Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

DataFog

DataFog Core

Find and transform sensitive data in your application.

PyPI version npm version crates.io version Rust CI status MIT license

Quickstart · Detection · Transformations · Local models · Documentation

DataFog Core is a Rust library with Python, Node.js, and browser WebAssembly bindings. Scan text or JSON for personal data and credentials, then redact, mask, remove, pseudonymize, or tokenize matched values. Findings use a consistent format across the bindings.

Rule-based scanning runs locally without model weights or downloads. Optional local model sessions add DataFog and GLiNER2 support in Rust, Python, and Node.js.

Stable: 0.4.4 · Published prerelease: 0.5.0-beta.2

0.5.0-beta.3 release candidate adds native workspace APIs, operation/conversation records and strict/progressive streaming. Publication and fresh registry verification are pending; see the beta.3 integration guide.

Quickstart

Choose the package for your application:

SDK Install stable Requirements
Python python -m pip install datafog-core Python 3.10+
Node.js npm install @datafog/node Node.js 24.x
Rust cargo add datafog-core Rust 1.88+
Browser / WASM npm install @datafog/wasm Browser with WebAssembly support

Python

from datafog_core import scan, scan_and_transform

findings = scan("Email jane@example.com")
print(findings[0].entity_type)   # EMAIL
print(findings[0].matched_text)  # jane@example.com

result = scan_and_transform(
    "Email jane@example.com",
    {"transform": {"default": {"strategy": "redact"}}},
)
print(result.text)  # Email [EMAIL]
Node.js example
import { scan, scanAndTransform } from "@datafog/node";

const findings = scan("Email jane@example.com");
console.log(findings[0].entityType); // EMAIL
console.log(findings[0].matchedText); // jane@example.com

const result = scanAndTransform("Email jane@example.com", {
  transform: { default: { strategy: "redact" } },
});
console.log(result.text); // Email [EMAIL]
Rust example
use datafog_core::{
    scan, scan_and_transform, PrivacyError, ScanAndTransformConfig,
    TransformationConfig, TransformationStrategy,
};

fn main() -> Result<(), PrivacyError> {
    let findings = scan("Email jane@example.com");
    println!("{}", findings[0].entity_type); // EMAIL
    println!("{}", findings[0].matched_text); // jane@example.com

    let result = scan_and_transform(
        "Email jane@example.com",
        &ScanAndTransformConfig::new(TransformationConfig::new(
            TransformationStrategy::Redact,
        )),
    )?;
    println!("{}", result.text); // Email [EMAIL]
    Ok(())
}
Browser / WASM example
import { init, scan, scanAndTransform } from "@datafog/wasm";

await init();
console.log(scan("Email jane@example.com"));

const result = scanAndTransform("Email jane@example.com", {
  transform: { default: { strategy: "redact" } },
});
console.log(result.text); // Email [EMAIL]

Python and Node.js packages include prebuilt native binaries for Linux x64 and ARM64 (GNU), macOS Intel and Apple Silicon, and Windows x64 (MSVC).

Detection coverage

Coverage Entity types Activation
Personal and financial data EMAIL, PHONE, SSN, CREDIT_CARD, IP_ADDRESS, DATE, ZIP_CODE On by default
Credentials API_KEY, BEARER_TOKEN, JWT, PRIVATE_KEY, CREDENTIAL_URI On by default
Labeled street addresses STREET_ADDRESS On by default; bounded English street-line rule
Labeled accounts ACCOUNT_NUMBER On by default; bounded lexical rule
Labeled US identifiers US_ROUTING_NUMBER, NPI On by default; requires a supported text label
German identifiers IBAN, VAT ID, tax ID, social security number, postal code, passport number, residence permit number Explicit German locale
UUIDs UUID Explicit detect_uuid setting
Names in JSON fields PERSON Structured scans, using name-field aliases or explicit JSON Pointer mappings

Use structured scans to find names by JSON field, or local models to detect names and unlabelled street addresses in prose.

Each finding includes its entity type, matched text, byte and code-point ranges, optional confidence, and detector name/version. Ranges are zero-based and end-exclusive. Node.js and WASM also provide UTF-16 ranges for JavaScript slice. Rule-based findings have no confidence score. Findings and ranges →

For scanning files from AI agents, use DataFog MCP.

