v0.1.0 — Initial PyPI Release

concept-file v0.1.0

Portable semantic search without a vector database — just files, find, and sort.

Highlights

• Published on PyPI — install with pip install concept-file (#45)
• 5 CLI tools — automatically added to PATH via pip install
  • concept-grep — semantic grep (search source code by meaning)
  • concept-embed — generate .concept files from text
  • concept-show — display .concept file contents
  • concept-sim — cosine similarity calculation
  • concept-plot — UMAP 2D/3D scatter plot visualization

Key Features

• Hybrid search — combines semantic similarity + keyword matching (#37)
• Tree-sitter summarization — automatically extracts structural summaries from 25 languages for higher-quality embeddings (#18)
• --index mode — bulk-index source code with concept-grep --index -r src/. Skips unchanged files using SHA-256 hash (#25)
• --summary — show file structure summary in search results (#43)
• --include / --exclude — filter files by glob pattern (#38)
• -g (graph) — visualize similarity scores as bar graphs (#14)
• Local LLM support — free, API-key-free embedding via LM Studio

Quick Start

pip install concept-file
export CONCEPT_API_BASE="http://localhost:1234/v1"
export CONCEPT_EMBED_MODEL="text-embedding-qwen3-embedding-0.6b"

concept-grep --index -r src/
concept-grep -r "user authentication" src/

Examples

• Java project similarity analysis
• Multilingual Wikipedia concept visualization (English & Japanese)
• Vue.js GitHub Issue semantic search & auto-clustering
• PostgreSQL 18.3 source code semantic search (2271 files)