# =============================================================================
# Custom Presidio Analyzer image — English + Italian NER (privacy-plus C1).
# -----------------------------------------------------------------------------
# Adds the Italian spaCy NER model to the community analyzer so the privacy-plus
# C1 detector can score Italian text natively, instead of the English-only stock
# image mis-firing on Italian (the "ciao" -> PERSON false positive).
#
# Presidio + spaCy models are community open-source. The models are baked at
# build time -> NO runtime download, NO external egress (sovereignty premise).
#
# This lab ships it_core_news_md (RAM-friendly on CPU-only clusters); the
# it_core_news_lg upgrade is a swap here + in nlp-config.yaml + a memory bump.
#
# Prebuilt: quay.io/asalvati/presidio-analyzer-multilang:0.1
# =============================================================================
FROM ghcr.io/data-privacy-stack/presidio-analyzer:2.2.363

USER 0

# Italian NER model, installed alongside the bundled English en_core_web_lg.
# NB: if the base image uses a virtualenv, ensure `python` targets it.
RUN python -m spacy download it_core_news_md

# Custom NLP engine config declaring BOTH languages. app.py reads NLP_CONF_FILE at
# runtime and builds the AnalyzerEngine with supported_languages = [en, it].
COPY nlp-config.yaml /opt/presidio/nlp-config.yaml
ENV NLP_CONF_FILE=/opt/presidio/nlp-config.yaml

# Drop back to non-root. On OpenShift the SCC assigns an arbitrary UID anyway; spaCy
# models install world-readable into site-packages, so any UID can read them.
USER 1000