Black-Bar

Review and remove sensitive information from documents on your computer.

Free and open source. Documents stay on your machine.

ND-2026-014

Nondisclosure Agreement

This Nondisclosure Agreement is between the National Institute of Standards and Technology (“NIST”) and Meridian Research, Inc. (“Recipient”) for the protection of proprietary information disclosed by NIST.

2. Recipient’s representative for receiving Proprietary Information: Alexandra R. Bennett. Recipient shall not disclose the Proprietary Information to any of its employees other than those who have a need to review it and which employees are legally obligated to honor the confidentiality provisions herein.

4. Recipient shall not make any copies of the Proprietary Information except as necessary for its employees who are entitled to review it under Section 2 above. Any copies which are made shall be identified as belonging to NIST and marked “confidential” or with a similar legend.

NIST model NDA · Illustrative recipient details

You decide what goes

Accept or reject suggestions, adjust redaction boxes, and add your own.

Remove the content

Accepted text and image pixels are removed. Hidden metadata is scrubbed.

Check the export

Independent extraction checks the result. Recoverable accepted content blocks export.

Review the document before it leaves

Open walkthrough

Current Black-Bar review app displaying the full NIST nondisclosure agreement, reviewed detections, and the redacted PDF preview — The current local app, reviewing a NIST model NDA with illustrative entries.

Review each suggestion

Accept or reject spans, adjust boxes, or redact every occurrence of a term.

Inspect the output

Check the redacted document beside the original before exporting.

Keep the record

A private manifest records file hashes, review decisions, and verification results.

Local detection, with room for your judgment

Find names, contact details, amounts, dates, and identifiers. Add your own watchlist; checksums validate supported ID types.

Default: regex, contract structure, and spaCy NER. Optional Presidio or GLiNER adds deeper and multilingual detection locally.

Scans and embedded imagesLocal OCR handles document images as well as text.
Readable replacementsUse consistent fictional replacements when the document needs to remain readable.
Extraction check before exportAccepted text and image pixels are removed, metadata is scrubbed, and extraction or OCR checks the result.
Private by defaultNo accounts, telemetry, or cloud processing. Keep manifests and replacement legends private.

Review app or command line

Codeberg

Use the review app for individual documents. The same pipeline supports batch and scripted runs.

Feature	Review app	Command line
Start	`redactor serve`	`redactor redact file`
Decisions	Accept or reject each suggestion	Accepts all detections
Best for	Careful individual review	Batch and scripted runs
Verification	Extraction-safety gate	Same extraction-safety gate

Run locally

Allow about 15 minutes for setup. Install once, then open the review workspace in your browser.

Open install guide

Prefer the redacted PDF for cloud AI. Keep the manifest and replacement legend private.

Setup commands · macOS / Linux

git clone https://codeberg.org/russkysong/black-bar.git
cd black-bar
python3 -m venv .venv
.venv/bin/pip install -e '.[dev]'
.venv/bin/python -m spacy download en_core_web_md
cd frontend && npm install && npm run build && cd ..
.venv/bin/redactor doctor
.venv/bin/redactor serve

Open http://127.0.0.1:8000 after the server starts.

Documents: PDF, Word, PowerPoint, Excel, HTML, Markdown, images
Processing: Local detection and redaction. Optional local models; no cloud processing.
Requirements: Python 3.10+, Node 18+, LibreOffice for Office proofs. Tesseract for scans.
Other uses: Command line and batch processing