Live MCP session — payloads captured 2026-04-02 from Graphonomous v0.2.0

A/B: Full Learning Pipeline — Interaction Ingestion vs Feedback Correction

Two tools, one pipeline: A uses learn_from_interaction to ingest a full user/assistant exchange (novelty 0.844, 5 claims extracted, 6 nodes created); B uses learn_from_feedback to boost one node's confidence and correct another's content (token-bucket → sliding-window). The system learns from conversations and improves from corrections.

Tools showcased: learn_from_interaction learn_from_feedback
Chat + tool-call storytelling
Live tool-call log
idle