Configuration and customization

Most users can work entirely through the GUI. Use configuration or a plugin when a setting must apply repeatedly, a new metric or view is needed, or an existing action should behave differently.

Configuration locations

The user configuration directory is ~/.phy/. Its main files and directories are:

~/.phy/phy_config.py
~/.phy/plugins/
~/.phy/TemplateGUI/state.json
~/.phy/screenshots/

Each dataset may also contain .phy/state.json and cached computations under .phy/.

At startup, phy combines packaged defaults, global GUI state, and dataset-local state. Local state has the highest precedence. Runtime changes are saved again when the GUI closes. This precedence matters when a plugin changes a value that was previously saved: launch once with --clear-state, and use --clear-cache as well when the setting changes a cached computation.

Plugins

A plugin is a Python class derived from phy.IPlugin. Put plugin files in ~/.phy/plugins/ and activate their class names in ~/.phy/phy_config.py:

c = get_config()
c.TemplateGUI.plugins = [
    'MyPlugin',
]

The filename is arbitrary, but plugin class names must be unique. See the plugin introduction and plugin cookbook.

Common customization tasks

• Change keyboard shortcuts
• Change spike limits and performance settings
• Understand or replace the similarity metric
• Add cluster metrics, columns, views, colors, and actions
• Use the in-process IPython console

Resetting safely

Use the narrowest reset that addresses the problem:

# Reset view layout and saved GUI options.
phy template-gui params.py --clear-state

# Rebuild cached dataset computations.
phy template-gui params.py --clear-cache

# Do both after changing a cached controller setting in a plugin.
phy template-gui params.py --clear-state --clear-cache

Clearing state does not delete curated spike_clusters.npy or cluster TSV files. Nevertheless, back up custom files under ~/.phy/ before removing that directory manually.