Skip to content

contributing.md

Latest commit

 

History

History
94 lines (64 loc) · 3.02 KB

contributing.md

File metadata and controls

94 lines (64 loc) · 3.02 KB

Contributing to FigMirror

Thanks for helping make research figures less painful. FigMirror is easiest to improve when contributions stay close to a visible user workflow: upload a reference, paste data, inspect iterations, export a polished figure.

Setup

git clone https://github.com/VILA-Lab/FigMirror.git
cd FigMirror

mkdir -p .artifacts/uv-cache
export UV_CACHE_DIR="$PWD/.artifacts/uv-cache"
uv sync --group dev
uv run pytest

For UI work, run the mock backend first:

mkdir -p .artifacts/figmirror-dev
uv run python scripts/figcopy_serve.py \
  --workspace .artifacts/figmirror-dev \
  --backend mock

Open http://127.0.0.1:8765/.

Good First PRs

  • Add a showcase reference/output pair under docs/assets/showcase/ and update the README gallery.
  • Improve a small web UI interaction in scripts/figcopy_static/.
  • Add a regression test for runner state, chat logs, backend availability, or export behavior under tests/.
  • Clarify an install or usage path in README or scripts/README_figcopy_serve.md.
  • Add a compact eval case to a skill bundle when a visual failure repeats.

Project Areas

Web UI

Primary files:

  • scripts/figcopy_serve.py
  • scripts/figcopy_static/workspace.js
  • scripts/figcopy_static/trajectory.js
  • scripts/figcopy_static/style.css
  • scripts/figcopy_runner/

Use --backend mock for fast UI iteration. Use --backend codex or --backend claude when testing real generation.

Skills and Prompts

Primary files:

  • .codex/skills/figmirror/
  • .claude/skills/figmirror/
  • .claude/agents/figure-preprocessor.md
  • .claude/agents/figure-illustrator.md
  • .claude/agents/figure-critic.md

Development/history bundles may live under resources/prompts/, but prompt changes intended for runtime should land in the Codex/Claude skill files above.

Prompt changes should keep the L1/L2 grounding contract intact: visual decisions come from the reference image or aesthetic library, and the reviewer preserve list should survive into the next iteration.

Tests

Run the suite before opening a PR:

uv run pytest

Focused tests live under tests/figcopy_runner/. Add tests near the behavior you changed.

Showcase Assets

Use small, committed assets for README examples. Keep heavy run workspaces under .artifacts/ or a project-specific external volume, and avoid committing raw iteration folders, caches, or large datasets.

Recommended naming:

docs/assets/showcase/<short-case-name>-reference.png
docs/assets/showcase/<short-case-name>-generated.png

If you add a new image pair, include a short note in the PR explaining the chart family, source paper or reference, and what style property the example proves.

PR Checklist

  • The README still opens with the title, demo slot, showcase, and Quick Start.
  • New public-facing docs are short enough to scan.
  • UI changes were checked in a browser with the mock backend.
  • Prompt changes were tested on at least one concrete reference/data pair.
  • uv run pytest passes, or the PR explains the remaining failure.