Skip to content

v2.0.43

Choose a tag to compare

@github-actions github-actions released this 10 Oct 11:35
· 99 commits to main since this release

What's Changed

  • fix(sdk): persist Codex step history by @kjgbot in #617
  • feat(babysitter): pin standalone launch artifact by @kjgbot in #616
  • regex_match gates fail in the standalone runtime: Cannot find module 're2js' by @agent-relay-code[bot] in #623
  • flows: refresh an expired agent-relay cloud login instead of asking for a re-login by @agent-relay-code[bot] in #614
  • No detached run mode: --local-agent binds the agent worker to the invoking terminal by @agent-relay-code[bot] in #622
  • unprovable_effects warns once per step and fires on nearly every real spec, so the warning channel carries no signal by @agent-relay-code[bot] in #621
  • flows check never talks to the daemon, so it structurally cannot catch spec/daemon skew by @agent-relay-code[bot] in #619
  • CLI: surface needs_human from Cloud (--wait) and add flows answer --cloud (answer + resume in one command) by @agent-relay-code[bot] in #624
  • relayflowd run.get times out under CPU load and kills the run: a flow's own test suite can destroy 30 steps of journaled work by @agent-relay-code[bot] in #613
  • flows check refuses a Cloud-bound f.slack flow locally (helper_slack.credential_missing) before printing REQUIRES by @agent-relay-code[bot] in #618
  • fix(sdk): pass agent task text after an end-of-options separator by @AgentRelayBot in #625
  • Authoring docs describe a 2.0.8-era surface; refresh against 2.0.19, starting with f.run's undocumented 30s default lease by @agent-relay-code[bot] in #615
  • docs(examples): software-factory is a plain example; Software Garden lives in agentrelay.com by @khaliqgant in #628
  • ci: run CI only for trunk PRs and main pushes by @kjgbot in #630
  • docs(agents): install merge-train snippet by @kjgbot in #631
  • revert: trunk-only CI gating (repo not in merge-train core set) by @kjgbot in #632
  • feat(babysitter): wake on real PR review feedback, not only verdicts by @khaliqgant in #633
  • feat(babysitter): standalone fixer that proposes a patch and never pushes by @khaliqgant in #638
  • feat(babysitter): keep the fixer's PR checkout across reused sandboxes (F3a) by @khaliqgant in #635
  • feat(sdk,kernel): f.agent resume continues a recorded CLI session by @khaliqgant in #636
  • feat(babysitter): resume the previous wake's agent session on a reused box (F4) by @khaliqgant in #637
  • feat(babysitter): fixer task modes for the merge-train agent (P2) by @khaliqgant in #644
  • chore(ops): remove the flows-watchdog Relayflow v1 schedule source by @khaliqgant in #629
  • feat(sdk): flows eval — re-run a flow version against a frozen suite as a deploy gate by @khaliqgant in #642
  • fix(ops): handle host rustup cargo shims correctly by @Rahulstark2 in #601
  • feat(sdk): runInCloud envSecrets option for per-run sandbox secrets by @khaliqgant in #645
  • fix(sdk): implement f.memory.learn so findings persist across runs by @khaliqgant in #641
  • feat(examples): self-improvement flow v1 — read runs, propose prompt+structure edits, open PR by @khaliqgant in #643
  • fix(babysitter): propose from a private git dir, never the checkout's .git by @khaliqgant in #646
  • fix(self-improvement): restore the shipped-source-models contract (unblocks cloud-runtime-artifact) by @khaliqgant in #651
  • fix(babysitter): ignore executable bits the sandbox's filesystem dropped by @khaliqgant in #652

New Contributors

Full Changelog: v2.0.42...v2.0.43