Commit Graph
96 Commits
Author SHA1 Message Date
James Peter 5082189656 feat(models): add GPT-5.4 fast mode support 2026-03-06 10:51:21 +10:00
James Peter 5d17a7ed69 feat(core): add memory config and model metadata updates 2026-03-01 16:41:21 +10:00
James Peter 2be7be994e feat(auto-drive): add configurable CLI routing entries 2026-02-16 14:02:38 +10:00
James Peter 2a4c39fd00 fix(auto-review): fallback from spark on usage-limit errors 2026-02-15 16:18:42 +10:00
James Peter d165389fe3 feat(models): gate spark behind pro-only auth capabilities 2026-02-15 14:05:23 +10:00
James Peter 60727b0682 fix(auto-drive): decouple auto review and cap long-session growth 2026-02-15 07:02:04 +10:00
James Peter fb1752750f feat(auto-drive): add per-turn model routing and schema rename 2026-02-14 16:31:45 +10:00
James Peter e4559ac4f9 fix(auto-drive): require strict decision fields without allOf 2026-02-14 10:56:18 +10:00
James Peter 873604d6fa fix(auto-drive-core): align coordinator schema with goal bootstrap flow 2026-02-14 09:20:10 +10:00
James Peter 46d537e4dc feat(auto-drive): enforce finish evidence and paste fallback 2026-02-14 09:13:17 +10:00
James Peter b86fcd5db8 feat(app-server): add app-server support 2026-02-11 08:17:08 +10:00
James Peter 66abab5ffc fix(auto-drive): stop on usage limit without reset (fixes #534) 2026-02-02 13:36:55 +10:00
James Peter 654d0bba89 fix(core/compact): harden auto-compact flow 2026-01-29 23:30:16 +10:00
James Peter ea0e3892b7 fix(auto-drive): track estimated tokens without usage 2026-01-22 08:54:48 +10:00
James Peter 5777f19e2d fix(tui/auto-drive): align Esc edit/stop behavior 2026-01-18 21:10:29 +10:00
James Peter 94eb8d23fe feat(browser): add navigation telemetry; migrate auto-drive history 2026-01-06 22:57:47 +10:00
James Peter 590f46be9b fix(core/auth): auto-switch on usage limit with notice 2026-01-05 16:27:21 +10:00
James Peter 7f6c12e8e3 fix(auto-drive): keep retrying after errors 2025-12-30 14:21:10 +10:00
James Peter d3efecb545 chore(auto-drive): refine timeboxed exec guidance 2025-12-25 14:47:49 +10:00
James Peter 736e6cf03e fix(auto-drive): drop unused time budget seed 2025-12-24 17:10:42 +10:00
James Peter 8dbfdbba50 feat(exec): add timeboxed auto-exec guidance 2025-12-24 17:08:49 +10:00
James Peter 376fc8ffc0 fix(auto-drive): tighten time budget guidance 2025-12-24 11:49:50 +10:00
James Peter f5c56198ca fix(auto-drive): respect retry budget and loop hints 2025-12-22 07:11:45 +10:00
James Peter d2dbf60af9 fix(tb2): harden exec env/shell; gate auto drive rails
- Add non-interactive env defaults to reduce hangs\n- Suppress bash job control for bash -lc exec\n- Make Auto Drive turn cap opt-in (default unlimited); add --turn-cap\n- Only bias coordinator finishing when --max-seconds is set
2025-12-17 05:37:54 +10:00
James Peter 0c323447a9 feat(exec): add max-seconds budget with countdown nudges 2025-12-16 08:01:48 +10:00
James Peter d53c71df9b fix(models): align presets with protocol types 2025-12-10 08:55:17 +10:00
James Peter 25bc53e1fa feat(auto-review): reuse worktree and pass turn context 2025-12-02 13:13:51 +10:00
James Peter 7640ff5ff4 fix(exec/auto-resolve): prevent stale follow-ups and snapshot races 2025-12-01 22:53:00 +10:00
James Peter 13226204c5 fix(auto-drive): force read-only agents without git repos 2025-11-26 16:59:44 +10:00
James Peter 7ffc4467c2 fix(auto-drive): durable retries and coordinator stability
- keep failed overlong coordinator outputs in the conversation, suppress compaction notices, and surface a retry action in the Auto Drive card so follow-up requests retain the error context
- add explicit Action app event, pass compaction notice intent through to the TUI, and raise the coordinator worker stack to 1 MiB with an opt-in backtrace hook for debugging validation overflows
- simplify bottom_spacer_hysteresis_requests_followup_frame to a known viewport pair to avoid FD exhaustion during repeated harness setup

