Fix self-hosted Workflow Copilot safety model setup by sardortaylakov · Pull Request #8872 · Skyvern-AI/skyvern · GitHub
Skip to content

Fix self-hosted Workflow Copilot safety model setup - #8872

Draft
sardortaylakov wants to merge 1 commit into
Skyvern-AI:mainfrom
sardortaylakov:fix/copilot-safety-model-setup
Draft

sardortaylakov wants to merge 1 commit into
Skyvern-AI:mainfrom
sardortaylakov:fix/copilot-safety-model-setup

Conversation

@sardortaylakov

Copy link
Copy Markdown
Contributor

Description

Self-hosted Copilot rejects every message with missing_handler when only the main model is configured. Add explicit safety-model selection to skyvern init llm, deployment guidance, a doctor warning, and an actionable chat error. Preserve saved model choices, PostHog precedence, and fail-closed screening. Setup also consistently writes provider settings to the selected environment file.

Fixes #8849.

How Has This Been Tested?

  • 492 Python tests passed across Copilot routes/safety, model configuration, setup, quickstart, and doctor, using an isolated PostgreSQL database.
  • pre-commit run --all-files, including 6,975 passing frontend tests.
  • Live Gemini/Docker demo has not been recorded.

This branch has not been deployed

No deployments
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.

Workflow Copilot blocked by raw-secret safety screen with missing_handler in self-hosted Docker setup

1 participant