ChatURP

    Documentation

    AI program review#

    Upload a robot program on its own and get it read back to you: what is clear, what will cost the next person time, and what looks like a bug. No support file, no log files, no robot needed — just the program.

    It costs 10 credits per review. Re-opening a review you have already paid for is free, from any device.

    What you need#

    PolyScope 5 — two files, and a third if you have it:

    FileNeeded?Why
    .urpyesthe program itself
    .installationyesthe I/O names, TCPs and payloads the program refers to. Without it the review cannot tell whether the signals your program uses are named
    .variablesoptionalstarting values. A program saved without one is perfectly normal

    PolyScope X — one file. A .urpx already carries its application, so there is nothing else to find.

    The quickest way is to drop in a UR support file (the .zip the robot writes). Pick the program from the list and the files below fill themselves in, including the right installation — a real support file usually holds several, and the review pairs your program with the one it actually references.

    What you get#

    The report opens on its own page and can be printed or saved as a PDF.

    At the top, a short written summary — what the findings add up to and what to deal with first.

    A graded checklist, each entry with the evidence behind it:

    • are the taught positions named so a colleague can read them on the pendant
    • are your variables named after what they hold
    • is the program grouped into Folders that match the real job
    • how deep the nesting goes
    • what the threads are for, and what happens if one stops
    • do the Comments explain why, or restate the obvious
    • Script nodes doing work a normal node would do better
    • the same sequence written out repeatedly instead of once as a Subprogram
    • empty nodes, suppressed nodes and branches nothing can reach
    • the TCPs and payloads the program relies on
    • whether the I/O the program uses is named

    Measured findings beside the graded ones — blends that are missing where the robot only passes through, blends that overlap so the controller cannot fit them, threads that never pause, and I/O the program uses that the installation does not name. These are counted rather than judged, and the report says plainly where it looked and found nothing.

    What it does not do#

    • It does not check safety. Nothing here reads your safety configuration or judges it, on purpose: a safety decision needs the cell, the people in it and the risk assessment, none of which is in a program file.
    • It does not run your program, so it cannot tell you about timing, reach, or anything that depends on where the robot actually is.
    • It is not an approval. It is a second pair of eyes on the program text.

    Where the files go#

    The program is sent to our servers and to the AI provider that writes the review. The provider keeps what it is sent for up to 30 days in its abuse-monitoring logs. The finished review is stored on your account so you can re-open it; the program file itself is not kept. Full detail is in the privacy policy.