fix(parametric): generate proper threaded geometry for screws/bolts

[zachdive]

Summary

Closes #42.

When the model generated a screw or bolt, the threaded region rendered as smooth cylinder or blank space, and the same gap shipped through to the STL export. Root cause: the LLM was hand-rolling threads with bare `cylinder()` / `linear_extrude()`, neither of which can represent helical thread geometry in OpenSCAD. At the default low `$fn`, the thread profile collapses into invisible slivers.

Fix

Adds explicit guidance + one worked example to the parametric-chat system prompt directing the model to use BOSL2 for any threaded fastener:

• `include <BOSL2/std.scad>` + the relevant module file (`screws.scad` for headed fasteners, `threading.scad` for plain threaded rod)
• Generate threaded geometry with `screw()`, `screw_hole()`, `threaded_rod()`, `nut()`, etc. — never bare primitives
• Set `$fn = 64` (96–128 for larger diameters) so threads actually resolve
• Expose diameter / length / pitch as snake_case parameters
• Use the BOSL2 spec string (`"M6x1"`, `"new images #8-32"`) so pitch is auto-picked

BOSL2 is already preloaded in the runtime (referenced from the README as a built-in library), so no library plumbing change is needed — only the prompt.

Test plan

• Manual: run "M8 hex bolt 30mm long" through the parametric chat — should now produce a model with visible helical threads (no blank gap), exportable to STL.
• Spot-check existing requests for nuts and threaded rods.

Suggested test prompts:

• `an M6 hex bolt 25mm with matching nut`
• `a threaded rod 8mm diameter 100mm long, 1.25mm pitch`
• `a plate with four M3 tapped holes`

Notes

This is a pure prompt change; no runtime, library, or schema changes. The change is in `supabase/functions/parametric-chat/index.ts` only.

Authored by Eve (Zach's AI agent) on behalf of Adam.

---

Summary by cubic

Fixes missing/blank threads by requiring BOSL2 for all screws, bolts, nuts, threaded rods, and tapped holes, and removing boolean thread overrides so models and STL exports render correct helical threads (fixes #42).

• Bug Fixes
  • Enforce BOSL2 includes and modules (std.scad, screws.scad, threading.scad); call screw(), screw_hole(), threaded_rod(), nut().
  • Remove boolean thread flags; use size strings like "M6x1" or "new images #8-32" and thread_len only for partial threads.
  • Forbid bare cylinder()/linear_extrude(); set $fn >= 64 (96–128 for larger diameters).
  • Add a minimal M8 hex bolt example.

^{Written for commit 5458cd3. Summary will update on new commits.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
cadam	Ready	Preview, Comment	May 11, 2026 11:48pm

[supabase]

This pull request has been ignored for the connected project sgprnbvihmydyrzvkcir due to reaching the limit of concurrent preview branches.
Go to Project Integrations Settings ↗︎ if you wish to update this limit.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

[cubic-dev-ai]

No issues found across 1 file

[greptile-apps]

Greptile Summary

This PR is a pure system-prompt change that steers the LLM away from bare cylinder()/linear_extrude() primitives toward BOSL2 modules (screw(), screw_hole(), threaded_rod()) for any threaded fastener request, fixing blank/smooth thread geometry in generated models and STL exports.

• Adds a "Threaded Fasteners (CRITICAL)" rule block covering required includes, banned primitives, $fn floor, snake_case parameter exposure, and the BOSL2 spec-string convention.
• Appends a worked M8 hex bolt few-shot example that the LLM can directly copy; critically, thread=true is absent — the bolt call is screw(bolt_spec, head=\"hex\", length=bolt_length) with no pitch override, so standard M8x1.25 geometry is produced correctly.
• No runtime, library, or schema changes; only the prompt string in supabase/functions/parametric-chat/index.ts is touched.

Confidence Score: 5/5

Safe to merge — pure prompt change with no runtime or schema impact, and the thread=true correctness issue from earlier review rounds is not present in this version.

The only change is the system-prompt string in a single Edge Function. BOSL2 guidance is technically accurate: includes are correct, the right modules are called, the previously-flagged thread=true boolean is gone, and $fn = 96 is appropriate for M8. The one finding is a naming nit in the worked example with no effect on runtime correctness.

No files require special attention.

Important Files Changed

Filename	Overview
supabase/functions/parametric-chat/index.ts	Adds BOSL2 threading guidance block and worked M8 bolt few-shot example to the system prompt. The thread=true bug from the previous review iteration is not present; inline and worked examples both call screw() without a thread argument. Minor: bolt_spec abbreviation contradicts the project naming rule baked into the same prompt.

_{Reviews (3): Last reviewed commit: "fix(parametric): remove boolean thread o..." | Re-trigger Greptile}

[cubic-dev-ai]

No issues found across 1 file