Skip to main content

VQL - Visual Query Language

before

uri2vql pipeline-analyze --image /tmp/screen.png --out app.vql.json --llm

1.png

after

uri2vql vql-to-image --program app.vql.json --out app.render.png

2.png

VQL objects app.vql.json

{
  "version": "1.0",
  "render_target": "svg",
  "scene": {
    "width": 2560.0,
    "height": 1600.0,
    "background": "#FFFFFF",
    "url": "file:///tmp/screen.png",
    "app": "desktop",
    "layers": [
      {
        "id": "screen_regions",
        "objects": [
          {
            "id": "region_000",
            "primitives": [
              {
                "shape_type": "rectangle",
                "params": {
                  "width": 213.0,
                  "height": 133.0
                }
              }
            ],
            "style": {
              "color": "#2F342E",
              "fill": true,
              "stroke_width": 0.0,

AI Cost Tracking

PyPI Version Python License AI Cost Human Time Model

  • 🤖 LLM usage: $11.5573 (8 commits)
  • 👤 Human dev: ~$659 (6.6h @ $100/h, 30min dedup)

Generated on 2026-06-24 using openrouter/qwen/qwen3-coder-next


VQL (Visual Query Language) — język wektorowego opisu fotografii, rysunków i zrzutów ekranu.

Samodzielna paczka Python z IR (VQLProgram), kompilatorem NL→VQL, walidacją i rendererami SVG/PNG. Sterowana przez DSL i bus CQRS/ES w paczkach *2vql.

Ekosystem: SUMD (opis) → DOQL (deklaracja) → Makefile/taskfile (automation) → testql (weryfikacja)

Dokumentacja: docs/README.md · Przykłady: examples/README.md · SUMD: SUMD.md

Instalacja

pip install vql              # core
pip install vql[png]         # eksport PNG (cairosvg)
pip install vql[desktop]     # zrzut ekranu (pillow, mss)
bash install-dev.sh          # pełny stack dev

Szybki start — rysowanie

from vql import nl_to_program, render_to_svg

program = nl_to_program("narysuj czerwone koło", width=400, height=400)
svg = render_to_svg(program)
dsl2vql -c 'COMPILE "narysuj niebieski kwadrat"'
rest2vql serve --port 8216

Szybki start — zrzut ekranu → VQL

# Automatyzacja LLM: vdisplay → imgl (bez dialogu GNOME, gdy mirror działa)
imgl capture -o /tmp/screen.png --verify --analyze --lang eng+pol

# Alternatywa: portal GNOME (docs/desktop-capture.md)
uri2vql capture-screen --interactive --out /tmp/screen.png

# All-in-one
uri2vql capture-and-analyze --out app.vql.json --diagnose

# Lub krok po kroku
uri2vql analyze-window --image /tmp/screen.png --out app.vql.json --grid 12
uri2vql adopt-ui --image /tmp/screen.png --out ui.vql.json --locale pl
uri2img2svg query "img2svg://svg?path=/tmp/screen.png&out=/tmp/screen.svg"

uri2vql query "vql://window/summary?file=app.vql.json&live=1&image=/tmp/screen.png"

Pełny skrypt: bash examples/full-pipeline.sh

Fotografia ↔ VQL (roundtrip)

Opis z obrazu i odtworzenie z metadanych — zależnie od typu obrazu:

# Test roundtrip (próbki + raport fidelity)
python examples/photo-roundtrip-test.py
make test-roundtrip

# Opis mozaiką kolorów
img2svg vql photo.png --out photo.vql.json --grid 20

# Wektoryzacja kolorowa (lepsza dla płaskich kształtów)
pip install 'img2svg[vtracer]'
img2svg vql photo.png --out photo.vql.json --method vtracer

# Odtworzenie
python -c "
from vql import VQLProgram, render_to_png
import json
p = VQLProgram.from_dict(json.load(open('photo.vql.json')))
render_to_png(p, 'reconstructed.png')
"

