ostruct-cli 0.8.29__py3-none-any.whl → 1.0.1__py3-none-any.whl

ostruct/cli/upload_manager.py

@@ -0,0 +1,448 @@
+"""Shared upload manager for multi-tool file sharing.
+
+This module implements a centralized upload management system that ensures files
+attached to multiple tools are uploaded only once and shared across all target tools.
+This prevents redundant uploads and improves performance for multi-target attachments.
+"""
+
+import logging
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Set, Tuple
+
+from openai import AsyncOpenAI
+
+from .attachment_processor import AttachmentSpec, ProcessedAttachments
+from .errors import CLIError
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class UploadRecord:
+    """Record of a file upload with status tracking."""
+
+    path: Path  # Original file path
+    upload_id: Optional[str] = None  # OpenAI file ID after upload
+    tools_pending: Set[str] = field(
+        default_factory=set
+    )  # Tools waiting for this file
+    tools_completed: Set[str] = field(
+        default_factory=set
+    )  # Tools that have received this file
+    file_size: Optional[int] = None  # File size in bytes
+    file_hash: Optional[str] = None  # File content hash for identity
+
+
+class UploadError(CLIError):
+    """Error during file upload operations."""
+
+    def __init__(self, message: str, **kwargs: Any):
+        """Initialize upload error with appropriate exit code."""
+        from .exit_codes import ExitCode
+
+        super().__init__(message, exit_code=ExitCode.USAGE_ERROR, **kwargs)
+
+    def __str__(self) -> str:
+        """Custom string representation without error type prefix."""
+        # For upload errors, we want clean user-friendly messages
+        # without the [USAGE_ERROR] prefix since our messages are already formatted
+        return self.message
+
+
+class SharedUploadManager:
+    """Manages file uploads across multiple tools to avoid duplicates.
+
+    This manager coordinates file uploads between Code Interpreter and File Search
+    tools, ensuring that files attached to multiple targets are uploaded only once
+    but shared across all requesting tools.
+    """
+
+    def __init__(self, client: AsyncOpenAI):
+        """Initialize the shared upload manager.
+
+        Args:
+            client: AsyncOpenAI client for file operations
+        """
+        self.client = client
+
+        # Map file identity -> upload record
+        self._uploads: Dict[Tuple[int, int], UploadRecord] = {}
+
+        # Queue of files needing upload for each tool
+        self._upload_queue: Dict[str, Set[Tuple[int, int]]] = {
+            "code-interpreter": set(),
+            "file-search": set(),
+        }
+
+        # Track all uploaded file IDs for cleanup
+        self._all_uploaded_ids: Set[str] = set()
+
+    def register_attachments(
+        self, processed_attachments: ProcessedAttachments
+    ) -> None:
+        """Register all attachments for upload management.
+
+        Args:
+            processed_attachments: Processed attachment specifications
+        """
+        logger.debug("Registering attachments for upload management")
+
+        # Register all attachment specs
+        for spec in processed_attachments.alias_map.values():
+            self._register_attachment_spec(spec)
+
+        logger.debug(
+            f"Registered {len(self._uploads)} unique files for upload, "
+            f"CI queue: {len(self._upload_queue['code-interpreter'])}, "
+            f"FS queue: {len(self._upload_queue['file-search'])}"
+        )
+
+    def _register_attachment_spec(self, spec: AttachmentSpec) -> None:
+        """Register a single attachment specification.
+
+        Args:
+            spec: Attachment specification to register
+        """
+        # Get file identity
+        file_id = self._get_file_identity(Path(spec.path))
+
+        # Create upload record if it doesn't exist
+        if file_id not in self._uploads:
+            self._uploads[file_id] = UploadRecord(
+                path=Path(spec.path),
+                tools_pending=set(),
+                tools_completed=set(),
+            )
+
+        # Handle directory vs file differently
+        if Path(spec.path).is_dir():
+            # For directories, we need to expand and register individual files
+            self._register_directory_files(spec, file_id)
+        else:
+            # For individual files, register with target tools
+            self._register_file_for_targets(file_id, spec.targets)
+
+    def _register_directory_files(
+        self, spec: AttachmentSpec, base_file_id: Tuple[int, int]
+    ) -> None:
+        """Register individual files from a directory attachment.
+
+        Args:
+            spec: Directory attachment specification
+            base_file_id: Base file identity for the directory
+        """
+        path = Path(spec.path)
+
+        # Expand directory to individual files
+        files = []
+        if spec.recursive:
+            if spec.pattern:
+                files = list(path.rglob(spec.pattern))
+            else:
+                files = [f for f in path.rglob("*") if f.is_file()]
+        else:
+            if spec.pattern:
+                files = list(path.glob(spec.pattern))
+            else:
+                files = [f for f in path.iterdir() if f.is_file()]
+
+        # Register each file individually
+        for file_path in files:
+            try:
+                file_id = self._get_file_identity(file_path)
+
+                if file_id not in self._uploads:
+                    self._uploads[file_id] = UploadRecord(
+                        path=file_path,
+                        tools_pending=set(),
+                        tools_completed=set(),
+                    )
+
+                self._register_file_for_targets(file_id, spec.targets)
+
+            except Exception as e:
+                logger.warning(f"Could not register file {file_path}: {e}")
+
+    def _register_file_for_targets(
+        self, file_id: Tuple[int, int], targets: Set[str]
+    ) -> None:
+        """Register a file for specific target tools.
