spaCy dashboards without the Streamlit boilerplate

What it does

spacy-streamlit wraps spaCy’s displacy visualizers and other NLP outputs into ready-made Streamlit components. You get dropdown model selectors, dependency parse trees, named entity highlights, text classification scores, token attribute tables, and word-vector similarity comparisons—either as a complete bundled UI or à la carte.

The interesting bit

The library is essentially a thin but thoughtful adapter layer: it handles the tedious Streamlit state management (model loading, sidebar config, color theming) so you can call spacy_streamlit.visualize() and have a working demo in four lines of Python. The individual component functions like visualize_ner() or visualize_parser() let you embed just the visualizations you need into larger apps without inheriting the full kitchen sink.

Key highlights

• One-shot full visualizer via visualize() with model dropdown, text input, and configurable components
• Modular functions: visualize_parser, visualize_ner, visualize_textcat, visualize_similarity, visualize_tokens
• Supports multiple spaCy models with automatic sidebar selection and metadata display
• Experimental color theming via hex code injection
• Includes runnable examples for out-of-the-box use and custom integration patterns

Caveats

• The color theming is explicitly marked “experimental” and implemented as a UI “hack”
• Requires spaCy models to be downloaded separately; no model bundling

Verdict

Worth it if you demo spaCy models frequently or need to prototype NLP interfaces fast. Skip if you’re building production analytics dashboards—this is scaffolding, not architecture.