@huggingface/transformers
State-of-the-art Machine Learning for the web. Run 🤗 Transformers directly in your browser, with no need for a server!
Supply chain provenance
Status for the latest visible version.
Without SLSA provenance there is no cryptographic link between this tarball and the public source — the axios compromise (March 2026) relied on exactly this gap.
Maintainers
Keywords
Accepted risks
Findings the reviewer chose to accept rather than block on.
| Source | Rule | Reason | Accepted by | When |
|---|---|---|---|---|
| provenance | missing-githead | AI (provenance): Pre-release major version (4.0.0-next.x) from trusted HuggingFace publisher xenova with clean track record; likely a CI/CD pipeline change during major version development. | ai | |
| publish-pattern | new-deps-added | AI (publish-pattern): @huggingface/tokenizers is a first-party Hugging Face dependency, expected for this package's v4 major version. Not a supply-chain risk. | ai | |
| source-diff | obfuscated-file:dist/transformers.web.js | AI (source-diff): Standard webpack bundle output for a large ML library. Webpack comments and external module declarations are clearly visible in the sample. Expected artifact of the documented build process. | ai | |
| source-diff | large-new-source-files | AI (source-diff): Large ML library ships pre-built webpack bundles and source maps for browser/Node targets. New files are expected dist artifacts, not injected code. | ai | |
| source-diff | obfuscated-file:dist/transformers.node.mjs | AI (source-diff): Standard webpack bundle output for Node.js target. Sample shows recognizable webpack module structure with legitimate imports (fs, onnxruntime-node, path, sharp, url). | ai | |
| source-diff | net-exec-file:dist/ort-wasm-simd-threaded.jsep.mjs | AI (source-diff): Emscripten-compiled ONNX Runtime WASM loader. Network+exec pattern is WASM binary fetch and WebAssembly.instantiate — standard and expected for any WASM-based ML runtime. | ai | |
| provenance | no-provenance | AI (provenance): Already marked accepted risk; Hugging Face/xenova is a well-established publisher with a strong track record. | ai |
Versions (showing 38 of 38)
| Version | Deps | Published |
|---|---|---|
| 4.2.0 | 5 / 8 | |
| 4.1.0 | 5 / 8 | |
| 4.0.1 | 5 / 8 | |
| 4.0.0 | 5 / 8 | |
| 3.8.1 | 4 / 13 | |
| 3.8.0 | 4 / 13 | |
| 3.7.6 | 4 / 13 | |
| 3.7.5 | 4 / 13 | |
| 3.7.4 | 4 / 13 | |
| 3.7.3 | 4 / 13 | |
| 3.7.2 | 4 / 13 | |
| 3.7.1 | 4 / 13 | |
| 3.7.0 | 4 / 13 | |
| 3.6.3 | 4 / 13 | |
| 3.6.2 | 4 / 13 | |
| 3.6.1 | 4 / 13 | |
| 3.6.0 | 4 / 13 | |
| 3.5.2 | 4 / 13 | |
| 3.5.1 | 4 / 13 | |
| 3.5.0 | 4 / 13 | |
| 3.4.2 | 4 / 13 | |
| 3.4.1 | 4 / 13 | |
| 3.4.0 | 4 / 13 | |
| 3.3.3 | 4 / 13 | |
| 3.3.2 | 4 / 13 | |
| 3.3.1 | 4 / 13 | |
| 3.3.0 | 4 / 13 | |
| 3.2.4 | 4 / 13 | |
| 3.2.3 | 4 / 13 | |
| 3.2.2 | 4 / 13 | |
| 3.2.1 | 4 / 13 | |
| 3.2.0 | 4 / 13 | |
| 3.1.2 | 4 / 13 | |
| 3.1.1 | 4 / 12 | |
| 3.1.0 | 4 / 12 | |
| 3.0.2 | 4 / 12 | |
| 3.0.1 | 4 / 12 | |
| 3.0.0 | 4 / 12 |
v4.2.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v4.1.0
2 findingsThis version has no gitHead field linking it to a source commit, but previous versions did. This suggests the publish environment changed. Published by: xenova.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v4.0.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v4.0.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.8.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.8.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.7.6
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.5
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.4
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.3
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.7.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.6.3
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.6.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.6.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.6.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.5.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.5.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.5.0
4 findingsNewly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
Newly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
Newly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.4.2
4 findingsNewly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
Newly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
Newly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.4.1
4 findingsNewly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
Newly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
Newly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.4.0
4 findingsNewly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
Newly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
Newly added source file contains lines over 3000 chars, suggesting minified or obfuscated code. New obfuscated files are a strong attack indicator.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.3.3
2 findingsNewly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.3.2
2 findingsNewly added file contains both network calls and dynamic code execution. This is a hallmark of dropper/loader malware.
[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.3.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.3.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Consider requesting the maintainer enable provenance via CI/CD.
v3.2.4
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.2.3
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.2.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.2.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.2.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.1.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.1.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.1.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.0.2
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.0.1
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.
v3.0.0
1 finding[Accepted risk] Package was published without Sigstore provenance. Only ~12% of npm packages have provenance, so this is common but not ideal.