griptape-nodes 0.52.1__py3-none-any.whl → 0.54.0__py3-none-any.whl

griptape_nodes/retained_mode/managers/model_manager.py

@@ -0,0 +1,1107 @@
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import re
+import sys
+import threading
+from dataclasses import dataclass
+from datetime import UTC, datetime
+from pathlib import Path
+from typing import TYPE_CHECKING, Any
+from urllib.parse import urlparse
+
+from huggingface_hub import list_models, scan_cache_dir, snapshot_download
+from huggingface_hub.utils.tqdm import tqdm
+from xdg_base_dirs import xdg_data_home
+
+from griptape_nodes.retained_mode.events.app_events import AppInitializationComplete
+from griptape_nodes.retained_mode.events.model_events import (
+    DeleteModelDownloadRequest,
+    DeleteModelDownloadResultFailure,
+    DeleteModelDownloadResultSuccess,
+    DeleteModelRequest,
+    DeleteModelResultFailure,
+    DeleteModelResultSuccess,
+    DownloadModelRequest,
+    DownloadModelResultFailure,
+    DownloadModelResultSuccess,
+    ListModelDownloadsRequest,
+    ListModelDownloadsResultFailure,
+    ListModelDownloadsResultSuccess,
+    ListModelsRequest,
+    ListModelsResultFailure,
+    ListModelsResultSuccess,
+    ModelDownloadStatus,
+    ModelInfo,
+    QueryInfo,
+    SearchModelsRequest,
+    SearchModelsResultFailure,
+    SearchModelsResultSuccess,
+)
+from griptape_nodes.retained_mode.griptape_nodes import GriptapeNodes
+from griptape_nodes.utils.async_utils import cancel_subprocess
+
+if TYPE_CHECKING:
+    from griptape_nodes.retained_mode.events.base_events import ResultPayload
+    from griptape_nodes.retained_mode.managers.event_manager import EventManager
+
+logger = logging.getLogger("griptape_nodes")
+
+
+HTTP_UNAUTHORIZED = 401
+HTTP_FORBIDDEN = 403
+
+MIN_CACHE_DIR_PARTS = 3
+
+
+@dataclass
+class SearchResultsData:
+    """Data class for model search results."""
+
+    models: list[ModelInfo]
+    total_results: int
+    query_info: QueryInfo
+
+
+class ModelDownloadTracker(tqdm):
+    """Custom tqdm progress bar that tracks aggregate model download progress."""
+
+    _file_lock = threading.Lock()
+    _current_model_id = ""
+
+    @classmethod
+    def set_current_model_id(cls, model_id: str) -> None:
+        """Set the current model being downloaded."""
+        cls._current_model_id = model_id
+
+    def __init__(self, *args, model_id: str = "", **kwargs):
+        if not model_id and self._current_model_id:
+            model_id = self._current_model_id
+
+        super().__init__(*args, **kwargs)
+        self.model_id = model_id
+        self.start_time = datetime.now(UTC).isoformat()
+
+        logger.debug(
+            "ModelDownloadTracker created - model_id: %s, total: %s, desc: %s, args: %s",
+            self.model_id,
+            self.total,
+            getattr(self, "desc", None),
+            args,
+        )
+
+        if self.model_id:
+            self._init_status_file()
+
+    def update(self, n: int = 1) -> None:
+        """Override update to track progress in status file."""
+        logger.debug(
+            "ModelDownloadTracker update - model_id: %s, n: %s, self.n: %s, total: %s",
+            self.model_id,
+            n,
+            self.n,
+            self.total,
+        )
+        super().update(n)
+        self._update_status_file()
+
+    def close(self) -> None:
+        """Override close to log download completion."""
+        super().close()
+        logger.debug(
+            "ModelDownloadTracker close - model_id: %s, self.n: %s, total: %s", self.model_id, self.n, self.total
+        )
+
+    def _get_status_file_path(self) -> Path:
+        """Get the path to the status file for this model."""
+        status_dir = xdg_data_home() / "griptape_nodes" / "model_downloads"
+        status_dir.mkdir(parents=True, exist_ok=True)
+
+        sanitized_model_id = re.sub(r"[^\w\-_]", "--", self.model_id)
+        return status_dir / f"{sanitized_model_id}.json"
+
+    def _init_status_file(self) -> None:
+        """Initialize the status file for this model."""