Fixes #432
Fixes #430
Fixes #433
Fixes #429
2025-11-26 09:09:46 +10:00
James Peter 6f6b99c579 chore(auto-drive): user-facing retry note without guidance
Replace the retry Thinking event text with a neutral message so the Auto Drive card shows the retry without exposing prompt-length guidance to the user. Build fast remains green.

fixes #432
2025-11-26 08:34:44 +10:00
James Peter 333b2855da chore(auto-drive): persist retry note into action history
Add assistant retry note into the conversation so the Auto Drive card shows context while retrying overlong prompt_sent_to_cli errors. Developer guidance string remains in the conversation; raw model output stays in assistant message.

fixes #432
2025-11-26 07:45:50 +10:00
James Peter ed24f70144 chore(auto-drive): place overlong prompt error in developer note
Assistant message now keeps the raw model output; developer note carries the fixed guidance string so the coordinator sees both the failure context and the explicit <=600 notice. Build fast remains green.

fixes #432
2025-11-26 07:38:57 +10:00
James Peter 1cb7fcf923 chore(auto-drive): clarify overlong prompt retry messaging
Replace the retry assistant insert with a fixed error string so the CLI clearly sees the overlong prompt error without echoing the long prompt. Persist the message in the conversation and simplify developer notes.

fixes #432
2025-11-26 07:33:31 +10:00
James Peter 7841f887a6 fix(auto-drive): show failed coordinator output when retrying
Surface the invalid assistant JSON as an assistant message before retrying recoverable validation errors (e.g., overlong prompt_sent_to_cli) so the coordinator can see what failed. Retains existing retry guidance path.

Refs #432
2025-11-26 07:21:47 +10:00
James Peter 5726826120 fix(core): agent dedupe and windows cmd resolution
- auto-drive retries overlong prompt_sent_to_cli instead of failing
- agents: canonical alias dedupe + PATHEXT-aware cmd resolution and Gemini config fallback

fixes #430
fixes #432
fixes #433
2025-11-26 06:56:15 +10:00
James Peter 24a0dd4c56 fix(auto-drive): restore 600-char CLI prompts and handle cutoffs 2025-11-26 01:15:06 +10:00
James Peter 3651bc85ae fix(auto-drive): enforce 4-140 char cli prompts 2025-11-25 23:18:37 +10:00
James Peter 936cca8f74 enhance(stream): carry request id, classify drops, log exhausted retries 2025-11-25 08:46:44 +10:00
James Peter f0cb7afda1 fix(core/tui): harden connectivity retries and reconnect UI 2025-11-24 10:22:34 +10:00
James Peter 6a7cac9d2b fix(models): clamp reasoning effort to supported levels (closes #422) 2025-11-22 11:44:10 +10:00
James Peter 6184ac4cb0 fix(auto/compact): bound remote compaction with timeout 2025-11-21 14:37:20 +10:00
James Peter 61209e062d feat(auto-drive): allow model selection and clamp verbosity 2025-11-21 03:03:28 +10:00
James Peter 698c53f25f feat(core/models): migrate defaults to gpt-5.1 2025-11-14 08:19:52 +10:00
James Peter 55e1c83c6a refactor(agent-models): canonicalize model guide 2025-11-12 12:51:12 +10:00
James Peter 76112d3623 added gpt-5-mini agent 2025-11-12 11:56:54 +10:00
James Peter 9759fceaee improve compaction 2025-11-12 11:06:32 +10:00
James Peter adae9ce74b prompt tweaks 2025-11-08 14:10:05 +10:00
James Peter 25bd02f9ce row spacer 2025-11-08 13:46:19 +10:00
James Peter a40fe2c347 fix(auto-drive): shorten rate-limit waits 2025-11-08 11:50:47 +10:00