OFP: Agentic AI Standard for Hallucination Mitigation and Multi-Agent Collaboration
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
Open Floor Protocol (OFP) defines a vendor-independent JSON API for natural-language agent messages, a Python Bacon Forge Python framework with ready agents (Pete, Atina, Zuse), and companion specs (Dialog Events, Assistant Manifest) used to run experiments: a four-agent chain evaluated on 300 prompts with a custom “total hallucination score” KPI and reports of about 90% reduction in hallucination after fact-checking metadata flows; floor-sentinel agents and a supply-chain demo show tool calls, manifests, and orchestration.
AI-written summary · about
chapters (18)
- 0:00Multi-agent AI growth and challenges
- 1:05Open Floor Agentic AI Protocol
- 3:59Vendor-independent multi-agent standard
- 5:10Open Floor Protocol APIs
- 5:55Estonian government use case
- 7:21Python framework demo
- 8:03Live API interactions demo
- 9:50Open Floor specs and GitHub
- 10:11Hallucination mitigation methods
- 12:20Measuring hallucination reduction
- 14:05Over 90% hallucination mitigation
- 15:51Prompt injection risks and fixes
- 16:53Floor sentinel agents
- 18:16AI travel agency example
- 19:22Supply chain demo
- 21:07Industries benefiting from OFP
- 21:27Contributing to Open Floor
- 22:33Q&A
related talks
30mBeaconforge.org, Open Agentic AI for Hallucination Mitigation
31mGenerative Conversational AI Interoperability
25mBeyond Orchestration: An Open Source Architecture for Agentic Systems
25mAI Orchestration: Getting Super Agents to Complete a Mission
26mDo you speak Agent? The convergence of APIs, MCP, A2A, and Agent Languages.
26mOverview of AI Agent Protocols
46mBoF: An Open Future for Conversational AI -- Why is it Important, and How Do We Get There?
19mAgent Harnesses, MCP and the Next Frontier for OSPOs