skills/browser-use__browser-use
1
0
Fork 0
mirror of https://github.com/browser-use/browser-use.git synced 2026-09-14 19:59:47 +08:00
Code Issues Packages Projects Releases Wiki Activity
9,690 Commits 564 Branches 148 Tags
codex/dev-core-extra-default
Commit Graph

182 Commits

Author SHA1 Message Date
MagMueller b727b36dcc fix(agent): skip screenshots on new tab pages 2026-05-25 20:38:12 -07:00
MagMueller 90a052371d Move user request before agent history 2026-05-23 09:48:15 -07:00
Saurav Panda a3bae60ed2 agent(prompts): move per-step metadata out of <agent_state> into a tail block 
The step counter (Step N maximum:M) and datetime.now() were rendered
inside <agent_state>, ahead of <browser_state> in the user message.
The cache miss already happens at the <agent_state> boundary today, so
this isn't a live cache regression — but the layout meant that any
future move of more-stable agent_state fields into the system prompt
would still leave per-step varying bytes in the middle of the prefix,
silently capping how far the cache could extend.

Pull both fields into a new _get_step_meta_description() and append it
at the very end of get_user_message(), after <agent_state>,
<browser_state>, <read_state>, <page_specific_actions>, and unavailable-
skills info. Everything above this tail block is now eligible to be
treated as the cacheable region.

Adds regression tests that lock the layout:
- <step_info> must appear after <agent_state> and <browser_state>
- <step_info> must not leak back into <agent_state>
- bytes before <step_info> must be identical across two different step
  numbers (the step counter must not be in the prefix)
 2026-05-22 17:58:13 -07:00
Saurav Panda 6d86be3a6b fix: detect skeleton screens and retry navigation for blank SPA pages 2026-03-19 18:48:47 -07:00
Saurav Panda 360f5842f5 improved scrolling instruction for agent 2026-03-10 20:58:18 -07:00
Saurav Panda 2895720eb0 fix: remove unused variable assignments 2026-03-07 14:37:31 -08:00
Saurav Panda 5ad365e31b added basic planing to agents 2026-02-02 12:28:29 -08:00
Laith Weinberger 2c5682618b only use special anthropic prompt for flash, as thinking & main sys are now min length 2026-01-31 16:03:45 -05:00
Laith Weinberger 2f24f53900 use longer prompts for opus/haiku 4.5 to allow caching 2026-01-23 16:20:25 -05:00
Magnus Müller 6bccc224ca Remove comma 2026-01-10 20:01:14 -08:00
Magnus Müller 5652ba5dd9 Small comment 2026-01-10 19:03:35 -08:00
Magnus Müller c4b9710a84 Less priming for scroll 2026-01-10 18:33:17 -08:00
Magnus Müller 1e0fd4d814 Remove CTA 2026-01-10 17:36:43 -08:00
mertunsall 110b96dc8d update system prompts 2025-12-10 18:50:10 +00:00
Gregor Žunič b1e4d6e686 primitive skills impl (with cookies) 2025-12-08 18:07:00 -08:00
Saurav Panda 7d2c6dc79e AI: add ai_step to replace extract_content on rerun 2025-12-01 21:34:09 +05:30
Mert Unsal 02cf38e864 Revert "allowing MUCH MORE with coordinates" 2025-11-28 04:00:22 -08:00
mertunsall 46cb1a0ae1 allowing MUCH MORE with coordinates 2025-11-27 14:48:23 -08:00
Cursor Agent e9cd88e6a5 Refactor page boundary indicators in prompts and formatting 
Co-authored-by: mailmertunsal <mailmertunsal@gmail.com>
 2025-11-24 22:10:08 +00:00
Cursor Agent d1b47eba50 Refactor: Reduce max actions per step to 3 
Co-authored-by: mailmertunsal <mailmertunsal@gmail.com>
 2025-11-22 01:29:59 +00:00
Saurav Panda 5e356108f9 feat: added AI summary for rerun task 2025-11-20 21:40:30 +05:30
Cursor Agent 487bd19e80 Refactor: Lower default max_actions_per_step to 4 
Co-authored-by: mailmertunsal <mailmertunsal@gmail.com>
 2025-11-15 06:31:10 +00:00
mertunsall a64c82ef2c sanitize everything now 2025-11-14 11:13:57 -08:00
mertunsall 0f5237a6e6 fix problems with utf-8 decoding 2025-11-08 19:00:40 -08:00
mertunsall 94aeb5d385 add system prompt special to anthropic 2025-11-08 09:56:15 -08:00
mertunsall d06d8fa8a2 image resizing the make sonnet 4.5 work 2025-11-07 22:20:14 -08:00
mertunsall 67d2f12665 default to PNG everywhere 2025-11-06 14:51:52 -08:00
mertunsall 03c9e94dc6 allow reading images such as jpg/png, as well as docx, fixes to the flash system prompt 2025-11-03 16:41:27 -08:00
Magnus Müller 9e93a92ba4 Include popup msg into browser state 2025-10-29 21:18:20 -07:00
Magnus Müller 86476e45df Linter 2025-10-28 20:53:25 -07:00
Magnus Müller f32fbc981d Remove scroll cta 2025-10-28 20:53:18 -07:00
Magnus Müller b678f27a3e Migration from highlight index to backendnode id 2025-10-21 21:02:28 -07:00
Magnus Müller 3eb6a5cb42 6 tokens less 2025-10-12 13:09:14 -07:00
Magnus Müller a50f2ca5b3 Small token changes 2025-10-12 13:09:10 -07:00
Magnus Müller ac9301b0d6 Reduce string for todo.md 2025-10-12 13:08:54 -07:00
Magnus Müller de0488031d Squashed commit of the following: 
commit 92d1732a40
Merge: fa4ff47a efcd2b33
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 08:18:35 2025 -0700

