Skip to content

Update Pi package namespace to @earendil-works - #20

Merged
Michaelliv merged 2 commits into
mainfrom
update-earendil-works-packages
May 31, 2026
Merged

Michaelliv merged 2 commits into
mainfrom
update-earendil-works-packages

Conversation

@Michaelliv

Copy link
Copy Markdown
Owner

Summary

  • switch Pi imports from @mariozechner/* to @earendil-works/*
  • add explicit @earendil-works/pi-ai peer/dev dependency for exported message types
  • pin @earendil-works/pi-* dev dependencies to 0.78.0 and peer dependencies to ^0.78.0
  • refresh package-lock.json for the updated package namespace

Verification

  • npm test

@Michaelliv
Michaelliv merged commit dc6ad26 into main May 31, 2026
1 check passed
HangxiangMa pushed a commit to HangxiangMa/pi-dynamic-workflows that referenced this pull request Sep 7, 2026
…n/s (Michaelliv#20) (Michaelliv#21)

Add a 'detailed' mode to the live 'Workflows running' panel alongside the
existing compact one-liner. Detailed mode renders per-phase progress (marker +
done/total + error count + token subtotal) and per-agent rows ([id] + status
icon + label + tokens + model), capped at progressPanelMaxAgents per phase with
a '... N earlier agents' overflow line, plus a run header with aggregate tokens,
cost, and a live tok/s rate. Compact mode is byte-for-byte unchanged.

New settings in ~/.pi/workflows/settings.json: progressPanelMode
('compact'|'detailed') and progressPanelMaxAgents (1-1000, default 8), live-read
each render (1s TTL) so toggles apply without a restart. New commands
/workflows-progress <compact|detailed|status> and /workflows-progress-max <N>.

token/s is summed from per-agent agent.tokens, NOT snapshot.tokenUsage.total:
real-pi testing showed the run-level aggregate is only finalized at run-end and
reads 0 for the entire live run, while per-agent tokens update on each agent
completion. Reuses statusIcon/shorten (display.ts) and shortModel (workflow-ui.ts)
via new exports; no duplication.

Tests: 16 new unit tests (settings normalize/clamp, token-rate math incl.
ramp/stall, detailed layout, per-phase overflow, mode selection, commands).
Verified end-to-end with a real deepseek-v4-flash run (live header: '31.7K tok
- 4434 tok/s').
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants