volidator-python 0.1.0

Zero-knowledge, server-blind audit logging for Python and agentic workflows.

Volidator Python SDK

Zero-knowledge, server-blind audit logging for Python and autonomous AI agentic workflows.

Volidator encrypts every log entry locally — on your server — before sending it. The Volidator backend stores only ciphertext and blind indexes. Your plaintext data never leaves your infrastructure unencrypted.

---

Installation

pip install volidator-python

---

Environment Setup

Generate a secure 32-byte encryption key (64 hex characters) with the vol-dek- prefix:

python -c "import os; print('vol-dek-' + os.urandom(32).hex())"

Add secrets to your environment:

VOLIDATOR_API_KEY="val_live_xxxxxxxx..."
VOLIDATOR_ENCRYPTION_KEY="vol-dek-xxxxxxxx..."

---

Quick Start

import os
from volidator import VolidatorClient

volidator = VolidatorClient(
    api_key=os.environ["VOLIDATOR_API_KEY"],
    encryption_key=os.environ["VOLIDATOR_ENCRYPTION_KEY"]
)

# Log a system event
volidator.log({
    "actor": "usr_12345",
    "action": "user.login.success",
    "target": "workspace_abc789",
    "metadata": {
        "device": "Workstation",
        "auth_provider": "Okta"
    }
})

---

AI Agent Auditing (VolidatorAgent)

For auditing LLM chains, agent reasoning loops, and multi-agent handoffs, use the volidator.agent namespace:

# Log a decision made autonomously by the model
volidator.agent.decision({
    "actor": "writer-agent-v1",
    "traceId": "run-xyz-456",
    "decision": "publish_article",
    "rationale": "grammar check and safety alignments passed verification checks",
    "confidenceScore": 0.98,
    "modelId": "gpt-4o"
})

# Log a tool execution call
volidator.agent.tool_call({
    "actor": "search-agent-v1",
    "traceId": "run-xyz-456",
    "toolName": "web_search",
    "toolInput": {"query": "AI regulation"},
    "toolOutput": {"results": [...]},
    "success": True,
    "latencyMs": 245
})