`filesets.resources`#

Extended FilesResource classes with FilesetFileSystem support.

These classes extend the SDK’s generated FilesResource classes to add high-level file operations (upload, download, list, delete) and fsspec filesystem access.

Module Contents#

Classes#

`AsyncFilesResource`	Extended AsyncFilesResource with high-level file operations.
`AsyncReadable`	Protocol for async file-like objects (e.g., anyio.open_file(), aiofiles).
`FilesResource`	Extended FilesResource with high-level file operations.
`ListFilesResponse`	Response from listing files in a fileset.
`Readable`	Protocol for file-like objects.

Data#

`AsyncContent`
`SyncContent`

API#

filesets.resources.AsyncContent#: None

class filesets.resources.AsyncFilesResource(client: nemo_platform._client.AsyncNeMoPlatform)#

Bases: nemo_platform.resources.files.AsyncFilesResource

Extended AsyncFilesResource with high-level file operations.

Provides convenient methods for uploading, downloading, and listing files. For fsspec filesystem access, use sdk.files.fsspec.

Initialization

async delete( *, remote_path: str, fileset: str | None = None, workspace: str | None = None, ) → None#

Delete a file from a fileset (async).

Parameters:

remote_path – Path of the file to delete. Can be a full path (e.g., “workspace/fileset#data/file.txt”) if fileset is not provided, or a relative path (e.g., “data/file.txt”) if fileset is provided.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.

Examples

# Delete a file with explicit fileset >>> await sdk.files.delete( … fileset=”my-fileset”, … remote_path=”data/old-file.txt” … )

# Delete using full path >>> await sdk.files.delete(remote_path=”my-fileset#data/old-file.txt”)

Download files from a fileset to a local path (async).

Parameters:

remote_path –
Path(s) within the fileset to download. Can be: - A single path (str): Full path (e.g., “workspace/fileset#data/”),

relative path (e.g., “data/”), or glob pattern (e.g., “*.json”).
- A list of paths (list[str]): Multiple specific file paths to download. When using a list, fileset and workspace must be provided explicitly.
Defaults to “” (root of fileset).
local_path – Local destination path (directory).
fileset – Fileset name. If not provided, inferred from remote_path (str only).
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
callback – Optional progress callback (e.g., RichProgressCallback).
max_workers – Maximum number of concurrent file transfers.

Examples

# Explicit fileset/workspace >>> await sdk.files.download( … fileset=”my-fileset”, … workspace=”default”, … remote_path=”data/”, … local_path=”./downloads/” … )

# Inferred from path >>> await sdk.files.download( … remote_path=”default/my-fileset#data/”, … local_path=”./downloads/” … )

# Download files matching a glob pattern >>> await sdk.files.download( … fileset=”my-fileset”, … remote_path=”*.json”, … local_path=”./downloads/” … )

# Download files matching a pattern in a subdirectory >>> await sdk.files.download( … fileset=”my-fileset”, … remote_path=”data/*.jsonl”, … local_path=”./downloads/” … )

# Download a list of specific files >>> await sdk.files.download( … fileset=”my-fileset”, … remote_path=[“config.json”, “tokenizer.json”, “vocab.txt”], … local_path=”./downloads/” … )

async download_content( *, remote_path: str, fileset: str | None = None, workspace: str | None = None, ) → bytes#

Download a file’s content from a fileset (async).

Parameters:

remote_path – Path of the file within the fileset.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, uses SDK default.

Returns:

The file content.

Return type:

bytes

Examples

# Load JSON >>> content = await sdk.files.download_content( … remote_path=”config.json”, … fileset=”my-fileset”, … ) >>> data = json.loads(content)

# Get text content >>> text = (await sdk.files.download_content( … remote_path=”readme.txt”, … fileset=”my-fileset”, … )).decode(“utf-8”)

property fsspec: nemo_platform.filesets.filesystem.filesystem.FilesetFileSystem#

Get a FilesetFileSystem instance pre-configured with this SDK client.