Detection settings
  • Select checks: scan(text, {"entities": ["EMAIL", "SSN"]}) in Python runs only the selected text detectors. Omit entities to use all enabled detectors. Selections must be nonempty, unique, exact entity names. Selecting a German type or UUID does not activate it; set locale or detect_uuid too.
  • German locale: {"locale": "de"} enables the seven German detectors. These use format/context rules; IBAN checksums and account existence are not verified. See the German reference for detector rules.
  • ENV and SQL: choose format: "env" or format: "sql" to handle email boundaries in assignments and quoted strings. Source-context rules →
  • Phone numbers: numbers without + use North American checks: ten digits, optionally preceded by 1, with area and exchange codes starting with 2–9. Numbers starting with + use a permissive 7–15 digit check, including shortened examples such as +1-555-1002. Country codes and assigned numbers are not validated; numeric identifiers can still match.
  • Decimal numbers and CSV: contiguous card digits after a digit and . or , are suppressed, so 3.4111111111111111 and 3,4111111111111111 do not produce card findings. Cards containing spaces or hyphens still undergo normal checks. Parse CSV/TSV into cells before scanning so delimiters and adjacent fields do not affect matching.

Use capabilities() in Rust or Python to discover supported entities and their activation rules.

Transformations

Strategy Result Application provider Browser / WASM
redact Replace with a label such as [EMAIL] None Supported
mask Replace characters, optionally revealing the first or last characters None Supported
remove Delete the matched span None Supported
pseudonymize Create a stable keyed HMAC-SHA-256 pseudonym Key provider Unsupported
tokenize Issue an opaque token that can be restored under the same scope Token provider Unsupported

Use transform with existing findings, or scan_and_transform / scanAndTransform to scan and transform together. Transformation records include source/output ranges and replacement metadata, without the original matched text.

Keep scan settings under scan and transformation policy under transform. A default strategy is required; entity selection, per-type overrides, and allowlists are optional. Configuration guide → · Strategy details →

Chrome extension integration in 0.5.0-beta.1

The developer prerelease adds persistent Workspace pseudonyms, approved save-before-handoff, strict restoration, follow-ups and host-owned encrypted backup/recovery, plus optional English GLiNER2.5 Small browser sessions. Use synthetic data while the snapshot format and vault integration remain provisional.

npm install --save-exact @datafog/wasm@0.5.0-beta.1

Integration lifecycle, all browser calls and errors → · Model assets and runtime setup → · Release scope and verification →

Local models (beta)

0.4.5-beta.4 adds local sessions for the DataFog PII model and GLiNER2 in Rust, Python, and Node.js on all five native platforms.

Install 0.5.0-beta.1 (Python 0.5.0b1) for the current native prerelease:

python -m pip install 'datafog-core==0.5.0b1'
npm install --save-exact @datafog/node@0.5.0-beta.1
cargo add datafog-core@0.5.0-beta.1 --features models

For GLiNER2, install its optional dependencies in the Python environment used by the runner:

python -m pip install 'datafog-core[gliner2]==0.5.0b1'

GLiNER2 supports the PII checkpoint and compatible general-purpose span models with caller-defined labels. The GLiNER2 runner ships in the Python package and also serves Node.js and Rust callers. Download and verify model files separately, then use ModelSession.open, text scan, and close. Reuse sessions across calls. Both backends return consistent results and errors, with cancellation and cleanup support.

DataFog setup and session API → · GLiNER2 setup →

Documentation and development

For existing datafog Python applications, follow the migration guide for changes to imports, results, and transformations.

For Rust development, run:

cargo fmt --all --check
cargo clippy --workspace --all-targets --all-features -- -D warnings
cargo test --workspace --all-features

See the development guide for the repository layout and Python, Node.js, and WASM package checks.

License

MIT. Built by DataFog.

