Add automatic API key sanitization to prevent sensitive data exposure in error messages and logs.

This addresses a security vulnerability where LLM API keys were previously exposed, reducing the blast radius if logs are shared or leaked.

Slack Thread

Summary by cubic

Adds automatic sanitization of API keys and other sensitive fields in exceptions, logs, and background task errors to prevent credential exposure. Introduces a sanitize_sensitive_data utility and enables redaction by default across the library.

• New Features

  • Sanitizes error messages in LLMException and ModelProviderError.
  • Sanitizes all log messages and arguments via BrowserUseFormatter.
  • Sanitizes background task exceptions in create_task_with_error_handling.
  • Exposes sanitize_sensitive_data in the public API for manual use.
  • Adds docs, example, and tests for sanitization.

• Migration

  • No changes required; sanitization is automatic.

^{Written for commit 12384ce. Summary will update automatically on new commits.}