This provides fsspec filesystem access. For high-level file operations, use sdk.files instead.

async list( *, remote_path: str = '', fileset: str | None = None, workspace: str | None = None, include_cache_status: bool = False, ) → filesets.resources.ListFilesResponse#

List all files in a fileset path (recursive, async), with optional glob pattern support.

Parameters:

remote_path – Path within the fileset to list. Can be a full path (e.g., “workspace/fileset#data/” or “fileset#data/”) if fileset is not provided, or a relative path (e.g., “data/”) if fileset is provided. Supports glob patterns (*, ?, []) for filtering files. Defaults to “” (root of fileset).
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
include_cache_status – Check and return cache status for each file. When False (default), external storage files return None for cache_status.

Returns:

ListFilesResponse with data (list of FilesetFile) and cache_status property.

Examples

# List all files in a fileset >>> response = await sdk.files.list(fileset=”my-fileset”) >>> for f in response.data: … print(f”{f.path}: {f.size} bytes”)

# List files in a subdirectory >>> await sdk.files.list( … fileset=”my-fileset”, … remote_path=”data/” … )

# List files matching a glob pattern >>> await sdk.files.list( … fileset=”my-fileset”, … remote_path=”*.json” … )

# List files matching a pattern in a subdirectory >>> await sdk.files.list( … fileset=”my-fileset”, … remote_path=”data/*.jsonl” … )

# Inferred from path >>> await sdk.files.list(remote_path=”my-fileset#data/”)

# Check cache status for external storage >>> response = await sdk.files.list(fileset=”my-fileset”, include_cache_status=True) >>> print(f”Cache status: {response.cache_status}”) >>> for f in response.data: … print(f”{f.path}: {f.cache_status}”)

async upload( *, local_path: str, remote_path: str = '', fileset: str | None = None, workspace: str | None = None, callback: fsspec.callbacks.Callback | None = None, max_workers: int | None = None, fileset_auto_create: bool = False, ) → nemo_platform.types.files.fileset.Fileset#

Upload files from a local path to a fileset (async).

Parameters:

local_path – Local source path (file or directory).
remote_path – Path within the fileset to upload to. Can be a full path (e.g., “workspace/fileset#data/” or “fileset#data/”) if fileset is not provided, or a relative path (e.g., “data/”) if fileset is provided. Defaults to “” (root of fileset).
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
callback – Optional progress callback (e.g., RichProgressCallback).
max_workers – Maximum number of concurrent file transfers.
fileset_auto_create – If True, create the fileset if it doesn’t exist. When no fileset is specified (neither as param nor in remote_path), a unique name is generated (e.g., “fileset-a1b2c3d4”).

Returns:

The fileset that was uploaded to. Check fileset.name to see: the generated name when using fileset_auto_create without specifying a fileset.

Return type:

Fileset

Examples

# Explicit fileset/workspace >>> await sdk.files.upload( … fileset=”my-fileset”, … workspace=”default”, … local_path=”./data/”, … remote_path=”uploads/” … )

# Inferred from path >>> await sdk.files.upload( … local_path=”./file.txt”, … remote_path=”default/my-fileset#file.txt” … )

# Auto-create fileset with specified name >>> fileset = await sdk.files.upload( … local_path=”./data/”, … fileset=”new-fileset”, … fileset_auto_create=True … ) >>> print(f”Uploaded to: {fileset.name}”)

# Auto-create fileset with generated name >>> fileset = await sdk.files.upload( … local_path=”./data/”, … fileset_auto_create=True … ) >>> print(f”Uploaded to: {fileset.name}”) # e.g., “fileset-a1b2c3d4”

async upload_content( *, content: filesets.resources.AsyncContent, remote_path: str, fileset: str | None = None, workspace: str | None = None, fileset_auto_create: bool = False, ) → nemo_platform.types.files.fileset.Fileset#

Upload in-memory data to a fileset (async).

Parameters:

content – Content to upload. Can be: - bytes: Raw byte content - str: Text content (will be UTF-8 encoded) - AsyncReadable: Async file-like object (e.g., anyio.open_file(), aiofiles) - AsyncIterator[bytes]: Async iterator yielding byte chunks (streamed)
remote_path – Destination path within the fileset.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, uses SDK default.
fileset_auto_create – If True, create the fileset if it doesn’t exist. When no fileset is specified (neither as param nor in remote_path), a unique name is generated (e.g., “fileset-a1b2c3d4”).

