Commit Graph

  • c39253ff50 Use UTF-8 for agent file writes Ghraven 2026-07-27 06:50:05 +08:00
  • 449e9ad5b9 Release 0.13.7 with Browser Harness 0.1.8 #5308 Laith Weinberger 2026-07-25 20:17:37 -07:00
  • 0964ad452a Call the model with minimal state after capture timeout (#5300) Magnus Müller 2026-07-24 18:29:05 -07:00
  • 05b68eb567 test(browser): verify state recovery after timeout #5300 MagMueller 2026-07-24 12:30:06 -07:00
  • 8c8e2db0ff Merge remote-tracking branch 'origin/main' into fix/minimal-state-timeout MagMueller 2026-07-24 11:53:46 -07:00
  • b909fbfba0 Simplify cross-origin iframe minimum size (#5299) Magnus Müller 2026-07-24 11:53:23 -07:00
  • e7cc40f5ad fix(browser): expose minimal state after capture timeout MagMueller 2026-07-24 11:45:53 -07:00
  • 9c8d12f7a7 refactor(dom): simplify cross-origin iframe size filter #5299 MagMueller 2026-07-24 11:38:52 -07:00
  • 29dc3db6fe Allow checkbox-sized cross-origin controls safely (#5298) Magnus Müller 2026-07-24 10:55:06 -07:00
  • c861d65b5f Merge remote-tracking branch 'origin/main' into fix/small-cross-origin-controls #5298 MagMueller 2026-07-24 10:52:06 -07:00
  • 07d1e39f85 Stop reusing cached DOM after browser-state timeouts (#5297) Magnus Müller 2026-07-24 10:49:26 -07:00
  • ffe93c190a fix(dom): allow compact cross-origin controls safely MagMueller 2026-07-24 10:45:35 -07:00
  • fa12474a87 fix(browser): stop reusing cached DOM after state timeout #5297 MagMueller 2026-07-24 10:42:33 -07:00
  • 245cdd8dfc Revert "fix(browser): recover with cached state after timeout" MagMueller 2026-07-24 10:38:26 -07:00
  • b7ac967ea5 Revert "fix(browser): bound cached state recovery attempts" MagMueller 2026-07-24 10:38:26 -07:00
  • 8a323269f9 Fix compact cross-origin payment field extraction (#5296) Magnus Müller 2026-07-24 09:47:06 -07:00
  • 58bf7f3b25 Revert "fix(dom): allow smaller embedded controls" #5296 MagMueller 2026-07-24 09:39:23 -07:00
  • e4d0818652 fix(dom): allow smaller embedded controls MagMueller 2026-07-24 09:30:16 -07:00
  • 53c3317fe5 fix(dom): index compact cross-origin payment fields MagMueller 2026-07-24 08:58:38 -07:00
  • 394ccc157e fix(browser): bound cached state recovery attempts MagMueller 2026-07-23 21:35:49 -07:00
  • bd45e4c22a fix(browser): recover with cached state after timeout MagMueller 2026-07-23 21:09:12 -07:00
  • 7f22c5a7b4 fix(browser): preserve DOM when state screenshots stall MagMueller 2026-07-23 21:00:34 -07:00
  • 9a0399a30f style: order DOM regression test imports MagMueller 2026-07-23 20:47:29 -07:00
  • ce0fc1b99b fix(dom): preserve structural state when AX tree fails MagMueller 2026-07-23 20:37:58 -07:00
  • 9bf7feffb6 fix(dom): bound JS listener CDP fanout MagMueller 2026-07-23 19:49:26 -07:00
  • 24e96d35a5 Keep selector indices unique across CDP sessions (#5291) Magnus Müller 2026-07-23 15:49:18 -07:00
  • 5a4b02c4bd Keep selector indices unique across CDP sessions #5291 MagMueller 2026-07-23 14:33:13 -07:00
  • 274f01cdd1 Fix nested cross-origin iframe extraction filtering (#5290) Magnus Müller 2026-07-23 15:43:14 -07:00
  • 2110918289 Scope paint-order filtering by iframe document #5290 MagMueller 2026-07-23 14:24:59 -07:00
  • 167f0c94d5 Scope DOM clickability cache by CDP session MagMueller 2026-07-23 14:21:52 -07:00
  • ac15362576 fix(async): avoid re-raising task callback errors GautamSharma99 2026-07-23 22:18:39 +05:30
  • d8e556b92f eval.yml: max-parallel 15 for the failed-task rerun (60min timeout, 429 headroom) cli-mcp Laith Weinberger 2026-07-23 00:57:04 -07:00
  • 96ec5dc137 Gate the skeleton-screen 'still loading?' hint on in-flight network requests (#5273) Magnus Müller 2026-07-22 20:47:52 -07:00
  • 6618c97ebd Merge branch 'main' into fix/skeleton-hint-network-gate #5273 Magnus Müller 2026-07-22 20:46:31 -07:00
  • 962cada3cb style: ruff format MagMueller 2026-07-22 20:33:49 -07:00
  • bf650fa5b1 Fix clear-to-empty on React controlled inputs (native value setter) (#5274) Magnus Müller 2026-07-22 20:33:09 -07:00
