truss-transfer
Speed up file transfers with the baseten.co + baseten_fs.
What it is and what it does
Truss-Transfer is a download utility that resolves Baseten Pointers—references to model files and data stored across cloud providers—into local directories. It wraps a Rust implementation with a Python API and CLI, allowing you to define which models or datasets to download from HuggingFace, Google Cloud Storage, AWS S3, or Azure, then fetch them on demand with optional caching and authentication via environment variables or secrets.
The package is designed for machine learning workflows where you need to download large model repositories at runtime. You define a manifest of models using PyModelRepo objects (specifying repo ID, revision, storage location, and file patterns), create a bptr manifest, then call lazy_data_resolve() to download files to a specified directory. It handles authentication through runtime secrets and supports filtering files by include/exclude patterns.
Use it for:
- Download HuggingFace model checkpoints to a local directory during model initialization in a serving container.
- Create a manifest of multiple models from different cloud providers and resolve them all to a shared volume.
- Set up a data loader that fetches model files on first use with automatic caching to avoid repeated downloads.
- Integrate with Baseten's b10fs caching layer to speed up repeated access to the same model files.
- Manage authentication for private model repositories using environment-based secrets.
Worth the install?
AI-flagged interpretation of the facts on this page — verify before relying
Downloads and resolves Baseten Pointers (bptr) to local directories, supporting HuggingFace, Google Cloud Storage, AWS S3, and Azure sources with optional caching.
Yes, if you are working within the Baseten ecosystem or need to download models from multiple cloud providers at runtime. The package is actively maintained, has no known vulnerabilities, and handles the boilerplate of cloud authentication and file resolution. Install friction is moderate due to compiled wheels, but pre-built binaries are available for common platforms. Not necessary if you only download from a single source or use standard tools like huggingface-hub directly.
Install
truss-transfer on PyPI
pip
pip install truss-transferuv
uv add truss-transferpoetry
poetry add truss-transferInstalling truss-transfer
Before you install
Medium install friction due to compiled wheels across multiple platforms and Python versions. Package is actively maintained with recent releases and no known vulnerabilities.
Quickstart
pip install truss-transfer
import truss_transfer
# Resolve a bptr to a local directory
result_dir = truss_transfer.lazy_data_resolve("/tmp/models")
print(f"Files resolved to: {result_dir}")
Requires Python >= 3.8. Compiled wheels are provided for Linux (x86_64, i686, ppc64le, aarch64, armv7l, musllinux variants) and macOS (x86_64, arm64); Windows support unclear from available wheels.
Verify before relying
- Whether the package requires system libraries or build tools beyond what pip provides
- Performance characteristics and typical download speeds for large model repositories
- Compatibility with Windows systems (wheels shown are Linux/macOS focused)
Package facts
| License | not declared (unclear) |
| Python support | supports the current Python release (>=3.8) |
| Install friction | medium — platform-specific wheel |
| Runtime dependencies | none |
| Maintenance | actively maintained — 72 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 389,417/month — #7,028 on PyPI (30-day window, as of 2026-08-14) |
| Known vulnerabilities | none known (OSV.dev, checked 2026-08-14) |
Evidence: truss_transfer-0.0.43-cp313-cp313t-macosx_10_12_x86_64.whl; truss_transfer-0.0.43-cp313-cp313t-macosx_11_0_arm64.whl; truss_transfer-0.0.43-cp313-cp313t-manylinux_2_17_i686.manylinux2014_i686.whl; truss_transfer-0.0.43-cp313-cp313t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl; truss_transfer-0.0.43-cp313-cp313t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl; truss_transfer-0.0.43-cp313-cp313t-musllinux_1_2_aarch64.whl; truss_transfer-0.0.43-cp313-cp313t-musllinux_1_2_armv7l.whl; truss_transfer-0.0.43-cp313-cp313t-musllinux_1_2_i686.whl; truss_transfer-0.0.43-cp313-cp313t-musllinux_1_2_x86_64.whl; truss_transfer-0.0.43-cp38-abi3-macosx_10_12_x86_64.whl; truss_transfer-0.0.43-cp38-abi3-macosx_11_0_arm64.whl; truss_transfer-0.0.43-cp38-abi3-manylinux_2_17_i686.manylinux2014_i686.whl; truss_transfer-0.0.43-cp38-abi3-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl; truss_transfer-0.0.43-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl; truss_transfer-0.0.43-cp38-abi3-manylinux_2_28_aarch64.whl; truss_transfer-0.0.43-cp38-abi3-manylinux_2_28_armv7l.whl; truss_transfer-0.0.43-cp38-abi3-musllinux_1_2_aarch64.whl; truss_transfer-0.0.43-cp38-abi3-musllinux_1_2_armv7l.whl; truss_transfer-0.0.43-cp38-abi3-musllinux_1_2_i686.whl; truss_transfer-0.0.43-cp38-abi3-musllinux_1_2_x86_64.whl
Tags
More Utilities packages
Converts domain names between Unicode and…
permissive · top 100 on PyPI
charset-normalizerDetects and normalizes text encoding from…
permissive · top 100 on PyPI
setuptoolsSetuptools is a Python build backend and…
permissive · top 100 on PyPI
pluggyPluggy provides a plugin system that lets you…
permissive · top 100 on PyPI
PygmentsPygments is a syntax highlighter that colorizes…
permissive · top 100 on PyPI
sixSix provides utility functions to write Python…
permissive · top 100 on PyPI
cached-pathProvides a unified interface to access local…
permissive · top 15,000 on PyPI
huggingface-hubClient library and CLI tool for downloading,…
permissive · top 100 on PyPI
spacesProvides utilities for working with Hugging…
permissive · top 15,000 on PyPI
pkgutil_resolve_nameProvides a backport of Python 3.9's…
permissive · top 1,000 on PyPI
trussTruss is a CLI tool for packaging ML models…
permissive · top 5,000 on PyPI
hf-transferAccelerates file transfers to and from the…
unclear · top 5,000 on PyPI
langchain-basetenIntegrates Baseten's chat models and embeddings…
permissive · top 15,000 on PyPI
hf-xethf-xet provides chunk-based deduplication and…
permissive · top 1,000 on PyPI
jsonpointerResolves JSON Pointers according to RFC 6901,…
permissive · top 1,000 on PyPI
hfA command-line interface for interacting with…
permissive · top 15,000 on PyPI