Skip to content

v1.1.0

Choose a tag to compare

View all tags
@GenDosa GenDosa released this 12 Mar 11:54
· 18 commits to main since this release
v1.1.0
2a1d132

Changelog

What's New

The first commit after checkpoint --init no longer wastes LLM credits describing its own config files. --init now writes a boilerplate MASTER_CONTEXT.md, and the CI workflow skips config-only commits. Real content populates after meaningful code is merged.

Security

  • Shell injection via PR titles — PR title was interpolated directly into bash; now passed through env: block
  • SSL verification on by default — set CHECKPOINT_SSL_VERIFY=false to opt out

Bug Fixes

  • get_diff() returns actual patch text instead of GitPython Diff object repr
  • graph.py no longer resets user's LLM config with a no-arg configure_llm() call
  • Ollama provider no longer blocked by API key check
  • max_tokens inline template default fixed from 2000 to 8000
  • Whitespace typo in config.py (config. model_dump)

Improvements

  • Diffs over 12k chars truncated before LLM call
  • Workflow concurrency groups cancel redundant runs
  • Catchup generation skipped when no checkpoints exist yet
  • Rate limit between catchup users reduced from 60s to 5s
  • API key pre-flight check in CI — fails fast with clear error message
  • Windows get_file_tree() fallback via pathlib

Cleanup

  • Removed dead config flags: vector_db, git_hook
  • Removed deprecated functions: get_active_authors, generate_file_dependency_mermaid, generate_class_hierarchy_mermaid
Assets 2
Loading