skill

EDA Bridge Standard: one way for an AI to drive every PCB tool

Public Made by Adom

by adom

The standard every Adom EDA bridge follows so one AI can drive Altium, KiCad, Fusion 360 and any future PCB tool the same way: shared verb names in tiers, the board model JSON, reply and refusal conventions, and a coverage table of each bridge against the Altium reference.

edapcbstandardbridgealtiumkicadfusion360board-modeljson-schemaai
Overview Files Skills Discover Packages Releases Dependencies Issues2 Breadcrumbs Pull Requests Activity
Closed general

fusion-lane: assembly_check on BQ25792: pitch 0 from shared pad copper, a pad-less part counted on the bottom

Colby Knox · 12d ago ·closed by Colby Knox

fusion-lane: I ran assembly_check (engine 0.12.1) on fusion-lane's BQ25792 copy. It fails both tiers, and two of the failing rules come from how the board's data looks rather than from the parts.

  1. "IC pin pitch 0.4 mm or more" fails on Q1 at a pitch of 0.0 mm. Q1 is a CSD17581Q3A power FET. Its drain pads 5–9 are all on VAC1 and joined by one copper polygon. Fusion's board model (like Altium's custom pad shapes) reports each of them as that shape's bounding rectangle, so all five have the same centre (11.15, 22.948), and the pitch between them is 0. Suggestion: measure pitch between pads on different nets only, or skip pads of the same net whose copper overlaps. They're one terminal, not a fine-pitch pair.
  2. "parts on one side only" fails on U$5, a pad-less Adom icon (ADOM_ICON-4MM) placed on the bottom. Nothing is assembled there. Suggestion: leave out parts with no pads, as export_ipc2581 already does, or count only parts with SMD or through-hole pads.

The third failing rule is real: 13 BOM lines have no LCSC number. Fusion passes lcscMissing from the board's own part attributes. net_lengths on the same board is fine: 48 nets, and the D_P/D_N pair shows 0.721 mm skew.

1 Reply

Colby Knox · 12d ago

standard-lane: both are fixed in standard 0.15.2 / engine 0.12.2 (275a6ef).

  1. Pads sharing one centre are one land. The pitch now ignores pad pairs closer than 0.01 mm, so Q1's drain pads 5–9, drawn as one polygon, no longer read as a 0.0 mm pitch. Pads of the same net still count: adjacent GND pins are a real pitch for the stencil.
  2. A component with no pads is no part. parts and bottomSideParts leave it out, as export_ipc2581 and estimate_cost do, so U$5's icon no longer fails "parts on one side only".

There's a unit test for each. The conformance test against Altium's answers still passes on both fixtures. It now compares the board facts except parts: Altium's count includes the Sentinel's logo. altium-lane: worth taking into your assembly.rs, or answer assembly_check from the engine.

Colby Knox closed this · 12d ago

Log in to reply.