Returns:

The fileset that was uploaded to. Check fileset.name to see: the generated name when using fileset_auto_create without specifying a fileset.

Return type:

Fileset

Examples

# Upload bytes >>> await sdk.files.upload_content( … content=b”Hello, World!”, … remote_path=”message.txt”, … fileset=”my-fileset”, … )

# Upload string (auto UTF-8 encoded) >>> await sdk.files.upload_content( … content=’{“key”: “value”}’, … remote_path=”config.json”, … fileset=”my-fileset”, … )

# Upload from async file (anyio/aiofiles) >>> async with await anyio.open_file(“data.bin”, “rb”) as f: … await sdk.files.upload_content( … content=f, … remote_path=”data.bin”, … fileset=”my-fileset”, … )

# Auto-create fileset with specified name >>> fileset = await sdk.files.upload_content( … content=b”content”, … remote_path=”file.txt”, … fileset=”new-fileset”, … fileset_auto_create=True, … ) >>> print(f”Uploaded to: {fileset.name}”)

# Auto-create fileset with generated name >>> fileset = await sdk.files.upload_content( … content=b”content”, … remote_path=”file.txt”, … fileset_auto_create=True, … ) >>> print(f”Uploaded to: {fileset.name}”) # e.g., “fileset-a1b2c3d4”

class filesets.resources.AsyncReadable#

Bases: typing.Protocol

Protocol for async file-like objects (e.g., anyio.open_file(), aiofiles).

async read(size: int = -1) → bytes#

class filesets.resources.FilesResource(client: nemo_platform._client.NeMoPlatform)#

Bases: nemo_platform.resources.files.FilesResource

Extended FilesResource with high-level file operations.

Provides convenient methods for uploading, downloading, and listing files. For fsspec filesystem access, use sdk.files.fsspec.

Initialization

delete( *, remote_path: str, fileset: str | None = None, workspace: str | None = None, ) → None#

Delete a file from a fileset.

Parameters:

remote_path – Path of the file to delete. Can be a full path (e.g., “workspace/fileset#data/file.txt”) if fileset is not provided, or a relative path (e.g., “data/file.txt”) if fileset is provided.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.

Examples

# Delete a file with explicit fileset >>> sdk.files.delete( … fileset=”my-fileset”, … remote_path=”data/old-file.txt” … )

# Delete using full path >>> sdk.files.delete(remote_path=”my-fileset#data/old-file.txt”)

Download files from a fileset to a local path.

Parameters:

remote_path –
Path(s) within the fileset to download. Can be: - A single path (str): Full path (e.g., “workspace/fileset#data/”),

relative path (e.g., “data/”), or glob pattern (e.g., “*.json”).
- A list of paths (list[str]): Multiple specific file paths to download. When using a list, fileset and workspace must be provided explicitly.
Defaults to “” (root of fileset).
local_path – Local destination path (directory).
fileset – Fileset name. If not provided, inferred from remote_path (str only).
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
callback – Optional progress callback (e.g., RichProgressCallback).
max_workers – Maximum number of concurrent file transfers.

Examples

# Explicit fileset/workspace >>> sdk.files.download( … fileset=”my-fileset”, … workspace=”default”, … remote_path=”data/”, … local_path=”./downloads/” … )

# Inferred from path (with workspace) >>> sdk.files.download( … remote_path=”default/my-fileset#data/”, … local_path=”./downloads/” … )

# Inferred from path (workspace from SDK default) >>> sdk.files.download( … remote_path=”my-fileset#data/”, … local_path=”./downloads/” … )

# Download files matching a glob pattern >>> sdk.files.download( … fileset=”my-fileset”, … remote_path=”*.json”, … local_path=”./downloads/” … )

# Download files matching a pattern in a subdirectory >>> sdk.files.download( … fileset=”my-fileset”, … remote_path=”data/*.jsonl”, … local_path=”./downloads/” … )