    Merge main

commit fa4ff47ae4
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 08:03:04 2025 -0700

    Laminar info mode

commit 8625aa9ba9
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 08:02:54 2025 -0700

    Default to flash for browser use llm

commit 69a517ad7c
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 08:02:44 2025 -0700

    Update example

commit c91467c51d
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 08:02:15 2025 -0700

    Update example

commit 4e2b3f1306
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 07:54:05 2025 -0700

    Linter

commit e8d6936a47
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 07:53:43 2025 -0700

    Example

commit 961d6fc451
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 07:52:23 2025 -0700

    Update default links

commit 0b2853859b
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Wed Oct 8 07:44:59 2025 -0700

    Default fast fasle

commit 9d67594b63
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 22:01:24 2025 -0700

    Simplify

commit f5712bdec0
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 21:45:00 2025 -0700

    Parsing json

commit 503af624dd
Merge: ef7fa1e4 0d2522b1
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 21:04:31 2025 -0700

    Merge branch 'flat-output' of https://github.com/browser-use/browser-use into flat-output

commit ef7fa1e401
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 21:04:28 2025 -0700

    Add prompt description handling for browser-use cloud service

    - Enhanced the Agent class to generate and pass prompt descriptions to the LLM when using the browser-use provider.
    - Updated the ChatBrowserUse class to accept an optional prompt_description parameter in the ainvoke method.
    - Modified the tracked_ainvoke function in TokenCost to forward additional keyword arguments.
    - Adjusted example model to initialize Laminar for tracing.

commit 0d2522b1f9
Author: Gregor Žunič <36313686+gregpr07@users.noreply.github.com>
Date:   Tue Oct 7 20:23:39 2025 -0700

    updated default base url

commit 0a2e6d6520
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 18:50:56 2025 -0700

    Hide api key

commit 96bba9692b
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 18:49:05 2025 -0700

    Add env var loading and validation to ChatBrowserUse, update example

commit c2e57a3bab
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 18:36:31 2025 -0700

    Linter

commit 3d4cc1dbec
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 16:28:10 2025 -0700

    Rename

commit 14ab0b75fe
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 16:21:52 2025 -0700

    Rename super_fast to fast parameter in ChatBrowserUse

commit 7fab3c47d3
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 16:08:10 2025 -0700

    Update system prompt

commit 2066e77b98
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 16:07:51 2025 -0700

    Del claude file

commit 297b767e91
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 16:01:50 2025 -0700

    Revert flash system prompt to JSON format

    - Reverted system_prompt_flash.md to main (JSON format)
    - Added examples/models/browser_use_cloud.py for llm-use cloud service
    - llm-use service will inject unstructured format on the server side

commit 99af55f8e9
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 15:58:40 2025 -0700

    Example

commit 657ff0d352
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 15:54:23 2025 -0700

    Remove unused

commit a42123d8e5
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 15:49:54 2025 -0700

    ignore

commit de4fe19020
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 15:48:01 2025 -0700

    Example

commit 27e5ab9947
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 14:29:01 2025 -0700

