pruv-openai 0.1.0

pruv verification layer for OpenAI Agents SDK

pruv-openai

Cryptographic verification for OpenAI Agents SDK. Every tool call, guardrail check, agent handoff, and LLM call — automatically recorded into a pruv identity chain.

pip install pruv-openai

Usage

from pruv_openai import OpenAIAgentWrapper

wrapped = OpenAIAgentWrapper(agent, agent_id="agent-id", api_key="pv_live_...")
result = await wrapped.run("analyze the quarterly report")
receipt = wrapped.receipt()

Implements the OpenAI Agents SDK TracingProcessor protocol. Your agent code stays unchanged. pruv runs underneath it.

How it works

PruvTraceProcessor implements the TracingProcessor protocol with automatic scope detection:

• tool spans → tool.execute scope
• handoff spans → agent.handoff scope
• llm / model / generation spans → llm.call scope
• guardrail spans → guardrail.check scope

Every span is posted to the pruv identity chain via POST /api/identity/act. The chain is tamper-evident — modify one entry and verification detects exactly where.

Links

• pruv.dev
• Documentation
• GitHub