Skip to content

Releases: SpaceK33z/autoauto

v1.7.0

13 Apr 13:29
32c925b

Choose a tag to compare

1.7.0 (2026-04-13)

Features

  • experiment: diagnose silent no-commit outcomes (c84031c)
  • mcp: add first-time setup guidance for unconfigured projects (49d7698)
  • sandbox: remote container execution via Modal (#40) (afd6571)

Bug Fixes

  • agent: validate model compatibility before starting runs (8da8d89)
  • home: include onPanelChange in useEffect dependency array (888655f)
  • mcp: validate model compatibility in set_config before saving (bbc523b)
  • model: include available models in opencode format error message (20b330a)
  • setup: lower default max_experiments to 10 for new programs (34c8fef), closes #39
  • test: stabilize concurrent e2e runs (996f445)

v1.6.0

12 Apr 23:44
d93ae8b

Choose a tag to compare

1.6.0 (2026-04-12)

Features

  • add Task tool to experiment agent for parallel subagent research (4c1779c)
  • agent-panel: show experiment notes as styled card instead of raw XML (a1069f5)
  • auto-fallback to alternate model on provider limits (#32) (b98f34b)
  • execution: filter ideas panel to selected experiment (442f551)
  • execution: make narrow-mode Agent/Ideas tab titles clickable (46ae4b7)
  • expand CLI with 6 new commands for agent-friendly control (#30) (d159237)
  • guidance: add human steering for mid-run experiment direction (#37) (a0aa923)
  • home: add double-click to open programs and runs (025e647)
  • mcp: add agent sessions and config tools (#36) (cd9e647)
  • mcp: add MCP server for external coding agents (#35) (aa55e46)
  • mcp: add resources for setup guide and program details (ebc8d9d)
  • measure: add Mann-Whitney U statistical significance testing (#38) (b363c39)
  • proactive quota usage warnings from Claude SDK rate limit events (#34) (a62bb98)
  • ui: add hint prop to CycleField component (1454385)

Bug Fixes

  • abort active run when deleting its queue entry (e7353d8)
  • app: use q instead of Esc to quit from home screen (e423da3)
  • daemon: clean up stuck runs from crashed daemons (a1e5e67)
  • execution: single-click row selection with toggle to deselect (af5ead1)
  • execution: update phase label on attach instead of staying on "Connecting..." (d77c4de)
  • finalize: handle removed worktree for queue-sourced runs (4e3d6b7)
  • finalize: teach agent to navigate worktrees when target branch is checked out elsewhere (4bf38fe)
  • prompt: improve experiment agent orientation and backlog awareness (f6e38e9)
  • queue: require manual start instead of auto-starting on HomeScreen mount (2d39b89)
  • resolve merge conflict in queue E2E tests (43836d3)
  • surface clear error messages when daemon fails to start (#29) (963aea0)
  • treat baseline abort as clean termination instead of crash (939736b)
  • ui: check terminal phase when attaching to alive daemon (152ea8d)
  • ui: remove duplicate experiment heading in detail view (c3a96d5)
  • ui: rewrite setting descriptions to communicate trade-offs (d1d686a)

Performance Improvements

  • prompts: encourage parallel tool calls in experiment agent (072bef3)

v1.5.0

10 Apr 10:48
b54fbe3

Choose a tag to compare

1.5.0 (2026-04-10)

Features

  • add --version/-v flag to CLI (8745515)
  • add abort option to stop confirmation dialog (6dedba4)
  • add click-to-focus for all panels in execution screen (f8dc59b)
  • add Ctrl+D debug frame snapshot in dev mode (b70bb59)
  • detect provider quota/rate-limit errors and stop experiment loop (4add972)
  • improve setup config defaults and confirmation flow (db5eb1e)
  • prevent macOS sleep during daemon runs via caffeinate (4a0a1ce)
  • redesign home screen layout with runs on top and mouse selection (aaaf19e)
  • show finalization summary when viewing finalized runs (7bc68e6)
  • split first-time setup into conversational and experiment agent model pickers (17ab0b5)
  • surface skills/prompts as first example in setup screen (19c8755)

Bug Fixes

  • exclude .autoauto-* files from dirty worktree check (1f9b1da)
  • handle caffeinate spawn failures gracefully in sandboxed environments (b0ca279)
  • improve pre-run screen keybindings for start and queue actions (4fb6a81)
  • install cross-platform native deps for darwin-x64 release build (828441b)
  • prevent flex layout overflow with explicit min dimensions and row heights (a886caf)
  • resolve 9 failing CI tests (491b534)
  • sanitize ANSI sequences and control chars in table cell formatting (1529193)
  • show phase-appropriate keyboard shortcuts on execution screen (0ee1803)

v1.4.0

09 Apr 12:48
71d87d5

Choose a tag to compare

1.4.0 (2026-04-09)

Features

  • add queue management CLI commands (#21) (60ba2eb)
  • codex provider cost estimation from config and token pricing (d15caa5)
  • dedicated UI for dirty working tree with agent-powered commit (4d1fadc)
  • keep_simplifications toggle in pre-run screen (59192e0)
  • opencode provider child session cost aggregation (fa47719)
  • redesign finalize as conversational multi-turn chat (#25) (cdf808e)
  • show current branch name on dirty tree prompt (d2697a5)
  • surface turn budget to experiment agents to prevent analysis paralysis (584ebcf)

Bug Fixes

  • clean up update draft when exiting via PostUpdatePrompt (e334128)
  • codex sandbox mode blocks git commands in dirty tree commit agent (5e961a0)
  • update pre-run E2E tests for budget cap field navigation (1c224bf)

v1.3.0

09 Apr 01:20
d9a9e61

Choose a tag to compare

1.3.0 (2026-04-09)

