LLM_INSTRUCTIONS.md

Universal LLM Instructions

This file serves as a central repository for instructions tailored to Large Language Models interacting with this codebase.

Core Directives

1. Versioning

   • The project version is stored in the VERSION and VERSION.md files.
   • Always increment the version for new builds or significant changes.
   • Update CHANGELOG.md with every version bump.
   • Reference the version in commit messages when bumping versions.

2. Submodules

   • ddc_onset: located at ddc_onset/
   • ffr-difficulty-model: located at ffr-difficulty-model/
   • Ensure submodules are synced, committed, and pushed when modified.
   • Keep docs/DASHBOARD.md aligned with submodule/project status changes.

3. Code Style

   • Target Python 3.8+ compatibility where practical.
   • Follow PEP 8 and prefer descriptive naming.
   • Be aware that the repository currently contains both legacy TensorFlow-oriented paths and newer PyTorch-oriented work for the active environment.

4. Testing / Validation

   • Validate command-line entry points when touched.
   • Prefer practical sanity checks for the active environment.
   • Keep documentation clear about what is validated versus what is only planned.

5. Documentation

   • Keep root/docs dashboards current.
   • Record substantial findings in docs/.
   • Update HANDOFF.md after meaningful progress.

Project Structure

• autochart.py: main CLI entry point
• infer/: inference logic and library code
• learn/: training logic
• scripts/: orchestration/utilities
• dataset/: data processing utilities
• docs/: project analysis, audit, and planning documentation

Model-Specific Notes

Claude (CLAUDE.md)

• Focus on architectural integrity and detailed explanations.
• Explain rationale and impact for meaningful changes.

Gemini (GEMINI.md)

• Use large-context analysis to understand project-wide implications.
• Prioritize efficient modernization and broad codebase synthesis.

GPT (GPT.md)

• Focus on practical implementation, architecture cleanup, and debugging.
• Keep changes actionable and consistent with the active repo state.

GitHub Copilot (copilot-instructions.md)

• Use version-aware commit messages when appropriate.
• Keep project structure/documentation in sync with implementation changes.

Handoff Protocol

When handing off work:

1. Summarize current project state.
2. List active tasks and known issues.
3. Reference docs/DASHBOARD.md and major audit/planning docs as needed.
4. Ensure changes are committed and pushed when appropriate.