Surface active experiment state during training (#59) - #78
Open
kunalgoesbyken wants to merge 2 commits into
Open
kunalgoesbyken wants to merge 2 commits into
kunalgoesbyken wants to merge 2 commits into
Conversation
- cmd_show now includes active_attempt (phase, status, diff) from attempt_state.json and pre-benchmark diff.patch - Dashboard enriches active nodes with active_attempt metadata and renders phase in the drawer - evo metrics <exp_id> [--tail] cli command for live training metrics - metrics.jsonl listed in dashboard logs tab with live tail
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This was referenced Jun 27, 2026
alvaroborras
added a commit
to alvaroborras/evo
that referenced
this pull request
Jun 27, 2026
…#1) * fix: preserve incomplete trailing lines in evo metrics --tail, handle truncation * Surface active experiment state during training (evo-hq#59) - cmd_show now includes active_attempt (phase, status, diff) from attempt_state.json and pre-benchmark diff.patch - Dashboard enriches active nodes with active_attempt metadata and renders phase in the drawer - evo metrics <exp_id> [--tail] cli command for live training metrics - metrics.jsonl listed in dashboard logs tab with live tail * feat: live observability dashboard for active experiments (evo-hq#59) Expand issue evo-hq#59 scope beyond PR evo-hq#78 to include: - Live diff panel: _DiffRefresher daemon rewrites diff.patch every 5s during benchmark - Metrics sparklines panel: tail metrics.jsonl with key filtering (--keys loss,lr) - Samples preview panel: new cmd_samples --tail for last N samples - Checkpoints panel: new cmd_checkpoints for saved model checkpoints - HTML export: self-contained report via new cmd_export (--full for untruncated) - Dashboard frontend: 5-panel Live tab auto-refreshing every 5s for active experiments - CLI improvements: cmd_metrics --keys filtering, hardened per-attempt dirs with EVO_METRICS_PATH/EVO_SAMPLES_PATH env vars - Tests: 40/40 passing (lifecycle, diff_refresher, export_html, active_attempt) - Skill docs updated with live dashboard monitoring and env var I/O contract --------- Co-authored-by: krockxz <71250874+krockxz@users.noreply.github.com> Co-authored-by: Fernandez, Alvaro <alvaro.fernandez@oliverwyman.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #59.
During long training runs (SFT, RL — often 1-2h) the dashboard and CLI showed nothing useful until the run finalized. The data existed on disk (attempt_state.json, diff.patch, metrics.jsonl) but evo didn't surface it.
Changes