codebasesearch 0.1.4

package/.thornsignore

@@ -0,0 +1,419 @@
+# Code Search Default Ignore Patterns
+# Ultra-comprehensive ignore list for all languages and frameworks
+
+# === VERSION CONTROL ===
+.git/
+.svn/
+.hg/
+.bzr/
+CVS/
+
+# === NODE / JAVASCRIPT / TYPESCRIPT ===
+node_modules/
+.npm/
+.yarn/
+.pnp.*
+.pnp/
+yarn-error.log
+npm-debug.log*
+.eslintcache
+.node_repl_history
+*.tsbuildinfo
+.next/
+.nuxt/
+dist/
+out/
+build/
+.cache/
+.parcel-cache/
+.vite/
+.turbo/
+
+# === PYTHON ===
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+env/
+venv/
+ENV/
+.venv
+pip-log.txt
+pip-delete-this-directory.txt
+.pytest_cache/
+.hypothesis/
+.mypy_cache/
+.dmypy.json
+.pyre/
+.pytype/
+*.egg-info/
+dist/
+build/
+*.whl
+.tox/
+.coverage
+htmlcov/
+
+# === RUST ===
+target/
+Cargo.lock
+**/*.rs.bk
+*.pdb
+
+# === GO ===
+vendor/
+*.exe
+*.exe~
+*.dll
+*.so
+*.dylib
+*.test
+*.out
+go.work
+
+# === JAVA ===
+target/
+*.class
+*.jar
+*.war
+*.ear
+*.nar
+.gradle/
+build/
+.mvn/
+!.mvn/wrapper/maven-wrapper.jar
+
+# === C / C++ ===
+*.o
+*.a
+*.so
+*.out
+*.exe
+*.obj
+*.dll
+*.dylib
+CMakeFiles/
+CMakeCache.txt
+cmake_build_debug/
+cmake_build_release/
+build/
+*.cmake
+!CMakeLists.txt
+
+# === C# / .NET ===
+bin/
+obj/
+*.suo
+*.user
+*.userosscache
+*.sln.docstates
+[Dd]ebug/
+[Rr]elease/
+x64/
+x86/
+[Bb]uild/
+bld/
+[Bb]in/
+[Oo]bj/
+*.nupkg
+*.snupkg
+project.lock.json
+project.fragment.lock.json
+artifacts/
+
+# === RUBY ===
+*.gem
+*.rbc
+/.config
+/coverage/
+/InstalledFiles
+/pkg/
+/spec/reports/
+/spec/examples.txt
+/test/tmp/
+/test/version_tmp/
+/tmp/
+.bundle/
+vendor/bundle/
+lib/bundler/man/
+
+# === PHP ===
+/vendor/
+composer.lock
+*.phar
+
+# === SCALA / SBT ===
+target/
+lib_managed/
+src_managed/
+project/boot/
+project/plugins/project/
+.history
+.cache
+.lib/
+
+# === SWIFT / XCODE ===
+*.xcodeproj/
+*.xcworkspace/
+Pods/
+DerivedData/
+*.moved-aside
+*.pbxuser
+!default.pbxuser
+*.mode1v3
+!default.mode1v3
+*.mode2v3
+!default.mode2v3
+*.perspectivev3
+!default.perspectivev3
+
+# === FIREBASE ===
+.firebase/
+firebase-debug.log
+firestore-debug.log
+ui-debug.log
+.firebaserc
+
+# === LLM / AI / AGENTIC FRAMEWORKS ===
+.cache/
+.llamaindex/
+.chroma/
+.vectorstore/
+.embeddings/
+.langchain/
+.autogen/
+.semantic-kernel/
+.openai-cache/
+.anthropic-cache/
+embeddings/
+vector-db/
+faiss-index/
+chromadb/
+pinecone-cache/
+weaviate-data/
+
+# === CLOUD PROVIDERS ===
+.aws/
+.azure/
+.gcloud/
+.terraform/
+*.tfstate
+*.tfstate.backup
+.terraformrc
+terraform.rc
+
+# === DOCKER / CONTAINERS ===
