presidio_demo / presidio_streamlit.py
presidio's picture
updated OpenAI model for synthesis to gpt-3.5-turbo-instruct
f94d289 verified
raw
history blame
12.9 kB
"""Streamlit app for Presidio."""
import logging
import os
import traceback
import dotenv
import pandas as pd
import streamlit as st
import streamlit.components.v1 as components
from annotated_text import annotated_text
from streamlit_tags import st_tags
from openai_fake_data_generator import OpenAIParams
from presidio_helpers import (
get_supported_entities,
analyze,
anonymize,
annotate,
create_fake_data,
analyzer_engine,
)
st.set_page_config(
page_title="Presidio demo",
layout="wide",
initial_sidebar_state="expanded",
menu_items={
"About": "https://microsoft.github.io/presidio/",
},
)
dotenv.load_dotenv()
logger = logging.getLogger("presidio-streamlit")
allow_other_models = os.getenv("ALLOW_OTHER_MODELS", False)
# Sidebar
st.sidebar.header(
"""
PII De-Identification with [Microsoft Presidio](https://microsoft.github.io/presidio/)
"""
)
model_help_text = """
Select which Named Entity Recognition (NER) model to use for PII detection, in parallel to rule-based recognizers.
Presidio supports multiple NER packages off-the-shelf, such as spaCy, Huggingface, Stanza and Flair,
as well as service such as Azure Text Analytics PII.
"""
st_ta_key = st_ta_endpoint = ""
model_list = [
"spaCy/en_core_web_lg",
"flair/ner-english-large",
"HuggingFace/obi/deid_roberta_i2b2",
"HuggingFace/StanfordAIMI/stanford-deidentifier-base",
"stanza/en",
"Azure AI Language",
"Other",
]
if not allow_other_models:
model_list.pop()
# Select model
st_model = st.sidebar.selectbox(
"NER model package",
model_list,
index=2,
help=model_help_text,
)
# Extract model package.
st_model_package = st_model.split("/")[0]
# Remove package prefix (if needed)
st_model = (
st_model
if st_model_package.lower() not in ("spacy", "stanza", "huggingface")
else "/".join(st_model.split("/")[1:])
)
if st_model == "Other":
st_model_package = st.sidebar.selectbox(
"NER model OSS package", options=["spaCy", "stanza", "Flair", "HuggingFace"]
)
st_model = st.sidebar.text_input(f"NER model name", value="")
if st_model == "Azure AI Language":
st_ta_key = st.sidebar.text_input(
f"Azure AI Language key", value=os.getenv("TA_KEY", ""), type="password"
)
st_ta_endpoint = st.sidebar.text_input(
f"Azure AI Language endpoint",
value=os.getenv("TA_ENDPOINT", default=""),
help="For more info: https://learn.microsoft.com/en-us/azure/cognitive-services/language-service/personally-identifiable-information/overview", # noqa: E501
)
st.sidebar.warning("Note: Models might take some time to download. ")
analyzer_params = (st_model_package, st_model, st_ta_key, st_ta_endpoint)
logger.debug(f"analyzer_params: {analyzer_params}")
st_operator = st.sidebar.selectbox(
"De-identification approach",
["redact", "replace", "synthesize", "highlight", "mask", "hash", "encrypt"],
index=1,
help="""
Select which manipulation to the text is requested after PII has been identified.\n
- Redact: Completely remove the PII text\n
- Replace: Replace the PII text with a constant, e.g. <PERSON>\n
- Synthesize: Replace with fake values (requires an OpenAI key)\n
- Highlight: Shows the original text with PII highlighted in colors\n
- Mask: Replaces a requested number of characters with an asterisk (or other mask character)\n
- Hash: Replaces with the hash of the PII string\n
- Encrypt: Replaces with an AES encryption of the PII string, allowing the process to be reversed
""",
)
st_mask_char = "*"
st_number_of_chars = 15
st_encrypt_key = "WmZq4t7w!z%C&F)J"
open_ai_params = None
logger.debug(f"st_operator: {st_operator}")
def set_up_openai_synthesis():
"""Set up the OpenAI API key and model for text synthesis."""
