Skip to content

Releases: ukanwat/overtime

0.7.3

Choose a tag to compare

@ukanwat ukanwat released this 17 Oct 03:01
  • Memories keep the person's exact words as a quote, and the rule is checked against them, so a short or misspelled correction can't be saved as its opposite.
  • The reflect pass spots patterns across several memories (saved as unconfirmed, never overriding the person's own rules) and keeps changeable facts current with a line of history; facts unchecked for a long time are marked possibly stale.
  • Snapshots of what's on disk aren't saved as memories; agents read the memory itself before tailoring to the person, and follow a saved rule as far as they can instead of stopping to ask.

0.7.2

Choose a tag to compare

@ukanwat ukanwat released this 17 Oct 02:44
  • A brand-new agent has the message tool from its first session, before it has its job.
  • An asleep agent's circle is a still, dim ○, so a screen of sleeping agents doesn't move.

0.7.1

Choose a tag to compare

@ukanwat ukanwat released this 17 Oct 02:08
  • A new look for the agent list: each agent shows a circle for its state (◐ working, ● needs you, ⊘ problem, ○ asleep, ◌ paused, ⊖ stopped, ◦ new) with the state as a word beside its name, and what it says it's doing underneath. The selected agent is a seamless highlighted block.
  • The message box is as wide as the conversation, so the agent list runs to the bottom and fits more agents.
  • Every message's left line sits in the same column.
  • New agents get a clickable Create button.
  • Scrolling while typing no longer types codes like <65;57;41M into the message box.
  • Memory: the background tidy-up pass is on. It reads an agent's whole history in order, can only write to memory/, never undoes your edits or touches skills, skips while offline, and never pauses the agent. Per-agent memory model and an on/off switch in settings.
  • Agent folders are no longer git repositories.
  • Helpers end their results with what they learned.

0.7.0

Choose a tag to compare

@ukanwat ukanwat released this 17 Oct 00:13
  • Agents have their own memory: a memory/ folder of notes with an index they always see, saved as they learn. Claude's built-in memory is off for agents, and memories Claude had saved are brought over on the first turn.
  • Each agent folder keeps a git history of its memory, skills, AGENT.md and INDEX.md.
  • Helpers can read their agent's memory but not change it.
  • New agents: pick backend, model, budget and visibility when creating one (app or overtime new).
  • Fixes: new-agent step keeps your choices when a backend isn't installed, wake times a week out show the date, clearer refusals in overtime answer and for agent-only commands, messaging an agent with no job says it isn't set up yet.
  • The background memory tidy-up pass is built but off for now.

Media

Choose a tag to compare

@ukanwat ukanwat released this 18 Oct 17:47

Images and recordings used by the README and docs, kept here so the repo stays small.

v0.6.1

Choose a tag to compare

@ukanwat ukanwat released this 16 Oct 22:46

Full Changelog: v0.6.0...v0.6.1

v0.6.0

Choose a tag to compare

@ukanwat ukanwat released this 16 Oct 21:34

Full Changelog: v0.5.11...v0.6.0

v0.5.11

Choose a tag to compare

@ukanwat ukanwat released this 16 Oct 19:30

Full Changelog: v0.5.10...v0.5.11

v0.5.10

Choose a tag to compare

@ukanwat ukanwat released this 13 Oct 17:50

Full Changelog: v0.5.9...v0.5.10

v0.5.9

Choose a tag to compare

@ukanwat ukanwat released this 12 Oct 17:38

Full Changelog: v0.5.8...v0.5.9