+        try:
+            with self._file_lock:
+                status_file = self._get_status_file_path()
+                current_time = datetime.now(UTC).isoformat()
+
+                logger.info(
+                    "ModelDownloadTracker initializing status file: %s (total_files=%s)", status_file, self.total
+                )
+
+                data = {
+                    "model_id": self.model_id,
+                    "status": "downloading",
+                    "started_at": current_time,
+                    "updated_at": current_time,
+                    "total_files": self.total or 0,
+                    "downloaded_files": 0,
+                    "progress_percent": 0.0,
+                }
+
+                with status_file.open("w") as f:
+                    json.dump(data, f, indent=2)
+
+                logger.info("ModelDownloadTracker status file initialized successfully")
+
+        except Exception:
+            logger.exception("ModelDownloadTracker._init_status_file failed")
+
+    def _update_status_file(self) -> None:
+        """Update the status file with current progress."""
+        if not self.model_id:
+            logger.warning("ModelDownloadTracker._update_status_file called with empty model_id")
+            return
+
+        try:
+            with self._file_lock:
+                status_file = self._get_status_file_path()
+                logger.info("ModelDownloadTracker updating status file: %s", status_file)
+
+                if not status_file.exists():
+                    logger.warning("Status file does not exist: %s", status_file)
+                    return
+
+                with status_file.open() as f:
+                    data = json.load(f)
+
+                current_time = datetime.now(UTC).isoformat()
+                progress_percent = (self.n / self.total * 100) if self.total else 0
+
+                logger.info(
+                    "ModelDownloadTracker updating progress: files=%d/%d, percent=%.1f%%",
+                    self.n,
+                    self.total,
+                    progress_percent,
+                )
+
+                data.update(
+                    {
+                        "downloaded_files": self.n,
+                        "progress_percent": progress_percent,
+                        "updated_at": current_time,
+                    }
+                )
+
+                with status_file.open("w") as f:
+                    json.dump(data, f, indent=2)
+
+                logger.debug("ModelDownloadTracker status file updated successfully")
+
+        except Exception:
+            logger.exception("ModelDownloadTracker._update_status_file failed")
+
+
+class ModelManager:
+    """A manager for downloading models from Hugging Face Hub.
+
+    This manager provides async handlers for downloading models using the Hugging Face Hub API.
+    It supports downloading entire model repositories or specific files, with caching and
+    local storage management.
+    """
+
+    def __init__(self, event_manager: EventManager | None = None) -> None:
+        """Initialize the ModelManager.
+
+        Args:
+            event_manager: The EventManager instance to use for event handling.
+        """
+        self._download_tasks = {}
+        self._download_processes = {}
+
+        if event_manager is not None:
+            event_manager.assign_manager_to_request_type(DownloadModelRequest, self.on_handle_download_model_request)
+            event_manager.assign_manager_to_request_type(ListModelsRequest, self.on_handle_list_models_request)
+            event_manager.assign_manager_to_request_type(DeleteModelRequest, self.on_handle_delete_model_request)
+            event_manager.assign_manager_to_request_type(SearchModelsRequest, self.on_handle_search_models_request)
+            event_manager.assign_manager_to_request_type(
+                ListModelDownloadsRequest, self.on_handle_list_model_downloads_request
+            )
+            event_manager.assign_manager_to_request_type(
+                DeleteModelDownloadRequest, self.on_handle_delete_model_download_request
+            )
+
+            event_manager.add_listener_to_app_event(AppInitializationComplete, self.on_app_initialization_complete)
+
+    def download_model(
+        self,
+        model_id: str,
+        local_dir: str | None = None,
+        revision: str = "main",
+        allow_patterns: list[str] | None = None,
+        ignore_patterns: list[str] | None = None,
+    ) -> str:
+        """Direct model download method that can be used without event system.
+
+        This method contains the core download logic without going through
+        the event system, avoiding recursion issues.
+
+        Args:
+            model_id: Model ID to download
+            local_dir: Optional local directory to download to
+            revision: Git revision to download
+            allow_patterns: Optional glob patterns to include
+            ignore_patterns: Optional glob patterns to exclude
+
+        Returns:
+            str: Local path where the model was downloaded
+
+        Raises:
+            Exception: If download fails
+        """
+        # Set up progress tracking
+        ModelDownloadTracker.set_current_model_id(model_id)
+
+        try:
+            # Build download kwargs
+            download_kwargs = {
+                "repo_id": model_id,
+                "repo_type": "model",
+                "revision": revision,
+                "tqdm_class": ModelDownloadTracker,
+            }
+
+            # Add optional parameters
+            if local_dir:
+                download_kwargs["local_dir"] = local_dir
+            if allow_patterns:
+                download_kwargs["allow_patterns"] = allow_patterns
+            if ignore_patterns:
+                download_kwargs["ignore_patterns"] = ignore_patterns
+
+            # Execute download with progress tracking
+            local_path = snapshot_download(**download_kwargs)  # type: ignore[arg-type]
+
+            return str(local_path)
+
+        finally:
+            # Clear the current model ID when done
+            ModelDownloadTracker.set_current_model_id("")
+
+    def _get_status_directory(self) -> Path:
+        """Get the status directory path for model downloads.
