Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.3.0] — 2026-02-27

Added

  • GitHub Action (action.yml): publish-ready composite action for GitHub Marketplace — auto-syncs rules on .ai-rules.yaml push and commits the generated files
  • examples/sync-ai-rules.yml: copy-paste workflow template for teams
  • Remote Rules (docs/r/*.txt): 16 presets compiled into plain-text URLs for zero-install use in Cursor, Windsurf, and other AI IDEs
  • docs-build.sh now generates docs/r/{preset}.txt and adds a Remote Rules section to the Presets docs page

Changed

  • README: added "Remote Rules" and "GitHub Action" sections before Quickstart

[1.2.0] — 2026-02-27

Added

  • npm package: npx ds-agent-rules init / npm install -g ds-agent-rules (package.json, bin/cli.js)
  • pip package: pip install ds-agent-rules / ds-agent-rules sync (pyproject.toml, src/ds_agent_rules/)
  • pip-prepare.sh helper to stage data files before python -m build
  • npm and PyPI version badges in README
  • Makefile npm-publish and pip-publish targets

Changed

  • sync.sh and new-project.sh support RULES_DIR_OVERRIDE env var for package-managed installs
  • README & README_zh-TW quickstart sections now show npm / pip / git clone options
  • Installation section expanded to 4 methods (npm, pip, standalone clone, git submodule)

[1.1.0] — 2026-02-26

Added

  • MkDocs Material documentation site with GitHub Pages deployment (mkdocs.yml, docs/, docs-build.sh)
  • GitHub Actions workflow for automatic docs deployment (.github/workflows/docs.yml)
  • make docs and make docs-serve targets
  • README badges: CI status, license, release version, GitHub stars
  • "Who Uses This" section in README

Changed

  • vlm.md (52→62 lines): added numeric thresholds (CLIPScore, POPE F1 > 0.85), dataset validation rules, quantization guidance, EXIF orientation pitfall
  • edge-inference.md (57→67 lines): added power profiling, structured pruning detail, calibration & validation rules, tokenizer overhead pitfall, lifecycle handling
  • chinese-nlp.md (56→65 lines): added numeric benchmarks (DRCD F1, CER < 5%), hybrid search guidance, currency/font rules, zero-width char cleaning
  • responsible-ai.md (63→74 lines): added intersectional analysis, rate limiting, watermarking, differential privacy (ε ≤ 10), SLA for incidents, hallucination monitoring, accessibility for generated images

[1.0.0] — 2026-02-26

Added

Core Architecture

  • Layered rule composition: core → overlay → snippets → team
  • sync.sh — main sync engine with flags: --preset, --dry-run, --output-dir, --validate, --init, --team, --list, --diff
  • new-project.sh — interactive project initializer with scaffold support
  • .ai-rules.yaml — per-project config with auto-detection
  • Multi-tool output: CLAUDE.md, AGENTS.md, .github/copilot-instructions.md, .gemini/styleguide.md, .cursorrules, .windsurfrules

Base Overlays (6)

  • core.md — universal coding standards (always loaded)
  • ds-ml.md — Data Science / Machine Learning
  • llm-eng.md — LLM / GenAI Engineering
  • data-eng.md — Data Engineering
  • software-eng.md — Software Engineering
  • research.md — Research / Academic

Snippets (23)

  • agentic-ai, audio-speech, chinese-nlp, ctr-prediction, cv, data-labeling, distributed-training, edge-inference, evaluation-framework, graph-ml, jax, llm-finetuning, mlops, nlp-general, prompt-engineering, pytorch, rag, responsible-ai, streaming-ml, synthetic-data, tabular-ml, time-series, vlm

Presets (11+)

  • llm-project, agentic-ai, distributed-llm, cv-project, recsys-project, tabular-project, ts-forecast, nlp-project, research-llm, full-stack-ai, data-platform

Project Quality

  • CI/CD with GitHub Actions (ShellCheck + bats)
  • Bats test suite for sync.sh
  • CONTRIBUTING.md with snippet format spec
  • PR and Issue templates
  • Makefile for common tasks
  • .gitignore