Skip to content

fix: improve action parser compatibility - #252

Open
Rohithmatham12 wants to merge 1 commit into
bytedance:mainfrom
Rohithmatham12:codex/action-parser-compat
Open

fix: improve action parser compatibility#252
Rohithmatham12 wants to merge 1 commit into
bytedance:mainfrom
Rohithmatham12:codex/action-parser-compat

Conversation

@Rohithmatham12

Copy link
Copy Markdown

Summary

  • support Doubao seed compact action output like click>point>point>x y by normalizing it into the existing structured action flow
  • parse documented <|box_start|>...<|box_end|> coordinate strings and comma/space separated point tags
  • replace pyautogui coordinate eval usage with safe coordinate parsing
  • generate a drag gesture for scroll(start_box=..., end_box=...), which is the common mobile/emulator scroll form
  • document supported parser formats and clarify that OpenAI-compatible deployment endpoints should include the /v1 suffix

Issues

Addresses #229.
Addresses #231.
Helps clarify #122.
Helps the deployment confusion discussed in #127.

Testing

python3 -m unittest tests.action_parser_test
python3 -m py_compile ui_tars/action_parser.py tests/action_parser_test.py
git diff --check

Note: make test could not run in my local environment because uv is not installed. Running the same parser unittest target directly passed; full unittest discovery additionally requires Pillow for tests/inference_test.py.

@CLAassistant

CLAassistant commented Aug 16, 2026

Copy link
Copy Markdown

CLA assistant check
All committers have signed the CLA.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants