Skip to content

fix: restore the prototype and establish regression coverage - #1

Open
ApexAiOfficial wants to merge 8 commits into
github:mainfrom
ApexAiOfficial:revival/pr-1-restore-prototype
Open

fix: restore the prototype and establish regression coverage#1
ApexAiOfficial wants to merge 8 commits into
github:mainfrom
ApexAiOfficial:revival/pr-1-restore-prototype

Conversation

@ApexAiOfficial

Copy link
Copy Markdown

Summary

This PR restores the Git SMS proof of concept to a working, testable baseline without expanding it into production deployment or GitHub Actions control.

The current application fails during import because natural_language_router.py imports create_repo and create_issue from commands.py, where those functions are not defined. The implementation also contains duplicated logic, eager model initialization, inconsistent configuration, unbounded GitHub requests, and no regression coverage.

Changes

  • fixes the import and startup path
  • consolidates command and summarization logic into canonical modules
  • allows credential-free, network-free module imports and startup
  • keeps repository and issue creation explicitly disabled
  • adds bounded GitHub API requests and safe TwiML error responses
  • supports repository, latest-issue, and numbered-issue summaries
  • prevents pull requests from being treated as issues
  • makes LLM and tokenizer initialization lazy
  • bounds repository and issue-thread prompts before LLM submission
  • updates setup and capability documentation while preserving the original project vision
  • adds Ruff and pytest CI for Python 3.10–3.12

Scope

This remains a read-only proof of concept.

Twilio signature verification, sender authorization, MessageSid idempotency, rate limiting, durable audit logging, GitHub App authentication, and GitHub Actions control are intentionally deferred to separate follow-up changes.

Validation

  • Ruff passes
  • 17 tests pass
  • CI passes on Python 3.10, 3.11, and 3.12

A self-review of the initial implementation found two important coverage gaps: the import test did not prove behavior in a fresh process, and the natural-language write paths lacked direct regression tests. Both are now covered.

Happy to rescope or split this differently if another first step would be easier to review. I’m opening this as a concrete starting point rather than assuming this is exactly the direction maintainers prefer.

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