Motivation
Run analysis today means reading runs/<id>/summary.json / events.jsonl / agent JSONL logs by hand. There is no visual leaderboard, value-series view, or spectator experience. Building this as custom Blockscout pages was considered and rejected: the data lives in run artifacts (not the explorer DB), and forking blockscout-frontend would add permanent upstream-tracking cost. Instead: a standalone Eris dashboard that consumes run artifacts and links out to Blockscout for chain-level detail.
Proposal
A self-contained dashboard/ package in this repo (dependency direction dashboard → sdk only; update check:boundaries accordingly). A small web app reading runs/<id>/ artifacts and the anvil RPC.
Phase 1 — post-run analytics (static artifacts only)
- Run list and run detail: leaderboard from
summary.json, per-agent value series from events.jsonl with stress/vuln event windows overlaid.
- Agent detail: decision timeline (
reason/signals from agents/<agentId>.jsonl), mempool self-reports (submitted / submit_failed / rejected), tx list.
Phase 2 — live spectator view
- Tail
events.jsonl during a run + poll RPC for fair/venue prices; live leaderboard and market view (fair vs venue prices, basis). Revisits the ADR 0008 direction. Phase 1 structure should keep this pluggable (same readers, streaming source).
Blockscout link integration
Design details (stack, charting, exact routes) to be worked out at implementation time.
Motivation
Run analysis today means reading
runs/<id>/summary.json/events.jsonl/ agent JSONL logs by hand. There is no visual leaderboard, value-series view, or spectator experience. Building this as custom Blockscout pages was considered and rejected: the data lives in run artifacts (not the explorer DB), and forking blockscout-frontend would add permanent upstream-tracking cost. Instead: a standalone Eris dashboard that consumes run artifacts and links out to Blockscout for chain-level detail.Proposal
A self-contained
dashboard/package in this repo (dependency directiondashboard → sdkonly; updatecheck:boundariesaccordingly). A small web app readingruns/<id>/artifacts and the anvil RPC.Phase 1 — post-run analytics (static artifacts only)
summary.json, per-agent value series fromevents.jsonlwith stress/vuln event windows overlaid.reason/signalsfromagents/<agentId>.jsonl), mempool self-reports (submitted / submit_failed / rejected), tx list.Phase 2 — live spectator view
events.jsonlduring a run + poll RPC for fair/venue prices; live leaderboard and market view (fair vs venue prices, basis). Revisits the ADR 0008 direction. Phase 1 structure should keep this pluggable (same readers, streaming source).Blockscout link integration
Design details (stack, charting, exact routes) to be worked out at implementation time.