EMRY-066: multi-run project dashboard #109

New issue

Closed

opened 2026-07-02 20:47:22 +00:00 by vxfemboy · 0 comments

vxfemboy commented

2026-07-02 20:47:22 +00:00

(Migrated from github.com)

emry web --run-dir shows one run; emry compare A B is two-run/final-only. Add a project view: emry web --project <log-dir> lists every run in the directory and overlays their curves for a chosen metric, so you can compare a whole sweep at once.

The CLI loads all runs via emry_store::{list_runs, load_baseline} and serves them to a self-hosted page (/project JSON + overlaid multi-line canvas chart, one color per run, metric selector, run legend/toggles).
Static snapshot from the run dirs' metrics.jsonl (no WS — these are historical runs); reload to refresh.
emry-web stays store-free; the CLI (which already depends on both) assembles the data.

`emry web --run-dir` shows one run; `emry compare A B` is two-run/final-only. Add a **project** view: `emry web --project <log-dir>` lists every run in the directory and overlays their curves for a chosen metric, so you can compare a whole sweep at once. - The CLI loads all runs via `emry_store::{list_runs, load_baseline}` and serves them to a self-hosted page (`/project` JSON + overlaid multi-line canvas chart, one color per run, metric selector, run legend/toggles). - Static snapshot from the run dirs' `metrics.jsonl` (no WS — these are historical runs); reload to refresh. - `emry-web` stays store-free; the CLI (which already depends on both) assembles the data.