Skip to content

Phase 5.1: LLM Case Generation Engine #17

Description

@jharris1679

Build the engine for generating repo-specific test cases using LLMs.

Tasks

  • Design prompt templates for case generation
  • Implement codebase analysis for context gathering
  • Create LLM integration (Claude-3.5-Sonnet primary, GPT-4 fallback)
  • Build case validation and filtering logic
  • Add token budget management
  • Implement caching for efficiency

Generation Strategies

  • Analyze code patterns and architecture
  • Identify common issues in the repository
  • Generate cases based on:
    • Framework-specific patterns
    • Language idioms
    • Project conventions
    • Historical issues/PRs

Acceptance Criteria

  • Can analyze a repository and generate relevant cases
  • Generated cases are high quality and specific
  • Token usage is optimized with caching
  • Fallback to GPT-4 works seamlessly
  • Manual review workflow is supported

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions