Utilities and Troubleshooting

Utilities and Troubleshooting

Common Options

These flags are available across most crisp and crispt commands.

Flag	Description
--inp <dir>	Input Directory. Folder containing your corpus.json. Defaults to crisp_input in many cases, but acceptable to always specify.
--out <dir>	Output Directory. Where to save results. If omitted, results are often just printed to terminal.
--clear	Clear Cache. Forces a reload of data, clearing any cached embeddings or models. Highly recommended when switching datasets.
--verbose / -v	Verbose Mode. Prints detailed logs and debugging info.

Troubleshooting

"No input data provided"

• Cause: You ran a command without specifying --source (for import) or --inp (for analysis).
• Fix: Check your paths. Ensure you imported data first.

"ML dependencies not available"

• Cause: You tried to run --kmeans, --regression, or similar commands without the ML extras.
• Fix: Install them: pip install crisp-t[ml].

Visualization Errors

• Cause: Running crispviz --wordcloud without running crisp --topics first.
• Fix: Ensure prerequisite analysis steps are done. crispviz only visualizes existing results.

Weird or Stale Results

• Cause: Caching.
• Fix: Run your command with --clear.

PDF Import Issues

• Cause: Corrupt PDF or encrypted file.
• Tip: Convert problematic PDFs to .txt manually and place them in the source folder.

Best Practices

1. Project Organization: Keep your raw data (raw_source) separate from your processed corpus (crisp_input) and your analysis results (crisp_output).
2. Incremental Analysis: Run one heavy analysis at a time (e.g., just --topics) and save it, rather than running --nlp --ml all at once on huge datasets.
3. Check IDs: Use crisp --linkage id whenever possible for precise matching between text and CSV data.