Compear

Compear is a browser-based CSV/XLSX comparison tool that finds semantically similar rows using in-browser embeddings. All processing runs client-side — no backend required.

Key features

• Upload CSV or XLSX (multi-sheet supported).
• Select one or more columns to use as the comparison text (embeddings are generated from the concatenated column values).
• Choose which columns are shown in the results table (display columns can be different from comparison columns).
• Pairwise semantic similarity (N²) with cosine scores; results sorted by similarity and paginated (default 50 rows).
• Switch between several preconfigured embedding models; model loading and inference happen in-browser via transformers-style pipelines.

Quick start (development)

• Install & run dev server:

npm install
npm run dev

• Build for production:

npm run build
npm run preview

Other useful scripts: npm run lint, npm run format, npm run type-check.

How it works (short)

• A Pinia store manages application state, model loading, uploaded files, selected columns, embeddings and similarity results.
• On store creation the selected model is initialized; model readiness gate prevents comparisons until the model is ready.
• Text for each CSV/XLSX row is created by concatenating the selected comparison columns; embeddings are generated and normalized; pairwise cosine similarities are computed and returned as result pairs.

Architecture & important files

• Model + embedding utilities: src/utils/embeddings.ts
• Pairwise similarity logic: src/utils/similarity.ts
• CSV parser: src/utils/csvParser.ts
• XLSX parser: src/utils/xlsxParser.ts
• Central state: src/stores/comparison.ts
• File upload / column selection UI: src/components/FileUploader.vue
• Results UI: src/components/ComparisonResults.vue

Available embedding models The project ships with several model names configured (see AVAILABLE_MODELS in src/utils/embeddings.ts) including:

• Xenova/bge-small-en-v1.5
• onnx-community/Qwen3-Embedding-0.6B-ONNX
• Xenova/all-MiniLM-L6-v2
• Xenova/all-MiniLM-L12-v2
• nomic-ai/nomic-embed-text-v1.5

Model loading occurs in-browser and may use WebGPU when available (the embedding util attempts to detect and enable WebGPU; otherwise it falls back to a WASM/CPU backend). Large models can take time and memory — choose a smaller model for faster, lighter usage.

Notes & developer tips

• Model initialization runs automatically when the store is created; switching models clears cached embeddings and results.
• Embeddings generation reports progress (store exposes comparisonProgress and comparisonPhase).
• The pairwise similarity step is O(N²) — for large files this can be slow or memory intensive; consider limiting input size or sampling rows.

Contributing

• Add models by updating AVAILABLE_MODELS in src/utils/embeddings.ts.
• Add new file formats by adding parsers under src/utils/ and updating the file load flow in the store.

License See the project root for license information.