Troubleshooting uploads

If you're signed in but not getting results from your uploads, the most common issues and fixes are below.

A PDF won't load or the agent can't find your part

Some PDFs don't load cleanly, especially long datasheets. Take a screenshot of the specific section you need, such as the package drawing or dimension table, crop it tightly around the part data, and upload the image instead. The agent (currently) reads a well-framed image more reliably than a full document.

The datasheet doesn't include what you need

Some datasheets never draw the footprint. They name a standard package, such as SOT-23, and assume the reader knows it. If the drawing or dimensions aren't in the document, the agent can't extract them from it. Use a source that contains the package drawing or dimension table for the part you're creating.

Nothing was generated

If the system doesn't have enough information to create a particular output, it won't generate that output. This is deliberate. It is not a failure or a crash. Check that your upload contains the data needed for what you asked for, such as a dimension table for a footprint or a mechanical drawing for a 3D model, and try again with a more complete source or a tightly cropped screenshot of the relevant section.

If your source is complete and you're still not getting results, check your prompt and node setup below.

Write a clear prompt

The prompt tells the agent what to make from what you've given it. A short, specific prompt works better than a vague one.

  1. Name the output you want: a symbol, a footprint, or a 3D model
  2. Describe what you're providing, such as "this is the dimension table for a 64-pin QFP" or "this is the package drawing from the datasheet"
  3. Include details the image alone can't confirm, such as pin count or package name

"Make the footprint for this 64-pin QFP using the dimension table shown" will outperform "make this part."

Prompts can be written in your preferred language. Multi-language prompts are supported.

Check your node setup

There are two ways to create an artifact: a one-off upload from the Create Artifact page, or a recipe in the recipe editor. The setup steps differ, so use the section that matches where you're working.

One-off uploads from Create Artifact

The Create Artifact page, in the top corner of the Artifacts section, is the simplest path from upload to artifact. The recipe is formatted for you. You choose a node card, write a prompt, and upload a file.

  1. Select a node card. Library and Schematic are currently active in the EDA section. Cards for PCB, SPICE, and Wire diagram, and the MDA cards (Model, Drawing, Assembly, and Sketch), are visible but not yet active.
  2. In the Library node, click either the Symbol or Footprint button to set the output type. This selection controls what the agent renders. It will not render a footprint, even if your prompt describes one, while Symbol is selected. The Schematic node has no Symbol or Footprint selection.
  3. Write your prompt in the prompt window.
  4. Click the + symbol to upload your file.
  5. Use the preview window to frame your data. Zoom and crop until only the part data you need is in frame.
  6. To adjust the frame after you've finished, click Edit section to reopen the preview window.

If your output doesn't match your prompt, check the Symbol or Footprint selection first. The button setting wins over the prompt.

Working in the recipe editor

The recipe editor gives you full control over the workflow by connecting nodes on the canvas. For the complete palette reference, see Recipe nodes, and for how to add, connect, configure, and run nodes, see Using nodes. The checks below cover the setup issues that most often block results.

Use the right node for the output. The Library node extracts component symbols and footprints from reference files. The Schematic node extracts schematics. Use Library when you're making an individual part, and Schematic when you're working from circuit-level source material.

In the Library node, map your file to the symbol or footprint role to match what you're creating. As on the Create Artifact page, the role setting controls what the agent renders. To check it, click the node and look under Configuration in the properties panel, which is also where the prompt lives.

If you're working from a PDF, the Extract image node sends exactly what's visible in its viewer. Zoom and crop until only your part data is in frame, lock the view, then run. If the agent missed your part, the frame probably included too much of the page.

To find where a workflow breaks, test one step at a time. Select a node and click Test step in the properties panel to run the recipe up to and including that node. Test the Extract image node to confirm your framing, then the Library or Schematic node to confirm the prompt and role mapping, instead of rerunning the whole recipe and guessing.

Presets come pre-wired, so if you're unsure how to connect nodes, start from a preset such as Extract footprint from image and add your file.

Still need help?

Email support@neurocad.com


Last updated June 11, 2026

Last updated June 11, 2026