cad-agent (Valpatel) errors

22 recurring problems engineers report in cad-agent (Valpatel), ranked by impact.

#14 by errors trackedMCADECADVerification

Recurring cad-agent (Valpatel) errors, by impact

  1. 1An agent-built robot turret rendered its ammunition outside the hopper and its subassemblies unconnected, and per-part FEA and a clash sweep caught neither problem.MCAD · Agent failure · 1 report74
  2. 2Before this, the fuse board had 6 hand-written footprints and only 7 of 43 footprints had a 3D model. The new parts library builds footprints and STEP bodies from datasheet numbers, each tied to the page it came from, with a registration check confirming the body sits on the board, pins go through their pads' holes and pin 1 lands on the right pad. After the change, 37 of 43 parts have a body, and the other 6 are waived with reasons.ECAD · Agent failure · 1 report74
  3. 3For a McMaster heat-set insert with a 4.0 mm maker's hole, the agent's hole_design module drew 4.20 mm, while fit_coupons centred its test sweep on 4.25 mm. The author traced the right number to the CNC Kitchen test data already cited in the repo and fixed the coupon. Two modules in one agent, same part, two numbers. This is the shared-source-of-truth problem at the scale of a single repo.MCAD · Agent failure · 1 report74
  4. 4OCCT Common and Cut booleans missed a 4 mm sphere pushed 0.03 to 0.1 mm into a cone wall at 5 of 50 positions, reading 0 mm3 outside, so the volume cross-check could not see it.MCAD · Agent failure · 1 report74
  5. 5OCCT's boolean can miss a shallow crossing between two nearly tangent curved faces, and in a ball-through-cone sweep 24 of 500 cases were misread as contains or touching, hiding a clash.MCAD · Agent failure · 1 report74
  6. 6On cad-agent excavator project, fusing nine nearly coincident swept solids in one call gave an invalid shape. The invalid tool then cut nothing from the boom, and only a range-of-motion collision check caught it, 14.7 cm3 of overlap. A cone meeting a pin hole raised Null TopoDS_Shape until the tool was nudged 0.01 mm, and a sub-micron split vertex made a valid solid fail the watertight export.MCAD · Agent failure · 1 report74
  7. 7OpenCASCADE booleans can fail without raising, and isValid is not enough: in OCCT 7.9 a three-way fuse of nearly coincident swept solids returns an empty result. They now check every boolean against volume bounds and retry cuts nudged by 0.01 mm, and weld exported STLs at 1 um because a valid solid can still export as an open STL.MCAD · Agent failure · 1 report74
  8. 8PrusaSlicer exits 0 and writes G-code even for parts it flags with Detected print stability issues or WipeTower path conflicts, so the agent slice step reported those parts as fine. The fix parses the log and only calls a part ok with no errors or stability issues.MCAD · Agent failure · 1 report74
  9. 9A CAD-agent toolkit corrected its own research notes after finding they overstated OrcaSlicer's MCP support. The 2.5 beta only has an experimental, off-by-default socket for one tab with no slicing commands, and the general server is still a draft PR, so their decision is nothing to pilot yet.MCAD · Agent failure · 1 report66
  10. 10A fastener calculation took a flank angle in radians, so passing 60 as degrees silently gave about 3 times the nut factor K, until a release reviewer found it.MCAD · Agent failure · 1 report66
  11. 11A project owner built a farm robot with cad-agent 0.31: welded steel chassis, bought motors, gearboxes and battery, 35 parts. The agent assumes every part is 3D printed, so it produced 35 failing 'Print files' rows that do not apply, and the full run stopped at the 3MF export. Turning print checks off is not the answer because the next revision adds a few FDM parts. The agent has no idea how each part is made.MCAD · Agent failure · 1 report66
  12. 12A range-of-motion collision check found four real collisions on a 46-part project. Before it, nothing between sampled poses was checked, and a boolean OCCT gets wrong without raising could read as a clear pair.Verification · Agent failure · 1 report66
  13. 13cad-agent adds PrusaSlicer. Flattening PrusaSlicer INI files by hand gave each filament setting once, so G-code for a five-tool XL named a filament for tool 1 only. Fix: let PrusaSlicer resolve presets itself with --save, which also refuses presets that do not fit the printer.MCAD · Agent failure · 1 report66
  14. 14cad-agent's common-issues doc listed impossible cable bend radius as review-only, and the new check refuses a bare number for bend radius because mm and multiples of OD differ by a factor of the diameter.Verification · Agent failure · 1 report66
  15. 15Fresh agents following the README hit two walls: the inserts check writes a verdict string ('pass', 'warn', 'fail') but no boolean, and 'fail' is truthy, so a checks row cannot read it. And a PLA lid clamped on an M3 insert caps torque at 0.20 N m with nothing saying what lifts the cap.MCAD · Agent failure · 1 report66
  16. 16PrusaSlicer reported long bridging and floating anchors on a big part without saying where, which meant guessing and rebuilding. Reading the G-code showed most were internal bridges up to 197 mm under a 2.5 mm cab roof, fixable in slicer settings, while the real ones were a 58 mm arch overhang and a bushing over a chamfer.MCAD · Agent failure · 1 report66
  17. 17Rebuild of an outside contributor checked booleans: fusing three swept tubes 1 um apart in OCCT 7.9 returns an empty compound that isValid passes. Shape.Volume can be off by several mm3 on curved, trimmed faces, enough to fail a good cut, and an inverted box also passes isValid.MCAD · Agent failure · 1 report66
  18. 18Snap-fit allowables were cited from V2.0 filament data sheets hosted by a reseller, while Bambu's V3.0 sheets show weaker results, and V3.0 PETG Basic is a reformulated filament sold under the same name. PETG-CF strain dropped from 1.68% to 1.00%, and the tool now keeps every edition and names the edition behind each number.MCAD · Agent failure · 1 report66
  19. 19Sourced FDM and sheet-metal process rules had sat in research/process-profiles.yaml since an earlier issue, but nothing read them. This PR adds the loader, a minimum-feature DFM check and a bend calculator, and unverified rows are refused.MCAD · Agent failure · 1 report66
  20. 20Step-by-step assembly playback could show an arriving part crossing an earlier one while the final pose was clear. The fix checks the actual insertion path from offset to seat against parts in install order, bound to hashes of the guide, model and meshes, and missing or stale evidence fails visibly. A follow-up, #517, says the real collection now has 233 instances and can need 27,028 ordered pair checks.Verification · Agent failure · 1 report66
  21. 21The slicer names a problem like Long bridging extrusions but not where it is. They added overhang regions with location and a G-code bridge report so the agent can point at the feature.MCAD · Agent failure · 1 report66
  22. 22The 3MF writer stamps zip entries with the write time, so rebuilding unchanged geometry yields different bytes, and the slice report keyed on sha256 marks every rebuild stale, so the whole set has to be sliced again.MCAD · Agent failure · 1 report58

How it fails

  1. Agent failure22

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.