Skip to content

feat: AppGuide support#21

Closed
saikrishna321 wants to merge 4 commits into
mainfrom
appcard
Closed

feat: AppGuide support#21
saikrishna321 wants to merge 4 commits into
mainfrom
appcard

Conversation

@saikrishna321
Copy link
Copy Markdown
Member

This pull request mainly improves code clarity, updates tool usage for compatibility, and enhances the formatting and readability of the documentation and HTML usage guide. The most significant changes are grouped below.

Codebase updates and compatibility improvements

  • Updated the call to list apps in AppResolver to use the new 'appium_app_lifecycle' tool with { action: 'list' } for compatibility with recent appium-mcp versions (src/agent/app-resolver.ts).
  • Changed the tap-by-coordinates implementation in tapAtCoordinates to use the newer 'appium_gesture' tool, and removed the fallback to 'mobile: clickGesture' for improved reliability and future-proofing (src/agent/element-finder.ts). [1] [2]
  • Added getScreenSizeForStark import and loadAppGuide import to loop.ts, and introduced variables to track the last app guide and active app ID for better app guide loading logic (src/agent/loop.ts). [1] [2] [3]
  • Changed episodic memory status logging to only print when debug mode is enabled, reducing output noise (src/agent/loop.ts).

Documentation and formatting improvements

  • Reformatted and prettified the HTML tables and sections in the usage guide for better readability and accessibility, without changing content or meaning (landing/usage.html). [1] [2] [3] [4] [5] [6]
  • Updated the CHANGELOG.md to use consistent list formatting for features and bug fixes.

saikrishna321 and others added 4 commits April 20, 2026 16:26
Co-authored-by: Srinivasan Sekar <srinivasan.sekar1990@gmail.com>
- Replace static "Reasoning…" spinner with randomly rotating fun verbs
  (Brewing, Cogitating, Pondering, etc.) that change every 2.5s
- Add step counter to spinner detail: (1/30 · vision · thinking on · model)
- Move verbose debug output behind MCP_DEBUG=1 flag:
  - Episodic memory status bullets
  - AppGuide injection/active bullets
  - "Pulling UI state" / "Consulting agent" bullets
  - LLM reasoning text (streaming and static)
- Remove misleading static 0/30 progress bar from goal box

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-authored-by: Srinivasan Sekar <srinivasan.sekar1990@gmail.com>
…ow improvements

- Thread AppGuide through planner and orchestrator for app-aware goal decomposition
- Add find_and_long_press meta-tool with vision and DOM mode support
- Migrate appium_click calls to appium_gesture for consistency
- Improve vision coordinate scaling with async screen size fetch
- Add natural language long_press step parsing in YAML flows
- Enhance preprocessor with appId tracking for AppGuide
- Update prompts with AppGuide context injection
- Various fixes across MCP client, device session, and flow execution

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-authored-by: Srinivasan Sekar <srinivasan.sekar1990@gmail.com>
Co-authored-by: Srinivasan Sekar <srinivasan.sekar1990@gmail.com>
@saikrishna321 saikrishna321 changed the title feature: AppGuide support feat: AppGuide support Apr 23, 2026
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.

1 participant