  • f60fdd10f1 hint wording: consider waiting (grammatical, per review) MagMueller 2026-07-22 20:29:35 -07:00
  • 923287e3a4 hint wording per Magnus: consider to wait MagMueller 2026-07-22 17:37:07 -07:00
  • 69f69bb06b skeleton hint: actionable data-driven wording; empty-page hint tells the model what to do MagMueller 2026-07-22 17:12:49 -07:00
  • 5adce7bf9b fix: gate skeleton-screen hint on in-flight network requests MagMueller 2026-07-22 16:58:59 -07:00
  • d4abb27612 review fixes: instanceof-guarded native setter (web components keep their own setter), drop dead honest-failure hunk, fix same descriptor bug in the concat-retry path #5274 MagMueller 2026-07-22 17:19:06 -07:00
  • 369678054c input clear: fail honestly when a pure clear leaves text behind MagMueller 2026-07-22 17:00:55 -07:00
  • 4ebc0f5d0d input clear: use native prototype value setter so React controlled inputs actually clear MagMueller 2026-07-22 14:37:43 -07:00
  • 854a7473e6 extract: stop auto-inheriting output_model_schema into per-page extraction (#5278) Magnus Müller 2026-07-22 20:15:57 -07:00
  • ba755a735c test: lock in that output_model_schema no longer auto-bridges into extraction_schema #5278 MagMueller 2026-07-22 20:05:08 -07:00
  • 0a64aaa4d4 extract: don't auto-inherit output_model_schema into per-page extraction MagMueller 2026-07-22 19:55:00 -07:00
  • 40717057a3 fix: handle file:// urls in _extract_start_url (#5276) laithrw 2026-07-22 17:43:25 -07:00
  • 581ab0d901 fix #5276 Laith Weinberger 2026-07-22 17:37:06 -07:00
  • 62df6b11f9 fix: handle file:// urls in _extract_start_url Laith Weinberger 2026-07-22 17:30:30 -07:00
  • b9c8dc88bb Fix tool name in flash prompts: replace_file_str -> replace_file (#5272) Magnus Müller 2026-07-22 17:09:37 -07:00
  • ff9f0836cf select_dropdown: native prototype value setter + async revert post-check for framework-controlled selects #5275 fix/react-select-revert MagMueller 2026-07-22 17:02:13 -07:00
  • daf7a3e112 fix: prompts referenced non-existent tool name replace_file_str (registered tool is replace_file) #5272 MagMueller 2026-07-22 16:55:32 -07:00
  • cb8e4dfcfe Revert "agent: retry-loop watchdog — flag identical action+outcome repeated 3+ times in multi_act" experiment/ts-no-rules-loop-watchdog MagMueller 2026-07-22 15:41:19 -07:00
  • 694e8bae40 Revert "browser state: surface recently completed fetch/XHR responses (url, status, age)" MagMueller 2026-07-22 15:41:19 -07:00
  • 2ebbf9731d input clear: use native prototype value setter so React controlled inputs actually clear MagMueller 2026-07-22 14:37:43 -07:00
  • abbfa8d34b browser state: surface recently completed fetch/XHR responses (url, status, age) MagMueller 2026-07-22 13:44:29 -07:00
  • 8ee25f30ce agent: retry-loop watchdog — flag identical action+outcome repeated 3+ times in multi_act MagMueller 2026-07-22 13:41:08 -07:00
  • 0117d2342a cap carried thought summary at 400 chars experiment/summary-cap MagMueller 2026-07-22 13:18:11 -07:00
  • 83bfa294a3 experiment: preserve Gemini signed turn history experiment/gemini-thought-signature-continuity Gregor Žunič 2026-07-22 13:02:00 -07:00
  • a96fff82cf fix: harden multi-turn reasoning continuity recording and replay experiment/multi-turn-reasoning-continuity MagMueller 2026-07-22 12:09:33 -07:00
  • 19b3f9d158 experiment: retain a transient native thought window experiment/flash-transient-thought-window Gregor Žunič 2026-07-22 11:59:01 -07:00
  • a1c1810134 fix: allow synthetic flash outputs without summaries experiment/flash-explicit-thought-summary Gregor Žunič 2026-07-22 11:09:47 -07:00
  • 1d41dff80a experiment: bound flash thought summaries Gregor Žunič 2026-07-22 11:07:00 -07:00
  • 83fe8da4d5 experiment: make thought summary length adaptive Gregor Žunič 2026-07-22 11:03:26 -07:00
  • a96577419f experiment: preserve explicit flash thought summaries Gregor Žunič 2026-07-22 11:00:07 -07:00
  • bbef184022 experiment: multi-turn reasoning continuity for ChatGoogle MagMueller 2026-07-22 10:57:02 -07:00
  • 405cac0e85 experiment: emit flash action before memory experiment/flash-memory-delta-decisive Gregor Žunič 2026-07-22 10:36:34 -07:00