if os.getenv("OPENAI_TYPE", default="openai") == "Azure":
openai_api_type = "azure"
st_openai_api_base = st.sidebar.text_input(
"Azure OpenAI base URL",
value=os.getenv("AZURE_OPENAI_ENDPOINT", default=""),
)
openai_key = os.getenv("AZURE_OPENAI_KEY", default="")
st_deployment_id = st.sidebar.text_input(
"Deployment name", value=os.getenv("AZURE_OPENAI_DEPLOYMENT", default="")
)
st_openai_version = st.sidebar.text_input(
"OpenAI version",
value=os.getenv("OPENAI_API_VERSION", default="2023-05-15"),
)
else:
openai_api_type = "openai"
st_openai_version = st_openai_api_base = None
st_deployment_id = ""
openai_key = os.getenv("OPENAI_KEY", default="")
st_openai_key = st.sidebar.text_input(
"OPENAI_KEY",
value=openai_key,
help="See https://help.openai.com/en/articles/4936850-where-do-i-find-my-secret-api-key for more info.",
type="password",
)
st_openai_model = st.sidebar.text_input(
"OpenAI model for text synthesis",
value=os.getenv("OPENAI_MODEL", default="gpt-3.5-turbo-instruct"),
help="See more here: https://platform.openai.com/docs/models/",
)
return (
openai_api_type,
st_openai_api_base,
st_deployment_id,
st_openai_version,
st_openai_key,
st_openai_model,
)
if st_operator == "mask":
st_number_of_chars = st.sidebar.number_input(
"number of chars", value=st_number_of_chars, min_value=0, max_value=100
)
st_mask_char = st.sidebar.text_input(
"Mask character", value=st_mask_char, max_chars=1
)
elif st_operator == "encrypt":
st_encrypt_key = st.sidebar.text_input("AES key", value=st_encrypt_key)
elif st_operator == "synthesize":
(
openai_api_type,
st_openai_api_base,
st_deployment_id,
st_openai_version,
st_openai_key,
st_openai_model,
) = set_up_openai_synthesis()
open_ai_params = OpenAIParams(
openai_key=st_openai_key,
model=st_openai_model,
api_base=st_openai_api_base,
deployment_id=st_deployment_id,
api_version=st_openai_version,
api_type=openai_api_type,
)
st_threshold = st.sidebar.slider(
label="Acceptance threshold",
min_value=0.0,
max_value=1.0,
value=0.35,
help="Define the threshold for accepting a detection as PII. See more here: ",
)
st_return_decision_process = st.sidebar.checkbox(
"Add analysis explanations to findings",
value=False,
help="Add the decision process to the output table. "
"More information can be found here: https://microsoft.github.io/presidio/analyzer/decision_process/",
)
# Allow and deny lists
st_deny_allow_expander = st.sidebar.expander(
"Allowlists and denylists",
expanded=False,
)
with st_deny_allow_expander:
st_allow_list = st_tags(
label="Add words to the allowlist", text="Enter word and press enter."
)
st.caption(
"Allowlists contain words that are not considered PII, but are detected as such."
)
st_deny_list = st_tags(
label="Add words to the denylist", text="Enter word and press enter."
)
st.caption(
"Denylists contain words that are considered PII, but are not detected as such."
)
# Main panel
with st.expander("About this demo", expanded=False):
st.info(
"""Presidio is an open source customizable framework for PII detection and de-identification.
\n\n[Code](https://aka.ms/presidio) |
[Tutorial](https://microsoft.github.io/presidio/tutorial/) |
[Installation](https://microsoft.github.io/presidio/installation/) |
[FAQ](https://microsoft.github.io/presidio/faq/) |
[Feedback](https://forms.office.com/r/9ufyYjfDaY) |"""
)
st.info(
"""
Use this demo to:
- Experiment with different off-the-shelf models and NLP packages.
- Explore the different de-identification options, including redaction, masking, encryption and more.
- Generate synthetic text with Microsoft Presidio and OpenAI.
- Configure allow and deny lists.
This demo website shows some of Presidio's capabilities.
[Visit our website](https://microsoft.github.io/presidio) for more info,
samples and deployment options.
