Violit Drawable Konva

Violit widget wrapping the same Konva canvas used by streamlit-drawable-konva (0.3.0 tracks Streamlit 0.6.x standalone).

Requires Violit ≥ 0.8.29 (app.register_js_widget).

Features

• vl_canvas — drawable Konva canvas (same tools as st_canvas)
• In-canvas tool picker — tools=[…] + display_tool_picker=True
• vl_image_comparison — before/after slider companion (same as st_image_comparison)
• Background color / image (upload or sample)
• Scene json_data with coordinates (x/y/width/height/points)
• Spline helpers: splines_from_json, sample_spline, …

Demo uses dark theme. Pass Violit State / callables into vl_canvas (not .value) so drawing mode and stroke update live. With display_tool_picker=True, tools can also be switched on the canvas itself.

Install (editable)

cd /devel/dev/cvrlab/violit-drawable-konva
uv venv .venv
source .venv/bin/activate
uv pip install -e ".[dev]"

Sync the browser IIFE from the Streamlit sibling build (required before demo / packaging):

# ensure standalone.js exists in streamlit-drawable-konva first
bash scripts/sync_standalone.sh

Until you sync, the package falls back to:

../streamlit-drawable-konva/streamlit_drawable_konva/frontend/build/standalone.js

Demo

source .venv/bin/activate
.venv/bin/violit run demo/app.py --reload --localhost --port 8031
# → http://localhost:8031

demo/app.py must end with app.run() (Violit does not auto-start the server).

Tests

pytest -q

Usage

import violit as vl
from violit_drawable_konva import ensure_registered, vl_canvas, vl_image_comparison

app = vl.App(title="Canvas", theme="dark")
ensure_registered(app)  # optional; also called by vl_canvas / vl_image_comparison

drawing_mode = app.state("polygon", key="drawing_mode")
stroke_width = app.state(3, key="stroke_width")
stroke_color = app.state("#111", key="stroke_color")
bg_file = app.state(None, key="bg_file")
payload = app.state({"image_data_url": None, "json_data": None}, key="c1")

app.file_uploader("Background", type=[".png", ".jpg"], bind=bg_file)

result = vl_canvas(
    app,
    drawing_mode=drawing_mode,  # State, not .value
    stroke_width=stroke_width,
    stroke_color=stroke_color,
    background_image=bg_file,   # UploadedFile / PIL / State / callable
    height=400,
    width=600,
    tools=["freedraw", "line", "rect", "polygon", "transform", "pan"],
    display_tool_picker=True,
    key="c1",
    bind=payload,
)
# result.json_data / result.image_data

vl_image_comparison(app, img_before, img_after, width=600, key="cmp")

Helpers mirror the Streamlit package: crop_box_from_json, objects_by_group, splines_from_json, spline_control_points, sample_spline, sample_catmull_rom.

Layout

violit_drawable_konva/
  api.py                # ensure_registered, vl_canvas, vl_image_comparison
  payload.py / spline.py / helpers.py
  static/standalone.js  # synced IIFE (mount + mountImageComparison)
demo/app.py
tests/
scripts/sync_standalone.sh

Publish (PyPI)

cd /devel/dev/cvrlab/violit-drawable-konva

# 1) Sync JS from Streamlit build (after streamlit 0.6.x is built)
bash scripts/sync_standalone.sh

# 2) Tests
.venv/bin/python -m pytest -q

# 3) Commit & push
git add -A
git status
git commit -m "Release 0.3.0: in-canvas tool picker and tools allow-list."
git push -u origin main
git tag v0.3.0 && git push origin v0.3.0

# 4) Build & publish
uv build
uv publish   # needs UV_PUBLISH_TOKEN or interactive PyPI token

Verify: https://pypi.org/project/violit-drawable-konva/

Bump version in pyproject.toml before each new upload (PyPI versions are immutable).

Quick PyPI update (0.2.0 → 0.3.0)

bash scripts/sync_standalone.sh
uv build && uv publish
pip install -U violit-drawable-konva   # expect 0.3.0

Manual port checklist

1. Build Streamlit frontend (npm run build) so standalone.js exists.
2. bash scripts/sync_standalone.sh
3. uv pip install -e ".[dev]" (violit>=0.8.29)
4. pytest -q
5. violit run demo/app.py --reload --localhost --port 8031
6. Commit / tag / uv publish (see above)