  • 3e8c0b92fc experiment: choose action before writing checkpoint Gregor Žunič 2026-07-22 10:30:07 -07:00
  • 2dc8e6513e merge thinking into memory as plain prefix (no thought_summary tags) experiment/summary-prefix-memory MagMueller 2026-07-22 10:20:34 -07:00
  • 6279429968 experiment: simplify reasoning checkpoint rules Gregor Žunič 2026-07-22 10:19:41 -07:00
  • 0375dd2ecc experiment: align memory evidence with next action Gregor Žunič 2026-07-22 10:11:32 -07:00
  • aff0ff87de experiment: enforce evidence checkpoint format Gregor Žunič 2026-07-22 10:05:52 -07:00
  • 4c9e09e847 experiment: preserve compact reasoning checkpoints Gregor Žunič 2026-07-22 09:56:46 -07:00
  • 65d5854a36 retry malformed-empty completions at the provider (2 immediate attempts) experiment/no-rules-malformed-retry MagMueller 2026-07-22 09:53:31 -07:00
  • 617a6c8dfb merge signature shared-state fix experiment/sig-no-rules MagMueller 2026-07-22 09:28:12 -07:00
  • 3244c5698a fix: share retained thought context across dataclasses.replace clones experiment/thought-signature-continuity MagMueller 2026-07-22 09:28:10 -07:00
  • dba15484ff composite: signatures + no rules block MagMueller 2026-07-22 09:21:36 -07:00
  • 428952225f ablation: drop thought_summary_rules block (model self-directs its exposed thinking) experiment/thought-summary-no-rules MagMueller 2026-07-22 09:00:44 -07:00
  • 971cbdc589 experiment: combine thought memory with evaluate history experiment/flash-thought-summary-evaluate-history-no-prompts Gregor Žunič 2026-07-22 08:44:19 -07:00
  • 7a3318988c experiment: retain Gemini thought signatures across agent steps MagMueller 2026-07-22 08:34:38 -07:00
  • e951881b24 thought-summary only: drop evaluate_rules (eval isolates the summary change) experiment/thought-summary-only MagMueller 2026-07-22 00:14:11 -07:00
  • cfb1b37c9b experiment: persist Gemini thought summaries in flash memory Gregor Žunič 2026-07-21 23:30:25 -07:00
  • 9197791e9e experiment: persist Gemini thought summaries in flash memory only experiment/flash-thought-summary-memory-only Gregor Žunič 2026-07-22 00:02:24 -07:00
  • 8f6d22cbd7 experiment: persist Gemini thought summaries in flash memory experiment/flash-thought-summary-memory Gregor Žunič 2026-07-21 23:30:25 -07:00
  • 1bd773f7ad fix: preserve evaluate code in agent memory fix/evaluate-remember-code Gregor Žunič 2026-07-21 22:32:42 -07:00
  • b3b7e12fa3 fix(browser): apply storage_state cookies when passed as an in-memory dict ayaangazali 2026-07-20 11:21:55 -07:00
  • 2be09b6c5e fix: dont send default params in gemini genconfig (#5173) Saurav Panda 2026-07-20 10:26:41 -07:00
  • a567c6dacf Merge branch 'main' into gemini-params #5173 Mark McDonald 2026-07-20 08:48:31 +05:30
  • 5df6bdb2d7 ci: retrigger cancelled/flaky test workflow. santhreal 2026-07-18 16:32:50 -07:00
  • c0d0542beb fix(mcp): mark read-only MCP tools with readOnlyHint annotations ayaangazali 2026-07-16 23:32:37 -07:00
  • 950eb03617 Release 0.13.6 with Browser Harness 0.1.6 (#5245) 0.13.6 Gregor Žunič 2026-07-16 22:28:49 -07:00
  • 169ebee82e Release 0.13.6 with Browser Harness 0.1.6 #5245 Gregor Žunič 2026-07-16 22:24:40 -07:00
  • 058e1cf73d Bump browser-use version to 0.13.5 + sync BH (#5243) 0.13.5 laithrw 2026-07-16 17:45:56 -07:00
  • 3cd3816c57 sync browser-use skill from harness main #5243 Laith Weinberger 2026-07-16 16:26:01 -07:00
  • a23c6ebdc7 Bump browser-use version to 0.13.5 Laith Weinberger 2026-07-16 16:20:23 -07:00
  • e9c3457140 support MCP registry (#5242) laithrw 2026-07-16 16:12:16 -07:00
  • 604392c392 support MCP registry #5242 Laith Weinberger 2026-07-16 15:54:33 -07:00
  • ceda06382e fix(mcp): surface CallToolResult.isError as a failed ActionResult Andrew Chen 2026-07-16 00:32:02 +08:00
  • 21dcbaa29c Fix ChatAnthropicBedrock.ainvoke() IndexError on an empty response.content list Andrew Chen 2026-07-15 15:08:10 +08:00
  • 77ff98ff5a style: reformat markdown chunking test for ruff 0.12.10 Diwak4r 2026-07-09 07:38:58 +05:45