Macierz jakości i biblioteki: docs/photo-roundtrip.md

Interfejsy

Warstwa Window / screenshot SVG
CLI uri2vql, img2vql img2svg, uri2img2svg, dsl2img2svg
URI vql://window/* img2svg://vectorize|svg|vql
REST POST /v1/window/* :8216 planowane
MCP vql_detect_ui, vql_diagnose_window, … —

REST: docs/rest-window-api.md · URI: docs/window-uri.md · SVG: docs/img2svg-uri.md

Desktop adopt + img2nl

Etap Narzędzie Co robi
Capture imgl capture (vdisplay) / uri2vql capture-screen PNG + provenance / portal
Grid adopt analyze-window scalone regiony + fingerprint
UI detect img2vql / adopt-ui okna, przyciski, bbox, relations
Wektoryzacja img2svg / uri2img2svg PNG → SVG / VQL
Diagnose diagnose-window routing LLM + auto-OCR
Compare compare-window phash — zmiana ekranu?
OCR semantyczny adopt-imgl / imgl tekst + role + interakcja

Pipeline: docs/window-pipeline.md · LLM capture: docs/vdisplay-imgl-automation.md · Portal: docs/desktop-capture.md

Przykłady

bash examples/live-capture-test.sh
bash examples/full-pipeline.sh
bash examples/img2nl-vql-flow.sh /tmp/screen.png
VQL_TEST_IMAGE=/tmp/screen.png bash examples/full-pipeline.sh

Development

make install-dev    # pełny stack *2vql
make test           # pytest core + dsl2vql
make test-all       # wszystkie pakiety
make serve          # rest2vql :8216
make goal           # test + commit + publish (goal.yaml)
bash project.sh     # SUMD + DOQL + testql + analiza

Zmienne LLM: skopiuj .env.example → .env (OPENROUTER_API_KEY, LLM_MODEL).

Struktura monorepo

src/vql/              — domena (schema, compiler, renderers, adopt/window)
packages/
  dsl2vql/ uri2vql/ nlp2vql/ cli2vql/ rest2vql/ mcp2vql/
  img2vql/            — detekcja UI + img2nl diagnose (v0.2)
  img2svg/            — obraz → SVG / VQL (v0.1)
  uri2img2svg/        — img2svg:// URI (v0.1)
  dsl2img2svg/        — VECTORIZE / TO_VQL DSL (v0.1)
docs/                 — dokumentacja
examples/             — skrypty demo (README w examples/)
testql-scenarios/     — kontrakty testql (CLI, pytest)

Skala (SUMD): 151 modułów · 278 funkcji · ~12.5k LOC · CC̄=4.2

Szczegóły: packages/README.md · docs/packages.md

Status

License

Licensed under Apache-2.0.

Metadata

Release files for vql 0.1.7

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for vql 0.1.7
File Size Uploaded
vql-0.1.7.tar.gz 3.3 MB Details

Built distribution (wheel)

Table of built distributions (wheels) for vql 0.1.7
File Interpreter ABI Platform
vql-0.1.7-py3-none-any.whl Python 3 none any Details

Total release size: 3.3 MB

Release files / vql-0.1.7.tar.gz

Download URL vql-0.1.7.tar.gz
Size 3.3 MB
Tags Source
SHA-256 checksum
How to use checksums
772391182dd40c5750c319475a9b540da8dcd488ad14c95c99b6a9fa0df06ec2
BLAKE2b-256 checksum
How to use checksums
e80a35098663aea00e1aabfda8507cf60fce530d226fddc5b80f1268e5c7f8ce
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.13.7

Release files / vql-0.1.7-py3-none-any.whl

Download URL vql-0.1.7-py3-none-any.whl
Size 81.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
43b268548ce0c1a764da6f7f7dc5e951ea343a3b3ae3243af3301d5b5bc900ce
BLAKE2b-256 checksum
How to use checksums
9a7b6e89afc8e1c25c6592cf0d32ee255497cdda78150c8370bc18aa3d9b4125
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.13.7

Release history Release notifications | RSS feed

This release

0.1.7 This release

2 release files

0.1.6

2 release files

0.1.5

2 release files

0.1.4

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page