vibe-openscad

bare/qwen3-0.6b / tier-1-cube-with-hole

render_error tier 1
error
openscad stl render failed (exit 1): ERROR: Parser error: syntax error in file ../../../../../tmp/render-4eQLMH/input.scad, line 12
Can't parse file '/tmp/render-4eQLMH/input.scad'!

bare (qwen3-0.6b)

render
no render
render_error 21.6s · 86/2202t
openscad stl render failed (exit 1): ERROR: Parser error: syntax error in file ../../../../../tmp/render-4eQLMH/input.scad, line 12
Can't parse file '/tmp/render-4eQLMH/input.scad'!

SCAD (11 lines)
length = 50;
width = 50;

// Create two squares for left and right sides
left_square = square(length/2, width);
right_square = square(length/2, width);

// Add a circular hole in the center
hole_circle_centered_at_midpoint = circle(20, radius=10);

model = combine(left_square, right_square, hole_circle_centered_at_midpoint)

meta

runIdbare_qwen3-0.6b-81de998b47b4-2026-04-29T08-19-53-615Z
matrixIdbare/qwen3-0.6b
signature81de998b47b4d512…
provideropenai-self-hosted
modelqwen3-0.6b
harnessbare
statusrender_error
duration21.63s (wallclock: load + prompt eval + generation + network)
TTFT5.686s
generation time5.74s (pure decode, excludes load + prompt eval)
tokens86 in / 2202 out
generation tok/s383.6 tok/s (pure decode rate)
wallclock tok/s101.8 tok/s (end-to-end)
createdAt2026-04-29T08:19:53.615Z
gitCommit1d991ed9e773
openscadOpenSCAD version 2026.04.27

model metadata (self-hosted)

セルフホスト LLM はクラウドの単一 provider と違い、 実質「(モデル開発者, weight publisher, 量子化)」の組で挙動が決まる。 run 時に LM Studio / Ollama から取得した値:

publisherlmstudio-community
typellm
archqwen3
quantizationQ8_0
max ctx32,768 tok
capabilitiestool_use
host GPUNVIDIA GeForce RTX 3090 (Vulkan)
host VRAM24 GB
host CPU12th Gen Intel(R) Core(TM) i9-12900K
host RAM32 GB

prompt

OpenSCAD で、50mm 角の立方体の中央に、直径 20mm の貫通穴を z 軸方向に開けたモデルを作成してください。

完成したコード全体を ```openscad ... ``` のフェンスで囲んで出力してください。コードのみで、追加の説明は不要です。