# Download a list of specific files >>> sdk.files.download( … fileset=”my-fileset”, … remote_path=[“config.json”, “tokenizer.json”, “vocab.txt”], … local_path=”./downloads/” … )

# With progress callback >>> from nemo_platform.filesets import RichProgressCallback >>> with RichProgressCallback(description=”Downloading”) as cb: … sdk.files.download( … remote_path=”my-fileset#”, … local_path=”./”, … callback=cb … )

download_content( *, remote_path: str, fileset: str | None = None, workspace: str | None = None, ) → bytes#

Download a file’s content from a fileset.

Parameters:

remote_path – Path of the file within the fileset.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, uses SDK default.

Returns:

The file content.

Return type:

bytes

Examples

# Load JSON (most common use case) >>> data = json.loads(sdk.files.download_content( … remote_path=”config.json”, … fileset=”my-fileset”, … ))

# Get text content >>> text = sdk.files.download_content( … remote_path=”readme.txt”, … fileset=”my-fileset”, … ).decode(“utf-8”)

# Get binary content >>> content = sdk.files.download_content( … remote_path=”model.bin”, … fileset=”my-fileset”, … )

property fsspec: nemo_platform.filesets.filesystem.filesystem.FilesetFileSystem#: Access the underlying fsspec filesystem.

list( *, remote_path: str = '', fileset: str | None = None, workspace: str | None = None, include_cache_status: bool = False, ) → filesets.resources.ListFilesResponse#

List all files in a fileset path (recursive), with optional glob pattern support.

Parameters:

remote_path – Path within the fileset to list. Can be a full path (e.g., “workspace/fileset#data/” or “fileset#data/”) if fileset is not provided, or a relative path (e.g., “data/”) if fileset is provided. Supports glob patterns (*, ?, []) for filtering files. Defaults to “” (root of fileset).
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
include_cache_status – Check and return cache status for each file. When False (default), external storage files return None for cache_status.

Returns:

ListFilesResponse with data (list of FilesetFile) and cache_status property.

Examples

# List all files in a fileset >>> response = sdk.files.list(fileset=”my-fileset”) >>> for f in response.data: … print(f”{f.path}: {f.size} bytes”)

# List files in a subdirectory >>> sdk.files.list( … fileset=”my-fileset”, … remote_path=”data/” … )

# List files matching a glob pattern >>> sdk.files.list( … fileset=”my-fileset”, … remote_path=”*.json” … )

# List files matching a pattern in a subdirectory >>> sdk.files.list( … fileset=”my-fileset”, … remote_path=”data/*.jsonl” … )

# Inferred from path >>> sdk.files.list(remote_path=”my-fileset#data/”)

# Check cache status for external storage >>> response = sdk.files.list(fileset=”my-fileset”, include_cache_status=True) >>> print(f”Cache status: {response.cache_status}”) >>> for f in response.data: … print(f”{f.path}: {f.cache_status}”)

upload( *, local_path: str, remote_path: str = '', fileset: str | None = None, workspace: str | None = None, callback: fsspec.callbacks.Callback | None = None, max_workers: int | None = None, fileset_auto_create: bool = False, ) → nemo_platform.types.files.fileset.Fileset#

Upload files from a local path to a fileset.

Parameters:

local_path – Local source path (file or directory).
remote_path – Path within the fileset to upload to. Can be a full path (e.g., “workspace/fileset#data/” or “fileset#data/”) if fileset is not provided, or a relative path (e.g., “data/”) if fileset is provided. Defaults to “” (root of fileset).
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, inferred from remote_path or uses the SDK’s default workspace.
callback – Optional progress callback (e.g., RichProgressCallback).
max_workers – Maximum number of concurrent file transfers.
fileset_auto_create – If True, create the fileset if it doesn’t exist. When no fileset is specified (neither as param nor in remote_path), a unique name is generated (e.g., “fileset-a1b2c3d4”).

Returns:

The fileset that was uploaded to. Check fileset.name to see: the generated name when using fileset_auto_create without specifying a fileset.

