| title | Open Redact |
|---|---|
| emoji | 🔒 |
| colorFrom | blue |
| colorTo | indigo |
| sdk | docker |
| app_port | 7860 |
| pinned | false |
open-redact is an open source api to help anonymize pdf. This can be used can be used to redact names and other identifiable information from resume before review to create a more equitable hiring process.
open-redact can redact any PII entity type that Microsoft Presidio supports (people's names, email addresses, phone numbers, credit card numbers, and more). The caller chooses which entity types to redact per request.
This package uses Microsoft Presidio to detect personally identifiable information (PII). Presidio combines regex-based recognizers (for entities such as email addresses) with a spaCy named entity recognition model (for entities such as people's names).
By default the spaCy model is en_core_web_lg, pinned in requirements.txt. You can swap in a smaller model to make development easier, a larger model for more accuracy, or a model for another language. Check out https://spacy.io/models/en for English options.
To use a different model, update the pinned model in requirements.txt and configure Presidio's NLP engine accordingly (see the Presidio customization docs). The default model is loaded automatically by the AnalyzerEngine in app/main/sensitive_text_check.py.
Clone from source and build an image using the included docker file
docker build --tag openredact:python .If not using the image, install dependencies (including the spaCy model) with
pip install -r requirements.txtWhen up and running the system auto generates swagger documentation which can be viewed at http://127.0.0.1:8000/docs#/ where the address and port should be updated for your deployment.
POST /redact_pdf accepts a multipart form with:
file— the PDF to redact (application/pdf).entities— one or more PII entity types to redact. Repeat the field (entities=PERSON&entities=EMAIL_ADDRESS) or pass a comma-separated list (entities=PERSON,EMAIL_ADDRESS). At least one entity type is required; a request with none returns400, as does an unsupported entity type.
GET /entities returns the full list of entity types you can request.
curl -X POST http://127.0.0.1:8000/redact_pdf \
-F "file=@resume.pdf;type=application/pdf" \
-F "entities=PERSON" \
-F "entities=EMAIL_ADDRESS" \
--output redacted.pdfFrom root run the following command to execute all unit tests
python -m pytest .Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.