H

hallucination-detection

Local dashboard that streams a Claude, GPT or Ollama reply chunk by chunk, shows where the model paused, flags claims worth checking and prices the prompt before you send it. Works fully offline with Ollama.

Agent supervision, memory distillation, and multi-model orchestration. Kybernetes steers AI agents — catching hallucination and drift before they compound.

A deterministic verification layer for AI systems. QWED verifies AI outputs using mathematics, symbolic reasoning, and formal methods (Z3, SMT, SymPy), creating an auditable trust boundary for agentic AI. Not generation. Verification.