Hawk-Eye

What it is

Hawk-Eye (installed as the hawk-scanner / hawk_scanner package) is a Python CLI for scanning data sources for PII and secrets. Beyond the capability list in the summary, the page documents concrete operational details an adopter needs before committing: install-time dependency quirks, a default-behavior gotcha, and a Python-library embedding path.

How to run it

Each source has its own subcommand (fs, s3, gcs, mysql, redis, firebase, slack, postgresql, couchdb, gdrive, text). Two YAML files drive it: connection.yml (source profiles + Slack notify block with redacted and suppress_duplicates options) and fingerprint.yml (regex fingerprints). Typical run:

hawk_scanner all --connection connection.yml --fingerprint fingerprint.yml --json output.json --debug

For CI/CD, connection config can be passed inline instead of by file, with --stdout --quiet:

hawk_scanner fs --connection-json '{"sources": {"fs": {"fs1": {"quick_scan": true, "path": "/path/KYC_PDF.pdf"}}}}' --stdout --quiet --fingerprint fingerprint.yml

It can also be embedded in Python, optionally with inline fingerprints:

from hawk_scanner.internals import system
pii = system.scan_file("/path/Resume.pdf")
print(pii)

Preconditions / gotchas

• Default runs everything. If you invoke it with no subcommand, it runs all scanners (firebase, fs, gcs, mysql, text, couchdb, gdrive, gdrive workspace, slack, postgresql, redis) by default — scope your command explicitly.
• PostgreSQL needs a manual dependency. psycopg2-binary is not shipped with the package (author cites Windows incompatibility): pip3 install psycopg2-binary.
• RedHat/OCR dependency. The cv2 dependency can error on RedHat until you install yum install mesa-libGL.
• Docker usage requires mounting both connection.yml and fingerprint.yml into /app and passing --platform linux/amd64.

Install

pip3 install hawk-scanner

Or from source: clone, pip3 install -r requirements.txt, then python3 hawk_scanner/main.py.

Caveats

Commercial support is offered via the author's LinkedIn/Twitter/Slack rather than a formal channel. The page is largely README/marketing prose; the feature list matches the summary, and no versioning, benchmark, or accuracy data is given for the "advanced algorithms" / OCR detection it claims.