irgolic/AutoPR

Built with Guardrails for structured LLM output validation, AutoPR used JSON Schema constraints to enforce code generation adherence, reasking the model when outputs violated the schema. Triggered by GitHub issue labels, it executed a four-step pipeline: planning, code generation, branch pushing, and pull request creation. The project pioneered autonomous PR generation before grammar-constrained language models became standard.

1,370 stars.