Starstruk Logo MarkStarstruk

v0.1.76

Skyvern-AI/skyvernv0.1.76Apr 16, 2025by wintonzheng

AI Summary

This early-stage release focuses on CUA (Computer Use Agent) integration, URL encoding fixes, and removing legacy Task V2 API dependencies. It introduces new actions like move and drag, and GPT-4.1 support.

Key Highlights

  • Integration with CUA (Computer Use Agent).
  • Removal of Task V2 API.
  • Added MoveAction and DragAction support.
  • GPT-4.1 support and updated prompts.
  • Fix for OpenAI client initialization issues.

Breaking Changes

  • Removed Task V2 API.

New Features

  • CUA feature - move action.
  • CUA feature - drag action.
  • CUA feature - keypress combo.
  • CUA feature - extraction data combined.
  • CUA feature - skyvern answers CUA questions.

Full Release Notes

## What's Changed
* revert "url encoding fix" by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2135
* improve task generation data extraction goal by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2137
* Retry url encode fix by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2136
* fix open ai client initialization issue by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2138
* remove task v2 api by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2140
* sdk get_run bug fix by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2139
* shu/fallback to skyvern actions when cua returns no action by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2141
* add required action fields for cua fallback actions by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2142
* add move action in frontend by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2143
* add CUA MoveAction by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2144
* cua move action better reasoning by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2145
* cua keypress combo by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2146
* shu/cua keypress combo by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2147
* update generate-task prompt by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2148
* Add any information that helps from the last message of CUA when there's no data extraction goal by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2149
* set task_id in skyvern_context when task block starts by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2114
* remove n8n dependency by @LawyZheng in https://github.com/Skyvern-AI/skyvern/pull/2151
* fix auto completion bug by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2152
* Log response text on http post by @suchintan in https://github.com/Skyvern-AI/skyvern/pull/2153
* shu/cua other action by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2154
* better data extraction goal prompt for task generation by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2155
* no need to verify completion when we decide to complete according to CUA response; extract data based on CUA reponse as well by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2156
* CUA feature - extraction data combined assisstant's message by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2157
* Let URL be empty string if there is a browser session id by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2158
* CUA feature - support drag action by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2160
* Revert "set task_id in skyvern_context when task block starts" by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2161
* GPT-4.1-support by @suchintan in https://github.com/Skyvern-AI/skyvern/pull/2159
* null_action means doing nothing and taking a screenshot by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2162
* CUA feature - skyvern answers CUA questions and makes a decision if being asked to by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2163
* try to fix screenshot timeout by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2164
* extend custom select by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2167
* reset skyvern_context after task v2 block is done by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2168
* Revert "Revert "set task_id in skyvern_context when task block starts"" by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2169
* fix llm cost for gpt4.1 by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2172
* VerificationCodeAction in frontend by @wintonzheng in https://github.com/Skyvern-AI/skyvern/pull/2173


**Full Changelog**: https://github.com/Skyvern-AI/skyvern/compare/v0.1.75...v0.1.76