+
+        Returns:
+            Path: Path to the status directory, creating it if needed
+        """
+        status_dir = xdg_data_home() / "griptape_nodes" / "model_downloads"
+        status_dir.mkdir(parents=True, exist_ok=True)
+        return status_dir
+
+    async def on_handle_download_model_request(self, request: DownloadModelRequest) -> ResultPayload:
+        """Handle model download requests asynchronously.
+
+        This method starts a background task to download models from Hugging Face Hub using the provided parameters.
+        It supports both model IDs and full URLs, and can download entire repositories
+        or specific files based on the patterns provided.
+
+        Args:
+            request: The download request containing model ID and options
+
+        Returns:
+            ResultPayload: Success result indicating download started or failure with error details
+        """
+        parsed_model_id = self._parse_model_id(request.model_id)
+        if parsed_model_id != request.model_id:
+            logger.debug("Parsed model ID '%s' from URL '%s'", parsed_model_id, request.model_id)
+
+        try:
+            download_params = {
+                "model_id": parsed_model_id,
+                "local_dir": request.local_dir,
+                "revision": request.revision,
+                "allow_patterns": request.allow_patterns,
+                "ignore_patterns": request.ignore_patterns,
+            }
+
+            task = asyncio.create_task(self._download_model_task(download_params))
+            self._download_tasks[parsed_model_id] = task
+
+            result_details = f"Started background download for model '{parsed_model_id}'"
+
+            return DownloadModelResultSuccess(
+                model_id=parsed_model_id,
+                result_details=result_details,
+            )
+
+        except Exception as e:
+            error_msg = f"Failed to start download for model '{request.model_id}': {e}"
+            return DownloadModelResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+
+    async def _download_model_task(self, download_params: dict[str, str | list[str] | None]) -> None:
+        """Background task for downloading a model using CLI command.
+
+        Args:
+            download_params: Dictionary containing download parameters
+        """
+        model_id = download_params["model_id"]
+        logger.info("Starting background download for model: %s", model_id)
+
+        # Build CLI command arguments
+        cmd = [sys.executable, "-m", "griptape_nodes", "models", "download", str(model_id)]
+
+        # Add optional parameters
+        if download_params.get("local_dir"):
+            cmd.extend(["--local-dir", str(download_params["local_dir"])])
+        if download_params.get("revision") and download_params["revision"] != "main":
+            cmd.extend(["--revision", str(download_params["revision"])])
+
+        try:
+            # Start subprocess
+            process = await asyncio.create_subprocess_exec(
+                *cmd,
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+            )
+
+            # Store process for cancellation
+            if isinstance(model_id, str):
+                self._download_processes[model_id] = process
+
+            # Wait for completion (this can be cancelled)
+            stdout, stderr = await process.communicate()
+
+            if process.returncode == 0:
+                logger.info("Successfully downloaded model '%s'", model_id)
+            else:
+                logger.error("Download failed for model '%s': %s", model_id, stderr.decode())
+
+        except asyncio.CancelledError:
+            logger.info("Download cancelled for model '%s'", model_id)
+            raise
+
+        except Exception:
+            logger.exception("Error downloading model '%s'", model_id)
+
+        finally:
+            if isinstance(model_id, str):
+                if model_id in self._download_tasks:
+                    del self._download_tasks[model_id]
+                if model_id in self._download_processes:
+                    del self._download_processes[model_id]
+
+    async def on_handle_list_models_request(self, request: ListModelsRequest) -> ResultPayload:  # noqa: ARG002
+        """Handle model listing requests asynchronously.
+
+        This method scans the local Hugging Face cache directory to find downloaded models
+        and returns information about each model including path, size, and metadata.
