MiroSharkMiroShark

Reports

How the ReACT report agent synthesizes post-mortems, consensus signals, belief trajectory SVGs, and run cost breakdowns.

When a simulation finishes, a ReACT report agent synthesizes the run into cited narrative and structured signals.

Report outputs

OutputDescription
ReACT report agentTool-calling loop over simulation_feed, market_state, graph search, belief_trajectory, and Nash equilibrium analysis. Every claim cites a post or trade.
signal.jsonDistills the final belief split into bullish, bearish, or neutral via a plurality rule with deterministic tie-breaking.
Trajectory SVGPure-stdlib SVG plotting per-round belief means, deltas, and confidence bands.
Consensus badgeShields.io-style SVG pill (green / grey / red) embeddable in any Markdown doc.
Run cost (cost.json)GET /api/simulation/<id>/cost.json reports estimated_cost_usd, token/latency totals, and by_model / by_phase breakdowns.
Quality diagnosticsScores entity coverage, graph density, agent activity, and belief movement to flag thin seeds or inert swarms.

On this page