This repository was archived by the owner on Oct 16, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Oct 16, 2026. It is now read-only.
EPIC: Dashboard results inspection and scale #1466
Description
Track the public roadmap for AgentV Dashboard result inspection as the product moves from small local runs to larger artifact-backed projects.
Scope:
- Keep the Dashboard as the zero-infra local cockpit over AgentV-owned run artifacts.
- Make run lists, result tables, comparison views, and detail drilldowns scale without reading every heavy artifact on every refresh.
- Load trace, transcript, file, and artifact bodies lazily from canonical run bundles or configured remote result sources.
- Preserve clear source/provenance indicators when results come from local or remote storage.
- Improve the review workflow with focused run detail, filtering, sorting, comparison, and live/progress affordances where they fit the local Dashboard boundary.
Non-goals:
- Do not turn the Dashboard into a hosted observability platform.
- Do not make a persistent database the canonical result store; run bundles and Git-backed artifacts remain authoritative.
- Do not duplicate Phoenix, Opik, or other external trace UIs inside AgentV beyond the local inspection experience AgentV needs.
Related public threads will be consolidated here when they are older implementation slices of the same Dashboard/results-inspection roadmap.
Activity
Keesan12 commentedon Jul 21, 2026
Results inspection at scale needs a verdict before it needs more charts. Which runs are trustworthy, which failed honestly, and which look successful but lack evidence?
That is the dashboard question driving MartinLoop: turn agent activity into software work a team can actually approve.
What single filter would save operators the most time in a large result set? I would appreciate your feedback, and the open-source repo is here if useful: https://github.com/Keesan12/martin-loop