+
+        Args:
+            file_id: File identity tuple
+            targets: Set of target tool names
+        """
+        record = self._uploads[file_id]
+
+        # Add to upload queues for tools that need uploads
+        for target in targets:
+            if target == "code-interpreter":
+                record.tools_pending.add(target)
+                self._upload_queue[target].add(file_id)
+            elif target == "file-search":
+                record.tools_pending.add(target)
+                self._upload_queue[target].add(file_id)
+            # "prompt" target doesn't need uploads, just template access
+
+    def _get_file_identity(self, path: Path) -> Tuple[int, int]:
+        """Get unique identity for a file based on inode and device.
+
+        This uses the file's inode and device numbers to create a unique
+        identity that works across different path representations of the same file.
+
+        Args:
+            path: File path to get identity for
+
+        Returns:
+            Tuple of (device, inode) for file identity
+
+        Raises:
+            OSError: If file cannot be accessed
+        """
+        try:
+            stat = Path(path).stat()
+            return (stat.st_dev, stat.st_ino)
+        except OSError as e:
+            logger.error(f"Cannot get file identity for {path}: {e}")
+            raise
+
+    async def upload_for_tool(self, tool: str) -> Dict[str, str]:
+        """Upload all queued files for a specific tool.
+
+        Args:
+            tool: Tool name ("code-interpreter" or "file-search")
+
+        Returns:
+            Dictionary mapping file paths to upload IDs
+
+        Raises:
+            UploadError: If any uploads fail
+        """
+        if tool not in self._upload_queue:
+            raise ValueError(f"Unknown tool: {tool}")
+
+        logger.debug(f"Processing uploads for {tool}")
+        uploaded = {}
+        failed_uploads = []
+
+        for file_id in self._upload_queue[tool]:
+            record = self._uploads[file_id]
+
+            try:
+                if record.upload_id is None:
+                    # First upload for this file
+                    record.upload_id = await self._perform_upload(record.path)
+                    self._all_uploaded_ids.add(record.upload_id)
+                    logger.info(
+                        f"Uploaded {record.path} -> {record.upload_id}"
+                    )
+                else:
+                    logger.debug(
+                        f"Reusing upload {record.upload_id} for {record.path}"
+                    )
+
+                uploaded[str(record.path)] = record.upload_id
+                record.tools_completed.add(tool)
+                record.tools_pending.discard(tool)
+
+            except UploadError as e:
+                # UploadError already has user-friendly message, don't duplicate logging
+                failed_uploads.append((record.path, str(e)))
+            except Exception as e:
+                logger.error(f"Failed to upload {record.path} for {tool}: {e}")
+                failed_uploads.append((record.path, str(e)))
+
+        if failed_uploads:
+            # If we have user-friendly error messages, present them cleanly
+            if len(failed_uploads) == 1:
+                # Single file failure - show the detailed error message
+                _, error_msg = failed_uploads[0]
+                raise UploadError(error_msg)
+            else:
+                # Multiple file failures - show summary with details
+                error_msg = (
+                    f"Failed to upload {len(failed_uploads)} files:\n\n"
+                )
+                for i, (path, error) in enumerate(failed_uploads, 1):
+                    error_msg += f"{i}. {error}\n\n"
+                raise UploadError(error_msg.rstrip())
+
+        logger.debug(f"Completed {len(uploaded)} uploads for {tool}")
+        return uploaded
+
+    async def _perform_upload(self, file_path: Path) -> str:
+        """Perform actual file upload with error handling.
+
+        Args:
+            file_path: Path to file to upload
+
+        Returns:
+            OpenAI file ID
+
+        Raises:
+            UploadError: If upload fails
+        """
+        try:
+            # Validate file exists and get info
+            if not file_path.exists():
+                raise FileNotFoundError(f"File not found: {file_path}")
+
+            file_size = file_path.stat().st_size
+            logger.debug(f"Uploading {file_path} ({file_size} bytes)")
+
+            # Perform upload
+            with open(file_path, "rb") as f:
+                upload_response = await self.client.files.create(
+                    file=f,
+                    purpose="assistants",  # Standard purpose for both CI and FS
+                )
+
+            logger.debug(
+                f"Successfully uploaded {file_path} as {upload_response.id}"
+            )
+            return upload_response.id
+
+        except Exception as e:
+            # Parse OpenAI API errors for better user experience
+            # Note: We don't log here as the error will be handled at a higher level
+            user_friendly_error = self._parse_upload_error(file_path, e)
+            raise UploadError(user_friendly_error)
+
+    def _parse_upload_error(self, file_path: Path, error: Exception) -> str:
+        """Parse OpenAI upload errors into user-friendly messages.