+
+        Args:
+            request: The list request (no parameters needed)
+
+        Returns:
+            ResultPayload: Success result with model list or failure with error details
+        """
+        try:
+            # Get models in a thread to avoid blocking the event loop
+            models = await asyncio.to_thread(self._list_models)
+
+            result_details = f"Found {len(models)} cached models"
+
+            return ListModelsResultSuccess(
+                models=models,
+                result_details=result_details,
+            )
+
+        except Exception as e:
+            error_msg = f"Failed to list models: {e}"
+            return ListModelsResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+
+    async def on_handle_delete_model_request(self, request: DeleteModelRequest) -> ResultPayload:
+        """Handle model deletion requests asynchronously.
+
+        This method removes a model from the local Hugging Face cache directory and
+        cleans up any associated download tracking records.
+
+        Args:
+            request: The delete request containing model_id
+
+        Returns:
+            ResultPayload: Success result with deletion confirmation or failure with error details
+        """
+        # Parse the model ID from potential URL
+        model_id = request.model_id
+
+        deleted_items = []
+
+        try:
+            deleted_path = await asyncio.to_thread(self._delete_model, model_id)
+            deleted_items.append(f"model files from '{deleted_path}'")
+
+        except FileNotFoundError:
+            logger.debug("No model files found for '%s' in cache", model_id)
+
+        except Exception as e:
+            error_msg = f"Failed to delete model files for '{model_id}': {e}"
+            return DeleteModelResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+
+        if not deleted_items:
+            error_msg = f"Model '{model_id}' not found (no cached files or download records)"
+            return DeleteModelResultFailure(
+                result_details=error_msg,
+                exception=FileNotFoundError(error_msg),
+            )
+
+        deleted_description = " and ".join(deleted_items)
+        result_details = f"Successfully deleted {deleted_description} for model '{model_id}'"
+
+        return DeleteModelResultSuccess(
+            model_id=model_id,
+            deleted_path=deleted_items[0] if deleted_items else "",
+            result_details=result_details,
+        )
+
+    async def on_handle_search_models_request(self, request: SearchModelsRequest) -> ResultPayload:
+        """Handle model search requests asynchronously.
+
+        This method searches for models on Hugging Face Hub using the provided parameters.
+        It supports filtering by query, task, library, author, and tags.
+
+        Args:
+            request: The search request containing search parameters
+
+        Returns:
+            ResultPayload: Success result with model list or failure with error details
+        """
+        try:
+            # Search models in a thread to avoid blocking the event loop
+            search_results = await asyncio.to_thread(self._search_models, request)
+        except Exception as e:
+            error_msg = f"Failed to search models: {e}"
+            return SearchModelsResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+        else:
+            result_details = f"Found {len(search_results.models)} models"
+            return SearchModelsResultSuccess(
+                models=search_results.models,
+                total_results=search_results.total_results,
+                query_info=search_results.query_info,
+                result_details=result_details,
+            )
+
+    def _search_models(self, request: SearchModelsRequest) -> SearchResultsData:
+        """Synchronous model search implementation.
+
+        Searches for models on Hugging Face Hub using the huggingface_hub API.
+
+        Args:
+            request: The search request parameters
+
+        Returns:
+            SearchResultsData: Dataclass containing models list, total results, and query info
+        """
+        # Build search parameters
+        search_params = {}
+
+        if request.query:
+            search_params["search"] = request.query
+        if request.task:
+            search_params["task"] = request.task
+        if request.library:
+            search_params["library"] = request.library
+        if request.author:
+            search_params["author"] = request.author
+        if request.tags:
+            search_params["tags"] = request.tags
+
+        # Validate and set sort parameters
+        valid_sorts = ["downloads", "likes", "updated", "created"]
+        sort_param = request.sort if request.sort in valid_sorts else "downloads"
+        search_params["sort"] = sort_param
+
+        # Only add direction for sorts that support it (downloads only supports descending)
+        if sort_param != "downloads":
+            # Convert direction to the format expected by HF Hub API (-1 for asc, 1 for desc)
+            direction_param = -1 if request.direction == "asc" else 1
+            search_params["direction"] = direction_param
+
+        # Limit results (max 100 as per HF Hub API)
+        limit = min(max(1, request.limit), 100)
+
+        # Perform the search
+        models_iterator = list_models(limit=limit, **search_params)
+
+        # Convert models to list and extract information
+        models_list = []
+        for model in models_iterator:
+            created_at = getattr(model, "created_at", None)
+            updated_at = getattr(model, "last_modified", None)
+
+            model_info = ModelInfo(
+                model_id=model.id,
+                author=getattr(model, "author", None),
+                downloads=getattr(model, "downloads", None),
+                likes=getattr(model, "likes", None),
+                created_at=created_at.isoformat() if created_at else None,
+                updated_at=updated_at.isoformat() if updated_at else None,
+                task=getattr(model, "pipeline_tag", None),
+                library=getattr(model, "library_name", None),
+                tags=getattr(model, "tags", None),
+            )
+            models_list.append(model_info)
+
+        # Prepare query info for response
+        query_info = QueryInfo(
+            query=request.query,
+            task=request.task,
+            library=request.library,
+            author=request.author,
+            tags=request.tags,
+            limit=limit,
+            sort=sort_param,
+            direction=request.direction,  # Keep the original user-friendly format
+        )
+
+        return SearchResultsData(
+            models=models_list,
+            total_results=len(models_list),
+            query_info=query_info,
+        )
+
+    async def on_app_initialization_complete(self, _payload: AppInitializationComplete) -> None:
+        """Handle app initialization complete event by downloading configured models and resuming unfinished downloads.
