Clearailhc/clearai-dsh

What it solves

ClearAI is designed to move beyond simple chat transcripts and task-completion loops in AI research. It solves the problem of creating a trustworthy, structured knowledge base (a domain ontology) where every piece of information is earned through a disciplined verification process rather than just being asserted by an LLM.

How it works

The platform operates using an "Epistemic Loop"—a seven-stage process (frame, hypothesize, plan, observe, verify, evaluate, record) that treats every conclusion as a hypothesis to be tested. Unlike standard agent loops that only track if a task is done, ClearAI tracks the evidence and grounds for every conclusion.

At runtime, these stages are compressed into four beats: plan, execute, observe, and reflect. The system automatically surfaces contradictions (conflict readings) and maintains a detailed ledger of evidence chains, boundaries, and support levels for every entry in the ontology.

Who it’s for

Researchers and scientists (e.g., in AI for Science, mathematics, or physical-world experiments) who need a rigorous, verifiable knowledge structure that tracks the history and evidence of their discoveries.

Highlights

• Domain Ontology: Automatically grows a vocabulary of concepts, predicates, and verified entries as research progresses.
• Epistemic Loop: A rigorous verification framework that separates the "doer" from the "judge" to ensure conclusions are trustworthy.
• Knowledge Inspector: A detailed view providing definitions, relations, assertions, and the full evidence chain for any node or edge in the graph.
• Conflict Detection: Automatically identifies and surfaces contradictory conclusions, leaving the final decision to the human user.
• DSH Integration: Delivered as a plugin for the DSH engine, adding an epistemic layer without modifying the core engine.