Skip to content

QUICKSTART.md

Latest commit

 

History

History
61 lines (47 loc) · 1.24 KB

QUICKSTART.md

File metadata and controls

61 lines (47 loc) · 1.24 KB

Quick Start with PIIgent

Installation

# Clone the repository
git clone https://github.com/shokrydev/piigent.git
cd piigent

# Sync the workspace (creates .venv and installs all dependencies)
uv sync

# Pull the required LLM for agentic detection
ollama pull ministral-3:8b

You can run the standard PII detection flow:

uv run demo/run_agentic_flow.py

Code example:

from graph.privacy_flow import run_flow

result = run_flow(
    document="""
    Entlassungsbrief - Charité Berlin
    Patient: Max Mustermann
    KVNR: A123456780
    PLZ: 10117 Berlin
    """,
    confidence_threshold=0.7,
    human_in_loop=False,
    preset='clinical',
)

print(result['anonymized_text'])

To explore where the flow fails and generate targeted test cases:

uv run demo/run_weakness_analysis.py

Code example:

from agents.core.weakness_analyzer import analyze_weaknesses

report = analyze_weaknesses(
    num_docs=20,
    max_rounds=3,
    enable_evolution=True,  # Automatically evolve prompts based on weaknesses
)

print(report["recommendations"])
print(report["evolution"]["mutations_applied"])  # Number of prompt mutations applied
print(report["evolution"]["evolved_genome_id"])  # ID of the improved genome