Return type:

Fileset

Examples

# Explicit fileset/workspace >>> sdk.files.upload( … fileset=”my-fileset”, … workspace=”default”, … local_path=”./data/”, … remote_path=”uploads/” … )

# Inferred from path >>> sdk.files.upload( … local_path=”./file.txt”, … remote_path=”default/my-fileset#file.txt” … )

# With workspace from SDK default >>> sdk.files.upload( … local_path=”./file.txt”, … remote_path=”my-fileset#file.txt” … )

# Auto-create fileset with specified name >>> fileset = sdk.files.upload( … local_path=”./data/”, … fileset=”new-fileset”, … fileset_auto_create=True … ) >>> print(f”Uploaded to: {fileset.name}”)

# Auto-create fileset with generated name >>> fileset = sdk.files.upload( … local_path=”./data/”, … fileset_auto_create=True … ) >>> print(f”Uploaded to: {fileset.name}”) # e.g., “fileset-a1b2c3d4”

upload_content( *, content: filesets.resources.SyncContent, remote_path: str, fileset: str | None = None, workspace: str | None = None, fileset_auto_create: bool = False, ) → nemo_platform.types.files.fileset.Fileset#

Upload in-memory content to a fileset.

Parameters:

content – Content to upload. Can be: - bytes: Raw byte content - str: Text content (will be UTF-8 encoded) - BinaryIO: File-like object (e.g., BytesIO, open file) - Iterator[bytes]: Generator or iterator yielding byte chunks
remote_path – Destination path within the fileset.
fileset – Fileset name. If not provided, inferred from remote_path.
workspace – Workspace name. If not provided, uses SDK default.
fileset_auto_create – If True, create the fileset if it doesn’t exist. When no fileset is specified (neither as param nor in remote_path), a unique name is generated (e.g., “fileset-a1b2c3d4”).

Returns:

The fileset that was uploaded to. Check fileset.name to see: the generated name when using fileset_auto_create without specifying a fileset.

Return type:

Fileset

Examples

# Upload bytes >>> sdk.files.upload_content( … content=b”Hello, World!”, … remote_path=”message.txt”, … fileset=”my-fileset”, … )

# Upload string (auto UTF-8 encoded) >>> sdk.files.upload_content( … content=’{“key”: “value”}’, … remote_path=”config.json”, … fileset=”my-fileset”, … )

# Upload from BytesIO >>> from io import BytesIO >>> sdk.files.upload_content( … content=BytesIO(b”content”), … remote_path=”data.bin”, … fileset=”my-fileset”, … )

# Auto-create fileset with specified name >>> fileset = sdk.files.upload_content( … content=b”content”, … remote_path=”file.txt”, … fileset=”new-fileset”, … fileset_auto_create=True, … ) >>> print(f”Uploaded to: {fileset.name}”)

# Auto-create fileset with generated name >>> fileset = sdk.files.upload_content( … content=b”content”, … remote_path=”file.txt”, … fileset_auto_create=True, … ) >>> print(f”Uploaded to: {fileset.name}”) # e.g., “fileset-a1b2c3d4”

class filesets.resources.ListFilesResponse#

Response from listing files in a fileset.

data#: List of files in the fileset.

Properties:

cache_status: Aggregate cache status of all files.

“caching” if any file is actively being cached
“not_cached” if any file is not cached (and none are caching)
“cached” if all files are fully cached
“not_cacheable” if all files cannot be cached
None if no cache information is available

property cache_status: nemo_platform.types.files.CacheStatus | None#

Get aggregate cache status of all files.

Returns the most relevant status based on priority: - “caching” if any file is actively being cached - “not_cached” if any file is not cached (and none are caching) - “cached” if all files are fully cached - “not_cacheable” if all files cannot be cached - None if no cache information is available

data: list[nemo_platform.types.files.FilesetFile]#: None

class filesets.resources.Readable#

Bases: typing.Protocol

Protocol for file-like objects.

read(size: int = -1) → bytes#

filesets.resources.SyncContent#: None

filesets.resources#

Module Contents#

Classes#

Data#

API#

`filesets.resources`#