Diogenesis SDK

Runtime security for AI applications. Monitors what your AI does — every import, file write, network connection, and subprocess — and alerts when behavior deviates from normal.

Zero dependencies. Pure Python. Works with any AI framework.

Install

pip install diogenesis-sdk

Quick Start

from diogenesis_sdk import activate, status, threat_summary

# Turn on monitoring
activate()

# Check what's happening
print(status())

If it worked, you'll see something like:

{'active': True, 'interceptors': 4, 'total_events': 1200, 'field_coherence': 1.0, ...}

That means Diogenesis is watching every import, file access, network call, and subprocess.

What Can It Do?

from diogenesis_sdk import activate, status, field_state, threat_summary

activate()

# System health — are interceptors running?
print(status())

# Behavioral voltage — per-module coherence scores
print(field_state())

# Threat analysis — what did the xenobot agents find?
print(threat_summary())

Example output:

Threats: 0 | Suspicious: 0
Agents: ['import_guardian', 'file_sentinel', 'network_watcher', 'subprocess_monitor']

Four agents are patrolling your application automatically.

Add Custom Threat Patterns

from diogenesis_sdk import activate, add_pattern, BehavioralPattern

activate()

# Detect: file read followed by network send (possible data leak)
pattern = BehavioralPattern(
    name="api_key_leak",
    description="Sensitive file read followed by outbound network call",
    event_sequence=[
        {"type": "file", "detail_contains": {"path": ".env"}},
        {"type": "network", "classification": "UNEXPECTED"},
    ],
    window_seconds=30,
    severity="CRITICAL",
)
add_pattern(pattern)

Create Custom Agents

from diogenesis_sdk import activate, XenobotAgent, add_agent, investigations

activate()

# Create an agent that specializes in file monitoring
agent = XenobotAgent("file_patrol", domain="file")
add_agent(agent)

# Check investigation findings
print(investigations())

Each agent uses a 5-phase investigation cycle: Observe, Question, Search, Synthesize, Crystallize.

Behavioral Voltage Field

Every module gets a "voltage" score — how closely its behavior matches baseline. Voltage drops before attacks complete.

from diogenesis_sdk import activate, field_state

activate()

state = field_state()
for name, info in state["modules"].items():
    voltage = info["voltage"]
    if voltage < 0.5:
        print(f"WARNING: {name} coherence low ({voltage:.2f})")
    else:
        print(f"OK: {name} stable at {voltage:.2f}")

Core Components

Features

• Zero dependencies — pure Python standard library only
• 104 automated tests — production-grade
• Python 3.8+ — works with any AI framework
• Behavioral baseline — learns "normal", alerts on deviation
• Graduated response — LOG → WARN → ALERT
• 5 built-in threat patterns — exfiltration, import chains, write bursts, network scans, shadow imports

Troubleshooting

"ModuleNotFoundError: No module named 'diogenesis_sdk'" Run: pip install diogenesis-sdk (note the hyphen, not underscore)

"ImportError: cannot import name 'activate'" Make sure you have version 0.2.0+: pip install --upgrade diogenesis-sdk

License

Apache License 2.0. Free for commercial and personal use.

Links

• Website: diogenicsecurity.com
• Source: github.com/AI-World-CEO/diogenesis-sdk
• PyPI: pypi.org/project/diogenesis-sdk

Built by Garry Anderson. Behavioral security for the age of autonomous AI.