"""
)
st.markdown(
"[![Pypi Downloads](https://img.shields.io/pypi/dm/presidio-analyzer.svg)](https://img.shields.io/pypi/dm/presidio-analyzer.svg)" # noqa
"[![MIT license](https://img.shields.io/badge/license-MIT-brightgreen.svg)](https://opensource.org/licenses/MIT)"
"![GitHub Repo stars](https://img.shields.io/github/stars/microsoft/presidio?style=social)"
)
analyzer_load_state = st.info("Starting Presidio analyzer...")
analyzer_load_state.empty()
# Read default text
with open("demo_text.txt") as f:
demo_text = f.readlines()
# Create two columns for before and after
col1, col2 = st.columns(2)
# Before:
col1.subheader("Input")
st_text = col1.text_area(
label="Enter text", value="".join(demo_text), height=400, key="text_input"
)
try:
# Choose entities
st_entities_expander = st.sidebar.expander("Choose entities to look for")
st_entities = st_entities_expander.multiselect(
label="Which entities to look for?",
options=get_supported_entities(*analyzer_params),
default=list(get_supported_entities(*analyzer_params)),
help="Limit the list of PII entities detected. "
"This list is dynamic and based on the NER model and registered recognizers. "
"More information can be found here: https://microsoft.github.io/presidio/analyzer/adding_recognizers/",
)
# Before
analyzer_load_state = st.info("Starting Presidio analyzer...")
analyzer = analyzer_engine(*analyzer_params)
analyzer_load_state.empty()
st_analyze_results = analyze(
*analyzer_params,
text=st_text,
entities=st_entities,
language="en",
score_threshold=st_threshold,
return_decision_process=st_return_decision_process,
allow_list=st_allow_list,
deny_list=st_deny_list,
)
# After
if st_operator not in ("highlight", "synthesize"):
with col2:
st.subheader(f"Output")
st_anonymize_results = anonymize(
text=st_text,
operator=st_operator,
mask_char=st_mask_char,
number_of_chars=st_number_of_chars,
encrypt_key=st_encrypt_key,
analyze_results=st_analyze_results,
)
st.text_area(
label="De-identified", value=st_anonymize_results.text, height=400
)
elif st_operator == "synthesize":
with col2:
st.subheader(f"OpenAI Generated output")
fake_data = create_fake_data(
st_text,
st_analyze_results,
open_ai_params,
)
st.text_area(label="Synthetic data", value=fake_data, height=400)
else:
st.subheader("Highlighted")
annotated_tokens = annotate(text=st_text, analyze_results=st_analyze_results)
# annotated_tokens
annotated_text(*annotated_tokens)
# table result
st.subheader(
"Findings"
if not st_return_decision_process
else "Findings with decision factors"
)
if st_analyze_results:
df = pd.DataFrame.from_records([r.to_dict() for r in st_analyze_results])
df["text"] = [st_text[res.start : res.end] for res in st_analyze_results]
df_subset = df[["entity_type", "text", "start", "end", "score"]].rename(
{
"entity_type": "Entity type",
"text": "Text",
"start": "Start",
"end": "End",
"score": "Confidence",
},
axis=1,
)
df_subset["Text"] = [st_text[res.start : res.end] for res in st_analyze_results]
if st_return_decision_process:
analysis_explanation_df = pd.DataFrame.from_records(
[r.analysis_explanation.to_dict() for r in st_analyze_results]
)
df_subset = pd.concat([df_subset, analysis_explanation_df], axis=1)
st.dataframe(df_subset.reset_index(drop=True), use_container_width=True)
else:
st.text("No findings")
except Exception as e:
print(e)
traceback.print_exc()
st.error(e)
components.html(
"""
<script type="text/javascript">
(function(c,l,a,r,i,t,y){
c[a]=c[a]||function(){(c[a].q=c[a].q||[]).push(arguments)};
t=l.createElement(r);t.async=1;t.src="https://www.clarity.ms/tag/"+i;
y=l.getElementsByTagName(r)[0];y.parentNode.insertBefore(t,y);
})(window, document, "clarity", "script", "h7f8bp42n8");
</script>
"""
)