
+            selected_model = trainable_models[0]
+            hud_console.info(
+                f"Auto-selected first trainable model (--yes mode): "
+                f"{selected_model.get('name', 'unnamed')}"
+            )
+        else:
+            selected_model = _select_model(models)
+
+        selected_model_id = selected_model["id"]
+        hud_console.success(
+            f"Selected model: {selected_model.get('name', 'unnamed')} (ID: {selected_model_id})"
+        )
+
     # Preflight check: Convert tasks to remote if needed
     hud_console.section_title("Preparing tasks for remote training")
     try:
@@ -153,8 +270,7 @@ def rft_command(
 
     # Prepare payload
     payload = {
-        "provider": provider,
-        "base_model": "o4-mini-2025-04-16",
+        "model_id": selected_model_id,
         "dataset": {"tasks": tasks},
         "config": {"parameters": {"reasoning_effort": reasoning_effort}},
     }
@@ -162,8 +278,6 @@ def rft_command(
     # Send request to service
     hud_console.section_title("Submitting RFT job")
 
-    import httpx
-
     base_url = settings.hud_rl_url
     url = f"{base_url}/training/jobs"
 

{hud_python-0.4.71 → hud_python-0.4.73}/hud/datasets/runner.py

@@ -29,6 +29,7 @@ async def run_single_task(
     job_id: str | None = None,
     task_id: str | None = None,
     group_id: str | None = None,
+    trace_id: str | None = None,
     trace_name: str | None = None,
     metadata: dict[str, Any] | None = None,
 ) -> Trace:
@@ -46,6 +47,7 @@ async def run_single_task(
         job_id: Job ID for telemetry grouping
         task_id: Task ID for telemetry
         group_id: Group ID for variance estimation runs
+        trace_id: Trace ID for telemetry (auto-generated if not provided)
         trace_name: Name for the trace (defaults to task prompt)
         metadata: Additional trace metadata
 
@@ -59,6 +61,7 @@ async def run_single_task(
         job_id=job_id,
         task_id=task_id,
         group_id=group_id,
+        trace_id=trace_id,
         attrs=metadata or {},
     ):
         agent = agent_type.cls.create(**(agent_params or {}))

{hud_python-0.4.71 → hud_python-0.4.73}/hud/tools/computer/openai.py

@@ -22,6 +22,8 @@ logger = logging.getLogger(__name__)
 class Coordinate(BaseModel):
     """A coordinate point with x and y values."""
 
+    model_config = {"extra": "forbid"}  # Ensures additionalProperties: false in JSON schema
+
     x: int = Field(..., description="X coordinate")
     y: int = Field(..., description="Y coordinate")
 

{hud_python-0.4.71 → hud_python-0.4.73}/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__ == "0.4.71"
+    assert hud.__version__ == "0.4.73"

{hud_python-0.4.71 → hud_python-0.4.73}/hud/version.py

@@ -4,4 +4,4 @@ Version information for the HUD SDK.
 
 from __future__ import annotations
 
-__version__ = "0.4.71"
+__version__ = "0.4.73"

{hud_python-0.4.71 → hud_python-0.4.73}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "hud-python"
-version = "0.4.71"
+version = "0.4.73"
 description = "SDK for the HUD platform."
 readme = "README.md"
 requires-python = ">=3.11, <3.13"