+
+        Args:
+            payload: The app initialization complete payload
+        """
+        # Get models to download from configuration
+        config_manager = GriptapeNodes.ConfigManager()
+        models_to_download = config_manager.get_config_value(
+            "app_events.on_app_initialization_complete.models_to_download", default=[]
+        )
+
+        # Find unfinished downloads to resume
+        unfinished_models = await asyncio.to_thread(self._find_unfinished_downloads)
+
+        # Combine new downloads and unfinished ones, avoiding duplicates
+        all_models = list(
+            dict.fromkeys(
+                [
+                    *[model_id for model_id in models_to_download if model_id],  # Filter empty strings
+                    *unfinished_models,
+                ]
+            )
+        )
+
+        if not all_models:
+            logger.debug("No models to download or resume")
+            return
+
+        logger.info(
+            "Starting download/resume of %d models (%d new, %d resuming)",
+            len(all_models),
+            len(models_to_download),
+            len(unfinished_models),
+        )
+
+        # Create download tasks for concurrent execution
+        download_tasks = []
+        for model_id in all_models:
+            task = asyncio.create_task(self._download_model_with_logging(model_id))
+            download_tasks.append(task)
+
+        # Wait for all downloads to complete
+        results = await asyncio.gather(*download_tasks, return_exceptions=True)
+
+        # Log summary of results
+        successful = sum(1 for result in results if not isinstance(result, Exception))
+        failed = len(results) - successful
+
+        logger.info("Completed automatic model downloads: %d successful, %d failed", successful, failed)
+
+    async def _download_model_with_logging(self, model_id: str) -> None:
+        """Download a single model with proper logging.
+
+        Args:
+            model_id: The model ID to download
+        """
+        logger.info("Auto-downloading model: %s", model_id)
+
+        # Create download request with default parameters
+        request = DownloadModelRequest(
+            model_id=model_id,
+            local_dir=None,  # Use default cache directory
+            revision="main",
+            allow_patterns=None,
+            ignore_patterns=None,
+        )
+
+        try:
+            # Run the download asynchronously
+            result = await self.on_handle_download_model_request(request)
+
+            if isinstance(result, DownloadModelResultFailure):
+                logger.warning("Failed to auto-download model '%s': %s", model_id, result.result_details)
+            elif not isinstance(result, DownloadModelResultSuccess):
+                logger.warning("Unknown result type for model '%s' download: %s", model_id, type(result))
+
+        except Exception as e:
+            logger.error("Unexpected error auto-downloading model '%s': %s", model_id, e)
+            raise
+
+    def _get_status_file_path(self, model_id: str) -> Path:
+        """Get the path to the status file for a model.
+
+        Args:
+            model_id: The model ID to get status file path for
+
+        Returns:
+            Path: Path to the status file for this model
+        """
+        status_dir = self._get_status_directory()
+
+        sanitized_model_id = re.sub(r"[^\w\-_]", "--", model_id)
+        return status_dir / f"{sanitized_model_id}.json"
+
+    def _read_model_download_status(self, model_id: str) -> ModelDownloadStatus | None:
+        """Read download status for a specific model.