+.dockerignore
+docker-compose.override.yml
+.docker/
+
+# === DATABASES ===
+*.db
+*.sqlite
+*.sqlite3
+*.sql
+*.bak
+*.dump
+
+# === LOGS & TEMP FILES ===
+*.log
+*.tmp
+*.temp
+*.swp
+*.swo
+*~
+.DS_Store
+Thumbs.db
+desktop.ini
+*.stackdump
+*.orig
+
+# === EDITORS & IDEs ===
+.vscode/
+.idea/
+*.iml
+.project
+.classpath
+.settings/
+*.sublime-project
+*.sublime-workspace
+.vs/
+
+# === COMPILED OUTPUT ===
+*.min.js
+*.min.css
+*.bundle.js
+*.chunk.js
+*.map
+
+# === DOCUMENTATION BUILD ===
+docs/_build/
+site/
+_site/
+.docusaurus/
+.vuepress/dist/
+
+# === TEST COVERAGE ===
+coverage/
+.nyc_output/
+lcov.info
+*.lcov
+
+# === PACKAGE MANAGERS ===
+.pnpm-store/
+.rush/
+.lerna/
+
+# === MONOREPO ===
+.nx/
+.turbo/
+
+# === METADATA / OS ===
+.Spotlight-V100
+.Trashes
+ehthumbs.db
+.fseventsd
+.TemporaryItems
+.AppleDouble
+.LSOverride
+
+# === SECURITY / SECRETS ===
+.env
+.env.local
+.env.*.local
+*.key
+*.pem
+*.p12
+*.pfx
+credentials.json
+secrets.yaml
+secrets.yml
+
+# === PROFILING / BENCHMARKS ===
+*.prof
+*.cpuprofile
+*.heapprofile
+
+# === LOCK FILES (optional - can be removed if needed) ===
+package-lock.json
+yarn.lock
+pnpm-lock.yaml
+Gemfile.lock
+poetry.lock
+Pipfile.lock
+
+# === CONFIGURATION & METADATA FILES (LOW-VALUE FOR ANALYSIS) ===
+*.config.js
+*.config.ts
+webpack.config.js
+rollup.config.js
+vite.config.js
+tsconfig.json
+jsconfig.json
+babel.config.*
+.babelrc
+.eslintrc.*
+.prettierrc.*
+.stylelintrc.*
+.editorconfig
+.nvmrc
+.env.example
+.env.template
+.env.sample
+*.local
+*.development
+*.production
+
+# === DOCUMENTATION & REFERENCE FILES ===
+*.md
+*.txt
+*.rst
+*.adoc
+docs/
+documentation/
+wiki/
+CHANGELOG*
+HISTORY*
+NEWS*
+UPGRADING*
+FAQ*
+CONTRIBUTING*
+SECURITY*
+LICENSE*
+LICENCE*
+COPYRIGHT*
+NOTICE*
+AUTHORS*
+THIRDPARTY*
+*.orig
+*.rej
+
+# === TEST & COVERAGE FILES ===
+*.test.*
+*.spec.*
+test/
+tests/
+__tests__/
+__mocks__/
+fixtures/
+spec/
+cypress/
+playwright/
+test-results/
+coverage/
+.nyc_output/
+lcov.info
+*.lcov
+.coverage
+pytest.ini
+tox.ini
+jest.config.*
+vitest.config.*
+
+# === GENERATED & BUILD ARTIFACTS ===
+dist/
+build/
+out/
+target/
+.next/
+.nuxt/
+.gatsby/
+.docusaurus/
+.vuepress/dist/
+site/
+public/
+static/
+.assets/
+.cache/
+.parcel-cache/
+.vite/
+.turbo/
+.tmp/
+temp/
+
+# === DATABASE & DATA FILES ===
+*.db
+*.sqlite
+*.sqlite3
+*.sql
+*.bak
+*.dump
+*.backup
+*.data/
+storage/
+logs/
+*.log
+
+# === IDE & EDITOR SPECIFIC ===
+.vscode/
+.idea/
+*.iml
+*.swp
+*.swo
+*~
+.DS_Store
+Thumbs.db
+desktop.ini
+*.sublime-*
+.vs/