    Add llm-use cloud client example

    - ChatLLMUseCloud: Browser-use compatible client for llm-use service
    - Implements BaseChatModel protocol with proper typing
    - Includes credit checking before/after tasks
    - Error handling for 401/402/500 responses
    - Converts messages between browser-use and HTTP API formats
    - Returns ChatInvokeCompletion with usage stats

    Usage:
      export LLM_USE_URL='https://your-app.railway.app'
      export LLM_USE_API_KEY='your-key'
      python examples/llm_use_cloud_client.py

commit a47a448005
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Tue Oct 7 13:03:59 2025 -0700

    Enhance response parsing in ChatBrowserUse to handle structured data from the API, including debug logging for received data keys.

commit e79c7dc633
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Mon Oct 6 19:39:29 2025 -0700

    Include browser-use llm

commit 584d977154
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Mon Oct 6 17:55:31 2025 -0700

    Linter

commit b0ac87076a
Merge: 1183b8fe 0076775f
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Mon Oct 6 15:01:37 2025 -0700

    merge main

commit 1183b8fe6b
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 16:26:34 2025 -0700

    Fix done output action model

commit 852449b1bc
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 16:23:31 2025 -0700

    More debug info

commit a57ba25d36
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 16:07:15 2025 -0700

    Multiline parser

commit 2153dbe6aa
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 15:50:14 2025 -0700

    Fix custom output

commit ec5c08ba0f
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 15:36:54 2025 -0700

    Fix files in output

commit 7444048e8f
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 15:28:09 2025 -0700

    Update codebase with unstructured output

commit a8f05b4611
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 14:55:20 2025 -0700

    Save systemprompt

commit d7d112c569
Author: Magnus Müller <67061560+MagMueller@users.noreply.github.com>
Date:   Sun Oct 5 14:51:24 2025 -0700

    new parser
 2025-10-08 08:39:24 -07:00
Magnus Müller 324ac74204 Linter 2025-10-05 19:21:12 -07:00
Magnus Müller b53a7c116a Update outdated function names 2025-10-05 19:12:52 -07:00
Magnus Müller a8e6af254e Switch all to jpeg 2025-10-03 21:32:47 -07:00
mertunsall 0a924437d6 improve the pdf viewer prompt 2025-09-23 02:52:38 +02:00
Cursor Agent 166ea02173 Fix: Check for None in override_system_message 
Co-authored-by: mailmertunsal <mailmertunsal@gmail.com>
 2025-09-21 03:31:55 +00:00
Magnus Müller 75005acd5d Merge branch 'main' into improve-prompt-dom-meta-data 2025-09-17 11:33:32 -07:00
Magnus Müller 6244c97876 Include information about dom metadata in prompt 2025-09-17 11:32:11 -07:00
Magnus Müller 8c1207a614 fix-pixel-prompt (#3142) 
Auto-generated PR for branch: fix-pixel-prompt
    
<!-- This is an auto-generated description by cubic. -->
---

## Summary by cubic
Fixes the browser prompt to show scroll context in pages instead of raw
pixels and correctly detect content above/below the viewport. This makes
the page summary clearer and avoids misleading pixel counts.

- **Bug Fixes**
- Compute has_content_above/below from page_info using viewport-height
ratios.
  - Replace "N pixels above/below" with "{pages:.1f} pages above/below".
  - Guard against divide-by-zero when calculating page counts.

<!-- End of auto-generated description by cubic. -->
 2025-09-17 11:23:12 -07:00
Magnus Müller 00d0574b96 Refactor AgentMessagePrompt to remove redundant elements_text assignments for pages above and below. This simplifies the code and improves readability. 2025-09-17 11:18:01 -07:00
Magnus Müller a6211d2df9 Refine page info reporting by removing viewport size details and current time from step info 2025-09-17 11:12:52 -07:00
Magnus Müller 9b6ae59006 Fix pixel below 2025-09-17 11:08:39 -07:00
zhcn a99971f39c Merge branch 'main' into feature/sample_images 2025-09-02 18:51:03 +08:00
Magnus Müller 3f9b1cb650 Include max steps back in 2025-09-01 23:54:37 -07:00
Magnus Müller 4cb585e8a0 Update prompts and adjust temperature setting 
- Fixed the closing tag for the file_system section in agent prompts for proper formatting.
- Increased the temperature parameter in ChatOpenAI from 0.05 to 0.2 to allow for more varied responses.
 2025-09-01 23:50:07 -07:00