+
+        Args:
+            model_id: The model ID to get status for
+
+        Returns:
+            ModelDownloadStatus | None: The status if found, None otherwise
+        """
+        status_file = self._get_status_file_path(model_id)
+
+        if not status_file.exists():
+            return None
+
+        try:
+            with status_file.open() as f:
+                data = json.load(f)
+
+            # Get file counts from simplified structure
+            total_files = data.get("total_files", 0)
+            downloaded_files = data.get("downloaded_files", 0)
+
+            # For simplified tracking, failed_files is calculated
+            failed_files = 0
+            if data.get("status") == "failed":
+                failed_files = total_files - downloaded_files
+
+            return ModelDownloadStatus(
+                model_id=data["model_id"],
+                status=data["status"],
+                started_at=data["started_at"],
+                updated_at=data["updated_at"],
+                total_files=total_files,
+                completed_files=downloaded_files,
+                failed_files=failed_files,
+                completed_at=data.get("completed_at"),
+                local_path=data.get("local_path"),
+                failed_at=data.get("failed_at"),
+                error_message=data.get("error_message"),
+            )
+
+        except (json.JSONDecodeError, KeyError) as e:
+            logger.warning("Failed to read status file for model '%s': %s", model_id, e)
+            return None
+
+    def _list_all_download_statuses(self) -> list[ModelDownloadStatus]:
+        """List all model download statuses from status files.
+
+        Returns:
+            list[ModelDownloadStatus]: List of all download statuses
+        """
+        status_dir = self._get_status_directory()
+
+        if not status_dir.exists():
+            return []
+
+        statuses = []
+        for status_file in status_dir.glob("*.json"):
+            try:
+                with status_file.open() as f:
+                    data = json.load(f)
+
+                model_id = data.get("model_id", "")
+                if model_id:
+                    status = self._read_model_download_status(model_id)
+                    if status:
+                        statuses.append(status)
+
+            except (json.JSONDecodeError, KeyError) as e:
+                logger.warning("Failed to read status file '%s': %s", status_file, e)
+                continue
+
+        return statuses
+
+    def _find_unfinished_downloads(self) -> list[str]:
+        """Find model IDs with unfinished downloads from status files.
+
+        Returns:
+            list[str]: List of model IDs with status 'downloading' or 'failed'
+        """
+        status_dir = self._get_status_directory()
+
+        if not status_dir.exists():
+            return []
+
+        unfinished_models = []
+        for status_file in status_dir.glob("*.json"):
+            try:
+                with status_file.open() as f:
+                    data = json.load(f)
+
+                status = data.get("status", "")
+                model_id = data.get("model_id", "")
+
+                if model_id and status in ("downloading", "failed"):
+                    unfinished_models.append(model_id)
+
+            except (json.JSONDecodeError, KeyError) as e:
+                logger.warning("Failed to read status file '%s': %s", status_file, e)
+                continue
+
+        return unfinished_models
+
+    async def on_handle_list_model_downloads_request(self, request: ListModelDownloadsRequest) -> ResultPayload:
+        """Handle model download status requests asynchronously.
+
+        This method retrieves download status for a specific model or all models
+        from the local status files stored in the XDG data directory.
+
+        Args:
+            request: The status request containing optional model_id
+
+        Returns:
+            ResultPayload: Success result with download status list or failure with error details
+        """
+        try:
+            # Get status information in a thread to avoid blocking the event loop
+            downloads = await asyncio.to_thread(self._get_download_statuses, request.model_id)
+
+            if request.model_id and not downloads:
+                result_details = f"No download status found for model '{request.model_id}'"
+            elif request.model_id:
+                result_details = f"Found download status for model '{request.model_id}'"
+            else:
+                result_details = f"Found {len(downloads)} download status records"
+
+            return ListModelDownloadsResultSuccess(
+                downloads=downloads,
+                result_details=result_details,
+            )
+
+        except Exception as e:
+            error_msg = f"Failed to get download status: {e}"
+            return ListModelDownloadsResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+
+    def _list_models(self) -> list[ModelInfo]:
+        """Synchronous model listing implementation using HuggingFace Hub SDK.
+
+        Uses scan_cache_dir to get information about cached models.
+
+        Returns:
+            list[ModelInfo]: List of model information
+        """
+        try:
+            cache_info = scan_cache_dir()
+            models = []
+
+            for repo in cache_info.repos:
+                # Calculate total size across all revisions
+                total_size = sum(revision.size_on_disk for revision in repo.revisions)
+
+                model_info = ModelInfo(
+                    model_id=repo.repo_id,
+                    local_path=str(repo.repo_path),
+                    size_bytes=total_size,
+                )
+                models.append(model_info)
+
+        except Exception as e:
+            logger.warning("Failed to scan cache directory: %s", e)
+            return []
+        else:
+            return models
+
+    def _delete_model(self, model_id: str) -> str:
+        """Synchronous model deletion implementation using HuggingFace Hub SDK.
+
+        Uses scan_cache_dir to find and delete the model from cache.