Features

  • add interactive finalize review with approve/refine flow (aeaabfd)
  • add notification support on run complete (2effcb7), closes #9
  • carry forward previous run context into new experiments (#17) (7500f6f)
  • include termination reason in carry-forward context (71383ba)
  • persist setup drafts and resume sessions on return (f12fff1), closes #7
  • sequential run queue for overnight batch execution (#19) (08e0570)
  • verify results — re-run measurements after run completes (#13) (7af8eef)

Bug Fixes

  • add animated spinner indicators during active run phases (404f17c)
  • clean up setup draft when program was successfully created (2cac919)
  • merge main, update tests for verify feature and revised FinalizeApproval (0a0b447)
  • remove maxTurns limit from setup phase chat (570dd58)
  • rename Abandon to Done in run complete prompt (11627d0)
  • resolve FinalizeApproval rendering crash and fix E2E tests (8a96e29)
  • show streaming indicator during multi-turn agent activity (20efb04)
  • use macos-latest for darwin-x64 build and allow partial uploads (c4d778b)

v1.2.0

08 Apr 15:03
549487a

Choose a tag to compare

1.2.0 (2026-04-08)

Features

  • add cleanup agent, evaluator snapshots, and compact TUI layout (1f6cbad)
  • add clipboard copy-on-select via OSC 52 (05b73c4)
  • add codex agent provider (d86d8d9)
  • add column allocation utilities and edge-case fixes (f27466b)
  • add compact tab bar for small terminal windows (ef29916)
  • add daemon-backed experiment runs (bf313c7)
  • add duplicate program detection and improve setup prompts (8a45afe)
  • add escalating exploration directives to escape local optima (3b47ac3)
  • add experiment loop, context packets, and agent spawning for autoresearch (736e686)
  • add first-time setup screen for provider/model selection (c1e8439)
  • add formatShellError utility for better shell error messages (2ad3ef0)
  • add headless CLI subcommands for Claude Code integration (0d6e1a0)
  • add ideas backlog toggle (6638034)
  • add ideas.md panel to execution screen with live updates (b9b3734)
  • add inline timestamps and tool events to agent panel (7ae96a8)
  • add measurement diagnostics sidecar file support (0257b8c)
  • add measurement validation with variance analysis and stability assessment (33d2190)
  • add mid-run settings overlay for max experiments (f6a19c4)
  • add model configuration settings and auth check on startup (8e47ccf)
  • add navigable results table and experiment detail view (fad0596)
  • add optional in-place run mode (skip worktree) (0d0d6dd)
  • add permission bypass, tool status display, and autoresearch expertise to setup agent (6ff2376)
  • add pre-run config screen with time estimates and model overrides (4618069)
  • add program artifact generation with Write/Edit tools and user review flow (b54fd7d)
  • add program deletion with cascading run cleanup (0e9f76a)
  • add program management and project root utilities (6a9fe19)
  • add program update mode with auto-analysis of previous runs (c3a1ded)
  • add provider-specific model selection (3ed490f)
  • add push-based async iterable utility for SDK streaming (7544a39)
  • add re-baseline logic after keeps and consecutive discards (649422b)
  • add results reading, run listing, events logging, and cost tracking (d5db6be)
  • add run deletion from home screen (8ceb4fd)
  • add run lifecycle, measurement, and git utilities for experiment loop (bf36174)
  • add run termination, abort handling, and execution screen (903ce72)
  • add screen navigation with home and setup screens (0833253)
  • add secondary metrics support (advisory, non-gating) (e610523)
  • add setup agent with system prompt, tools, and ideation mode (16b94ae)
  • add setup mode chooser with codebase analysis option (db1d943)
  • add simplicity criterion — auto-keep experiments that simplify code (76d830e)
  • add tool status parsing, phase-aware indicators, and detach (dfd6893)
  • add TUI execution dashboard with stats, results table, and agent panel (d46da14)
  • allow deferred finalize on completed runs (39b3532)
  • initial project scaffolding with OpenTUI + Claude Agent SDK (2011300)
  • make agent max_turns configurable per-program (default 50) (2a84da0)
  • make max_experiments required instead of optional (849cb2b)
  • redesign home screen with programs and runs panels (29b6771)
  • render agent responses as markdown with syntax highlighting (128b796)
  • replace cleanup with finalize — group experiments into independent branches (b723a6f)
  • replace input with multiline textarea and auto-focus on typing (be02739)
  • separate build from measurement, add animated tool status, and improve setup UX (fc9b6f7)
  • show model label in setup/update chat title (3241c97)
  • sort programs by most recently used run on home screen (f4a7e97)

Bug Fixes

  • add contextual guidance for direct setup mode (0d44a84)
  • add flexGrow to select components so items render visibly (4eacda4)
  • add missing maybeRebaseline call after no_commit/agent_error outcomes (06c9eeb)
  • apply OpenTUI best practices across TUI components (0d13b06)
  • auto-stage .gitignore after adding .autoauto entry (fe5b5f1)
  • daemon agent provider init, stats header emoji width glitch, mktemp macOS guidance (811efed)
  • enable release-please labeling for proper release creation (5ad57c6)
  • improve tool event display with abbreviated paths (f957fe1)
  • kill child processes on ralph.sh exit to prevent orphaned claude sessions (854d788)
  • navigate to setup screen after first-time configuration (68d896c)
  • poll stream file in daemon watcher for macOS reliability ([cfd9f72](https://github.com/SpaceK...
Read more