package/README.md

@@ -0,0 +1,129 @@
+# code-search
+
+Ultra-simple semantic code search with Jina embeddings and LanceDB. Supports both CLI and MCP protocol interfaces.
+
+## Quick Start
+
+### CLI
+```bash
+bunx code-search "your search query"
+```
+
+### MCP (for Claude Code & IDE plugins)
+```bash
+bunx code-search --mcp
+```
+
+Example:
+```bash
+claude mcp add -s user code-search -- bunx code-search
+```
+
+## Features
+
+- **Semantic search** across entire repositories using Jina embeddings (512-dim vectors)
+- **Embedded vector database** (LanceDB) - no external servers or setup required
+- **Auto-indexing** - automatically scans and indexes repository before each search
+- **Comprehensive ignore patterns** - respects .gitignore and ignores build artifacts, node_modules, etc. across all languages
+- **Single-shot execution** - no persistent processes, no background daemons
+- **MCP protocol support** - integrates with Claude Code and other MCP-compatible tools
+- **Auto-gitignore** - automatically adds `.code-search/` to .gitignore on first run
+- **Auto-recover from corruption** - automatically detects and clears corrupted model cache on Protobuf errors
+- **Performance optimized** - 5MB file size limits, smart chunking, batch embedding generation
+
+## Usage
+
+### Search from CLI
+
+```bash
+bunx code-search "authentication middleware"
+bunx code-search "database connection pool"
+bunx code-search "error handling"
+```
+
+### Search from custom repository
+
+```bash
+bunx code-search --repo /path/to/repo "query"
+```
+
+**Default Search Directory**: When no path is specified, searches the **current working directory** (project root), not the Claude Code plugins directory. In Claude Code, this defaults to your project context.
+
+### MCP Tool (in Claude Code)
+
+The `search` tool accepts:
+- `query` (required): Natural language search string
+- `repository_path` (optional): Path to repository (defaults to current directory)
+
+Example:
+```
+search query="middleware validation" repository_path="/path/to/repo"
+```
+
+## How It Works
+
+1. **Scans** the repository for code files (25+ language types supported)
+2. **Respects** .gitignore and comprehensive ignore patterns
+3. **Chunks** large files into manageable segments
+4. **Generates embeddings** using Jina embeddings v2 small (512 dimensions)
+5. **Stores** vectors in embedded LanceDB database
+6. **Searches** using semantic similarity
+7. **Returns** ranked results with line numbers and code snippets
+8. **Auto-adds** `.code-search/` to .gitignore
+
+## Supported Languages
+
+JavaScript, TypeScript, Python, Go, Rust, Java, C/C++, C#, Ruby, PHP, Scala, Swift, Shell, SQL, R, Lua, Perl, Groovy, XML, JSON, YAML, TOML, HTML, CSS, SCSS, Vue, and more.
+
+## Storage
+
+Search index is stored in `.code-search/lancedb/` (automatically added to .gitignore).
+
+First run downloads the Jina model (~120MB) to `~/.cache/huggingface`.
+
+## Performance
+
+- **First run**: ~30-60s (downloads model + indexes repository)
+- **Subsequent runs**: Sub-second search queries (index already exists)
+- **Large repos** (10k+ files): May take 1-2 minutes for full indexing
+
+### Technical Optimizations
+
+- **5MB file size limit**: Files larger than 5MB are skipped to prevent memory issues
+- **Smart chunking**: Files >1000 lines auto-split into overlapping chunks (200-line overlap) for better semantic context
+- **Batch embedding**: Chunks processed in batches of 32 for efficient API usage
+- **Binary detection**: 47 binary file extensions ignored (.zip, .exe, .jpg, .mp4, etc.)
+- **Auto-recovery**: Detects Protobuf parsing errors in cached models and auto-clears corrupted cache
+- **5-minute timeout**: Model loading has timeout to prevent indefinite hangs
+
+## Installation Details
+
+The package includes:
+
+- **bin/code-search.js** - CLI entry point for direct use
+- **mcp.js** - MCP server for integration with Claude Code
+- **src/** - Core modules (embeddings, scanning, vector store, search)
+- **.thornsignore** - Comprehensive ignore patterns (all languages/frameworks)
+
+## Ignored Files & Directories
+
+By default, the tool ignores:
+- All build artifacts (dist/, build/, target/, node_modules/, etc.)
+- Version control (.git/, .svn/, .hg/, etc.)
+- IDE files (.vscode/, .idea/, etc.)
+- Lock files (package-lock.json, yarn.lock, etc.)
+- Dependencies and caches
+- Test files and coverage reports
+- Secrets and credentials
+
+Configure custom ignores via `.codesearchignore` file.
+
+## Privacy
+
+All processing happens locally. No data is sent to external servers. The Jina model is downloaded once and cached locally.
+
+## License
+
+MIT
+# Triggered npm publishing
+

package/bin/code-search.js

@@ -0,0 +1,27 @@
+#!/usr/bin/env node
+
+// MUST patch sharp before any other imports
+import fs from 'fs';
+import path from 'path';
+import { fileURLToPath } from 'url';
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const distPath = path.join(__dirname, '..', 'node_modules', '@huggingface', 'transformers', 'dist', 'transformers.node.mjs');
+
+if (fs.existsSync(distPath)) {
+  let content = fs.readFileSync(distPath, 'utf-8');
+  if (!content.includes('SHARP_REMOVED_FOR_WINDOWS_COMPATIBILITY')) {
+    content = content.replace(/import \* as __WEBPACK_EXTERNAL_MODULE_sharp__ from "sharp";\n/, '// SHARP_REMOVED_FOR_WINDOWS_COMPATIBILITY\n');
+    content = content.replace(/module\.exports = __WEBPACK_EXTERNAL_MODULE_sharp__;/g, 'module.exports = {};');
+    content = content.replace(/} else \{\s*throw new Error\('Unable to load image processing library\.'\);\s*\}/, '} else {\n    loadImageFunction = async () => { throw new Error(\'Image processing unavailable\'); };\n}');
+
+
+    try { fs.writeFileSync(distPath, content); } catch (e) {}
+  }
+}
+
+import('../src/cli.js').then(m => m.run(process.argv.slice(2)))
+  .catch(err => {
+    console.error('Error:', err.message);
+    process.exit(1);
+  });