Metadata

Release files for datafog-core 0.5.0b3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for datafog-core 0.5.0b3
File Size Uploaded Provenance Actions
datafog_core-0.5.0b3.tar.gz 297.9 kB With provenance Details for datafog_core-0.5.0b3.tar.gz

Built distributions (wheels)

Table of built distributions (wheels) for datafog-core 0.5.0b3
File Actions
datafog_core-0.5.0b3-cp310-abi3-win_amd64.whl CPython 3.10 abi3 Windows x86-64 Details for datafog_core-0.5.0b3-cp310-abi3-win_amd64.whl
datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_x86_64.whl CPython 3.10 abi3 Linux glibc 2.28+ x86-64 Details for datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_x86_64.whl
datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_aarch64.whl CPython 3.10 abi3 Linux glibc 2.28+ ARM64 Details for datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_aarch64.whl
datafog_core-0.5.0b3-cp310-abi3-macosx_11_0_arm64.whl CPython 3.10 abi3 macOS 11.0+ ARM64 Details for datafog_core-0.5.0b3-cp310-abi3-macosx_11_0_arm64.whl
datafog_core-0.5.0b3-cp310-abi3-macosx_10_12_x86_64.whl CPython 3.10 abi3 macOS 10.12+ x86-64 Details for datafog_core-0.5.0b3-cp310-abi3-macosx_10_12_x86_64.whl

Total release size: 12.0 MB

Release files / datafog_core-0.5.0b3.tar.gz

Download URL datafog_core-0.5.0b3.tar.gz
Size 297.9 kB
Tags Source
SHA-256 checksum
How to use checksums
5d966a827b4f16474158fcd2e4e3e23037d4706f2161058441d71e77b22baf86
BLAKE2b-256 checksum
How to use checksums
c3f3f8784aae3cbf53d6731f5aff661d27fed1952036b19b026baf5f3576f48f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / datafog_core-0.5.0b3-cp310-abi3-win_amd64.whl

Download URL datafog_core-0.5.0b3-cp310-abi3-win_amd64.whl
Size 2.1 MB
Tags CPython 3.10 Windows x86-64 abi3
SHA-256 checksum
How to use checksums
e1c58a8cbafb8c371573243475167200d54016852afdb96f67026919c92443e1
BLAKE2b-256 checksum
How to use checksums
fc248c0a0d423613e7c784c83b7d71aff71fb47bc92893b598ba5ad50e7be2bf
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_x86_64.whl

Download URL datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_x86_64.whl
Size 2.6 MB
Tags CPython 3.10 Linux glibc 2.28+ x86-64 abi3
SHA-256 checksum
How to use checksums
4e8c42f8cd2a251d92005e40ea59cbf6629789138cf03f124e5dcf827197677f
BLAKE2b-256 checksum
How to use checksums
358f8e73e486f7bb88ee5b6ef4f8aa02b49fe4ba59a29af859a632e85e860735
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_aarch64.whl

Download URL datafog_core-0.5.0b3-cp310-abi3-manylinux_2_28_aarch64.whl
Size 2.5 MB
Tags CPython 3.10 Linux glibc 2.28+ ARM64 abi3
SHA-256 checksum
How to use checksums
c7b68c719dcea1d5fc78a43de18fb76e1abcac4226cba1b6bee119ac0fb04ac6
BLAKE2b-256 checksum
How to use checksums
3b820d881dd1b33e513c44851f3ad97b3b00561604b654b630480d25dd475ac2
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / datafog_core-0.5.0b3-cp310-abi3-macosx_11_0_arm64.whl

Download URL datafog_core-0.5.0b3-cp310-abi3-macosx_11_0_arm64.whl
Size 2.2 MB
Tags CPython 3.10 abi3 macOS 11.0+ ARM64
SHA-256 checksum
How to use checksums
6b10f66c95d4a20520033bab1733cef0c018753d649ec753cf28582c92edb12e
BLAKE2b-256 checksum
How to use checksums
71414bf2c23c22b915a3f764ba7f4a40ed6a70c07c871e87fd6e5d78900d8fb7
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Release files / datafog_core-0.5.0b3-cp310-abi3-macosx_10_12_x86_64.whl

Download URL datafog_core-0.5.0b3-cp310-abi3-macosx_10_12_x86_64.whl
Size 2.3 MB
Tags CPython 3.10 abi3 macOS 10.12+ x86-64
SHA-256 checksum
How to use checksums
a22626f77dbf02570741eab27e7ed9321f32fc9b9d3d42bbcec3cff6751278ef
BLAKE2b-256 checksum
How to use checksums
dc95f9ee546879acfd9325d1840ccc5c23b9df0aa4c28c5e12c0c5ad4be3ceb9
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

Security for datafog-core 0.5.0b3

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 6 files in this release have been uploaded with provenance metadata.

Select from 6 files

Attestation for datafog_core-0.5.0b3.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 29, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page