Skip to content

feat(studio): add end-to-end Skill creation mode - #711

Merged
yaozheng-fang merged 8 commits into
mainfrom
feat/studio-skill-creation
Jul 22, 2026
Merged

feat(studio): add end-to-end Skill creation mode#711
yaozheng-fang merged 8 commits into
mainfrom
feat/studio-skill-creation

Conversation

@yaozheng-fang

@yaozheng-fang yaozheng-fang commented Jul 22, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • add Agent, temporary-session, and Skill-creation modes to the new-chat composer
  • use separate SANDBOX_CHAT_CODEX and SANDBOX_SKILL_CREATOR CodeEnv Tools; unavailable modes are disabled with an administrator configuration hint
  • create only new Sessions at runtime; run Doubao and DeepSeek Skill candidates in parallel and persist cross-instance status
  • stream public reasoning summaries, tool activity, file changes, and final replies into the shared conversation renderer
  • keep candidate stage status above the conversation while rendering only Sandbox conversation events below it
  • add the Skill Beta badge, model-aware placeholder, consistent tool typography, and the updated new-chat layout
  • validate, preview, download, and publish generated Skill archives through AgentKit
  • auto-provision two independent Ready CodeEnv Tools during initial cloud deployment when IDs are omitted
  • preserve existing cloud environment variables during Studio update and override only explicitly supplied Tool IDs
  • include AgentKit Skill center pagination, runtime management layout updates, documentation, and bundled WebUI assets

Security

  • keep raw Ark credentials out of the browser, VeFaaS environment, and per-job Session variables
  • bind only hosted credential tickets and relay URLs to the dedicated CodeEnv Tools
  • require an HTTPS Volcengine credential relay
  • redact environment secrets, structured credentials, bearer tokens, and JWTs from failure output
  • validate archive paths, symlinks, UTF-8 content, file count, and expanded size

Validation

  • frontend tests: 136 passed
  • CLI tests: 194 passed
  • focused backend regression tests: 73 passed
  • Ruff passed
  • focused Pyright passed with 0 errors
  • frontend and documentation production builds passed
  • frontend README Markdown lint passed
  • commit hardcoded-secret scan passed
  • live temporary chat produced thinking, command lifecycle, and final reply events; the Session was deleted
  • live AgentKit A/B generation completed for both models; both candidates succeeded, the ZIP passed integrity validation, and test Sessions were deleted

# Conflicts:
#	frontend/README.md
#	frontend/src/App.tsx
#	pyproject.toml
#	veadk/webui/assets/CodeEditor-C4sQGBmQ.js
#	veadk/webui/assets/CodeEditor-DZ9F0sL6.js
#	veadk/webui/assets/CodeEditor-jNiMd4Xa.js
#	veadk/webui/assets/MarkdownPromptEditor-CvXMyNxy.js
#	veadk/webui/assets/MarkdownPromptEditor-Cy35qj6G.js
#	veadk/webui/assets/MarkdownPromptEditor-Dd-SgWF0.js
#	veadk/webui/index.html
yield 'event: done\ndata: {"reason": "failed"}\n\n'

return StreamingResponse(
_stream(),
@yaozheng-fang
yaozheng-fang merged commit e99dc84 into main Jul 22, 2026
16 checks passed
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.

3 participants