+
+        Args:
+            model_id: The model ID to delete
+
+        Returns:
+            str: Information about what was deleted
+
+        Raises:
+            FileNotFoundError: If the model is not found in cache
+        """
+        cache_info = scan_cache_dir()
+
+        # Find the repo to delete
+        repo_to_delete = None
+        for repo in cache_info.repos:
+            if repo.repo_id == model_id:
+                repo_to_delete = repo
+                break
+
+        if repo_to_delete is None:
+            error_msg = f"Model '{model_id}' not found in cache"
+            raise FileNotFoundError(error_msg)
+
+        # Get all revision hashes for this repo
+        revision_hashes = [revision.commit_hash for revision in repo_to_delete.revisions]
+
+        if not revision_hashes:
+            error_msg = f"No revisions found for model '{model_id}'"
+            raise FileNotFoundError(error_msg)
+
+        # Create delete strategy for all revisions of this repo
+        delete_strategy = cache_info.delete_revisions(*revision_hashes)
+
+        # Execute the deletion
+        delete_strategy.execute()
+
+        return f"Deleted model '{model_id}' (freed {delete_strategy.expected_freed_size_str})"
+
+    def _get_model_info(self, model_dir: Path) -> dict[str, str | int | float] | None:
+        """Get information about a cached model.
+
+        Args:
+            model_dir: Path to the model directory in cache
+
+        Returns:
+            dict | None: Model information or None if not a valid model directory
+        """
+        try:
+            # Extract model_id from directory name
+            # HuggingFace cache format: models--{org}--{model}--{hash}
+            dir_name = model_dir.name
+            if not dir_name.startswith("models--"):
+                return None
+
+            # Parse the model ID from the directory name
+            parts = dir_name.split("--")
+            if len(parts) >= MIN_CACHE_DIR_PARTS:
+                # Reconstruct model_id as org/model
+                model_id = f"{parts[1]}/{parts[2]}"
+            else:
+                model_id = dir_name[8:]  # Remove "models--" prefix
+
+            # Calculate directory size
+            total_size = sum(f.stat().st_size for f in model_dir.rglob("*") if f.is_file())
+
+            return {
+                "model_id": model_id,
+                "local_path": str(model_dir),
+                "size_bytes": total_size,
+                "size_mb": round(total_size / (1024 * 1024), 2),
+            }
+
+        except Exception:
+            # If we can't parse the directory, skip it
+            return None
+
+    def _download_model(self, download_params: dict[str, str | list[str] | None]) -> Path:
+        """Model download implementation.
+
+        Args:
+            download_params: Dictionary containing download parameters
+
+        Returns:
+            Path: Local path where the model was downloaded
+        """
+        # Validate parameters and build download kwargs
+        download_kwargs = self._build_download_kwargs(download_params)
+
+        # Set the current model ID for progress tracking
+        model_id = download_params["model_id"]
+        if isinstance(model_id, str):
+            ModelDownloadTracker.set_current_model_id(model_id)
+
+        # Execute download with progress tracking
+        local_path = snapshot_download(**download_kwargs)  # type: ignore[arg-type]
+
+        # Clear the current model ID when done
+        ModelDownloadTracker.set_current_model_id("")
+
+        return Path(local_path)
+
+    def _build_download_kwargs(self, download_params: dict[str, str | list[str] | None]) -> dict:
+        """Build kwargs for snapshot_download with validation.
+
+        Args:
+            download_params: Dictionary containing download parameters
+
+        Returns:
+            dict: Validated download kwargs for snapshot_download
+        """
+        param_model_id = download_params["model_id"]
+        local_dir = download_params["local_dir"]
+        revision = download_params["revision"]
+        allow_patterns = download_params["allow_patterns"]
+        ignore_patterns = download_params["ignore_patterns"]
+
+        # Build base kwargs with custom progress tracking
+        download_kwargs: dict[str, Any] = {
+            "repo_id": param_model_id,
+            "repo_type": "model",
+            "revision": revision,
+            "tqdm_class": ModelDownloadTracker,
+        }
+
+        # Add optional parameters
+        if local_dir is not None and isinstance(local_dir, str):
+            download_kwargs["local_dir"] = local_dir
+        if allow_patterns is not None and isinstance(allow_patterns, list):
+            download_kwargs["allow_patterns"] = allow_patterns
+        if ignore_patterns is not None and isinstance(ignore_patterns, list):
+            download_kwargs["ignore_patterns"] = ignore_patterns
+
+        return download_kwargs
+
+    def _parse_model_id(self, model_input: str) -> str:
+        """Parse model ID from either a direct model ID or a Hugging Face URL.
