explanation-traces

What problem does it solve?

Explanation traces prevent confident-sounding but unverifiable rationalizations by showing users a factual timeline of routing, agent selection, skill phases, and gate outcomes from the recorded session trace.

Core Features & Use Cases

• Reads recorded decision evidence: Presents a human-readable timeline sourced only from session-trace.json, never from memory or post-hoc inference.
• Query-focused trace filtering: Narrows results to the specific decision the user asks about (e.g., routing, a chosen agent, gate verdicts).
• Honest gap reporting: Flags missing or incomplete evidence fields so answers do not overclaim when the trace is partial.
• Troubleshoots missing tracing: If the trace file is absent, guides enabling hooks and points to the expected trace schema.

Quick Start

Ask your agent to explain routing decisions and show the decision log for your current session by querying the trace, for example: "show trace for why you chose that agent".