monora 1.0.0

Lightweight governance and trace SDK for AI systems

Monora v1 SDK

Lightweight governance and trace SDK for AI systems

Monora provides minimal viable trust through immutable event logs, policy enforcement, and comprehensive auditability for LLM applications.

Features

• 🔒 Immutable Event Logs: Cryptographic hash chains for tamper detection
• 📋 Policy Enforcement: Model allowlists/denylists with data classification controls
• 🧭 Provider Registry: Explicit model-to-provider mapping with unknown-model alerts
• 🧾 Versioned Registry: Registry versioning, history, and provider deprecation metadata
• 🔍 Full Auditability: JSON-lines event logs with CLI reporting tools
• ⚡ Non-blocking: Background worker with bounded queue for zero user-code latency
• 🎯 Simple API: Decorator-based interface with sensible defaults
• 🧩 Auto-Instrumentation: Optional OpenAI/Anthropic patching for drop-in logging
• ✅ Completeness Checks: Event sequencing and security review reports
• 🔌 Pluggable Sinks: Stdout, file, and HTTPS endpoints
• 🚨 Violation Alerts: Callback or webhook notifications for policy violations
• 🧹 Data Handling: Regex redaction rules tied to data classifications
• 🛡️ Signed Attestations: Optional GPG-signed security review bundles

Installation

pip install -e .

# With YAML config support
pip install -e ".[yaml]"

# With HTTPS sink support
pip install -e ".[https]"

# Development dependencies
pip install -e ".[dev]"

Quick Start

Minimal Example (Dev Mode)

import monora

# Initialize with defaults (stdout logging, no policies)
monora.init()

@monora.llm_call(purpose="customer_support")
def ask_gpt(prompt: str, model: str = "gpt-4o-mini"):
    # Your LLM call here
    return {"response": "Hello!"}

# Use trace context for grouping events
with monora.trace("ticket_123"):
    response = ask_gpt("How do I reset my password?")

See README.md for full documentation and examples.

Testing

pytest

License

MIT