+
+        Args:
+            model_input: Either a model ID (e.g., 'microsoft/DialoGPT-medium')
+                        or a Hugging Face URL (e.g., 'https://huggingface.co/microsoft/DialoGPT-medium')
+
+        Returns:
+            str: The parsed model ID in the format 'namespace/repo_name' or 'repo_name'
+        """
+        # If it's already a simple model ID (no URL scheme), return as-is
+        if not model_input.startswith(("http://", "https://")):
+            return model_input
+
+        # Parse the URL
+        parsed = urlparse(model_input)
+
+        # Check if it's a Hugging Face URL
+        if parsed.netloc in ("huggingface.co", "www.huggingface.co"):
+            # Extract the path and remove leading slash
+            path = parsed.path.lstrip("/")
+
+            # Remove any trailing parameters or fragments
+            # The model ID should be in the format: namespace/repo_name or just repo_name
+            model_id_match = re.match(r"^([^/]+/[^/?#]+|[^/?#]+)", path)
+            if model_id_match:
+                return model_id_match.group(1)
+
+        # If we can't parse it, return the original input and let huggingface_hub handle the error
+        return model_input
+
+    def _get_download_statuses(self, model_id: str | None = None) -> list[ModelDownloadStatus]:
+        """Get download statuses for a specific model or all models.
+
+        Args:
+            model_id: Optional model ID to get status for. If None, returns all statuses.
+
+        Returns:
+            list[ModelDownloadStatus]: List of download statuses
+        """
+        if model_id:
+            # Get status for specific model
+            status = self._read_model_download_status(model_id)
+            return [status] if status else []
+        # Get all download statuses
+        return self._list_all_download_statuses()
+
+    async def on_handle_delete_model_download_request(self, request: DeleteModelDownloadRequest) -> ResultPayload:
+        """Handle model download status deletion requests asynchronously.
+
+        This method removes download tracking records for a specific model
+        from the local status files stored in the XDG data directory.
+        If the model is currently downloading or failed, it also cancels
+        the download task and deletes any cached model files.
+
+        Args:
+            request: The delete request containing model_id
+
+        Returns:
+            ResultPayload: Success result with deletion confirmation or failure with error details
+        """
+        model_id = request.model_id
+
+        try:
+            # Check current download status first
+            download_status = await asyncio.to_thread(self._read_model_download_status, model_id)
+
+            # Cancel active download process if it exists
+            if model_id in self._download_processes:
+                process = self._download_processes[model_id]
+                await cancel_subprocess(process, f"download process for model '{model_id}'")
+                del self._download_processes[model_id]
+
+            if model_id in self._download_tasks:
+                task = self._download_tasks[model_id]
+                if not task.done():
+                    task.cancel()
+                    logger.info("Cancelled active download task for model '%s'", model_id)
+                del self._download_tasks[model_id]
+
+            # Delete status file
+            deleted_path = await asyncio.to_thread(self._delete_model_download_status, model_id)
+
+            # Only delete cached model if it's not completed
+            if download_status and download_status.status != "completed":
+                try:
+                    await asyncio.to_thread(self._delete_model, model_id)
+                except FileNotFoundError:
+                    logger.debug("No cached model files found for '%s'", model_id)
+
+            result_details = f"Successfully deleted download status for model '{model_id}'"
+
+            return DeleteModelDownloadResultSuccess(
+                model_id=model_id,
+                deleted_path=deleted_path,
+                result_details=result_details,
+            )
+
+        except FileNotFoundError:
+            error_msg = f"Download status for model '{model_id}' not found"
+            return DeleteModelDownloadResultFailure(
+                result_details=error_msg,
+                exception=FileNotFoundError(error_msg),
+            )
+
+        except Exception as e:
+            error_msg = f"Failed to delete download status for '{model_id}': {e}"
+            return DeleteModelDownloadResultFailure(
+                result_details=error_msg,
+                exception=e,
+            )
+
+    def _delete_model_download_status(self, model_id: str) -> str:
+        """Delete download status file for a specific model.
+
+        Args:
+            model_id: The model ID to remove download status for
+
+        Returns:
+            str: Path to the deleted status file
+
+        Raises:
+            FileNotFoundError: If the status file is not found
+        """
+        status_file = self._get_status_file_path(model_id)
+
+        if not status_file.exists():
+            msg = f"Download status file not found for model '{model_id}'"
+            raise FileNotFoundError(msg)
+
+        status_file.unlink()
+        return str(status_file)