Skip to main content
Use waveorder-tutorial to run a first WaveOrder reconstruction on your own microscopy data, or on a small sample dataset from Zenodo. The Skill works one stage at a time and asks before each step, so expect a conversation rather than a single answer.
This Skill calls no Biohub tools, so none of your microscopy data reaches Biohub. Your AI client runs the wo command, napari, and the Skill’s scripts on your own computer. A first reconstruction usually takes a couple of hours of tuning.

​
Before you begin

  • Connect Biohub MCP to a client that can run commands on your computer, such as Claude Code, Codex, or Cursor; see Connect your client.
  • Install uv, which the Skill uses to install WaveOrder and napari and to run its scripts.
  • Use a computer with a display, because the Skill shows every result in a napari window.
  • Bring a Z stack of a thin, weakly scattering sample, or let the Skill download its sample dataset.

​
Run the Skill

Try WaveOrder on its sample data

Reconstruct your own data

If your client answers without the Skill, ask it to read skill://waveorder-tutorial/SKILL.md and follow it, or see The assistant ignores a Skill.

​
What the Skill guides

  1. It explains what WaveOrder does, checks that it suits your sample, and installs WaveOrder and napari with uv once you agree.
  2. It asks for the dimensions, contrast type, voxel size, and imaging conditions one at a time, starting from the file’s metadata, and stops for a single 2D image, because WaveOrder needs a Z stack.
  3. It converts TIFF, NumPy, or plain zarr data to OME-Zarr, then crops one position and one time point (and one cell when it is much smaller than the field of view) into roi.zarr for fast iteration.
  4. It drafts 2D or 3D phase or fluorescence configs, including both signs of invert_phase_contrast for label-free data.
  5. It tunes the forward model on roi.zarr with parameter sweeps shown side by side in napari, and raises regularization last.
  6. It tests the tuned config on other positions, tries automatic tuning for parameters that drift, and estimates what the full dataset needs.

​
Expected result

The result includes:
  • A working folder with the converted data, draft configs, sweeps, and the final config.
  • For each config parameter, where its value came from: metadata, you, a manual sweep, or automatic tuning.
  • A wall-time, memory, and storage estimate for reconstructing the full dataset.
The Skill shows every result in napari and asks you what you see. It does not judge reconstructions from static images.

​
What runs on your computer

  • If the wo command is missing, the Skill installs waveorder[all], which includes napari, with uv pip install.
  • For the sample data, it downloads a brightfield defocus dataset of about 24 MB from Zenodo into ~/.waveorder_tutorial_data.
  • It saves a copy of each bundled script it needs, such as the OME-Zarr converter and the parameter sweep, into your working folder and runs it there.

​
Skill files