+
+        Args:
+            file_path: Path to the file that failed to upload
+            error: The original exception from OpenAI API
+
+        Returns:
+            User-friendly error message with actionable suggestions
+        """
+        error_str = str(error)
+        file_ext = file_path.suffix.lower()
+
+        # Check for unsupported file extension error
+        if (
+            "Invalid extension" in error_str
+            and "Supported formats" in error_str
+        ):
+            return (
+                f"❌ Cannot upload {file_path.name} to Code Interpreter/File Search\n"
+                f"   File extension '{file_ext}' is not supported by OpenAI's tools.\n"
+                f"   See: https://platform.openai.com/docs/assistants/tools/code-interpreter#supported-files\n\n"
+                f"💡 Solutions:\n"
+                f"   1. Use template-only routing (recommended):\n"
+                f"      Change: --dir ci:configs {file_path.parent}\n"
+                f"      To:     --dir configs {file_path.parent}\n"
+                f'      Your template can still access content with: {{{{ file_ref("configs") }}}}\n\n'
+                f"   2. Rename file to add .txt extension:\n"
+                f"      {file_path.name} → {file_path.name}.txt\n"
+                f"      Then use: --dir ci:configs {file_path.parent}\n"
+                f"      (OpenAI will treat it as text but preserve YAML content)"
+            )
+
+        # Check for file size errors
+        if (
+            "file too large" in error_str.lower()
+            or "size limit" in error_str.lower()
+        ):
+            return (
+                f"❌ File too large: {file_path.name}\n"
+                f"   OpenAI tools have a file size limit (typically 100MB).\n\n"
+                f"💡 Solutions:\n"
+                f"   1. Use template-only routing: --dir configs {file_path.parent}\n"
+                f"   2. Split large files into smaller chunks\n"
+                f"   3. Use file summarization before upload"
+            )
+
+        # Generic upload error with helpful context
+        return (
+            f"❌ Failed to upload {file_path.name}\n"
+            f"   {error_str}\n\n"
+            f"💡 If this is a configuration file, try template-only routing:\n"
+            f"   Change: --dir ci:configs {file_path.parent}\n"
+            f"   To:     --dir configs {file_path.parent}"
+        )
+
+    def get_upload_summary(self) -> Dict[str, Any]:
+        """Get summary of upload operations.
+
+        Returns:
+            Dictionary with upload statistics and status
+        """
+        total_files = len(self._uploads)
+        uploaded_files = sum(
+            1 for record in self._uploads.values() if record.upload_id
+        )
+        pending_files = total_files - uploaded_files
+
+        ci_files = len(self._upload_queue["code-interpreter"])
+        fs_files = len(self._upload_queue["file-search"])
+
+        return {
+            "total_files": total_files,
+            "uploaded_files": uploaded_files,
+            "pending_files": pending_files,
+            "code_interpreter_files": ci_files,
+            "file_search_files": fs_files,
+            "total_upload_ids": len(self._all_uploaded_ids),
+            "shared_files": total_files
+            - ci_files
+            - fs_files
+            + len(self._all_uploaded_ids),  # Files used by multiple tools
+        }
+
+    async def cleanup_uploads(self) -> None:
+        """Clean up all uploaded files.
+
+        This method deletes all files that were uploaded during this session.
+        Should be called when the operation is complete to avoid leaving
+        temporary files in OpenAI storage.
+        """
+        if not self._all_uploaded_ids:
+            logger.debug("No uploaded files to clean up")
+            return
+
+        logger.debug(
+            f"Cleaning up {len(self._all_uploaded_ids)} uploaded files"
+        )
+        cleanup_errors = []
+
+        for file_id in self._all_uploaded_ids:
+            try:
+                await self.client.files.delete(file_id)
+                logger.debug(f"Deleted uploaded file: {file_id}")
+            except Exception as e:
+                logger.warning(
+                    f"Failed to delete uploaded file {file_id}: {e}"
+                )
+                cleanup_errors.append((file_id, str(e)))
+
+        if cleanup_errors:
+            logger.error(f"Failed to clean up {len(cleanup_errors)} files")
+        else:
+            logger.debug("Successfully cleaned up all uploaded files")
+
+        # Clear tracking
+        self._all_uploaded_ids.clear()
+
+    def get_files_for_tool(self, tool: str) -> List[str]:
+        """Get list of upload IDs for a specific tool.
+
+        Args:
+            tool: Tool name ("code-interpreter" or "file-search")
+
+        Returns:
+            List of OpenAI file IDs for the tool
+        """
+        if tool not in self._upload_queue:
+            return []
+
+        file_ids = []
+        for file_id in self._upload_queue[tool]:
+            record = self._uploads[file_id]
+            if record.upload_id:
+                file_ids.append(record.upload_id)
+
+        return file_ids