A 100% offline semantic code search tool that uses embeddings to find code by meaning rather than just keyword matching.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install semantic-grep
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install semantic-grep using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is semgrepll - Local Semantic Code Search?
semgrepll is a powerful local semantic grep utility designed for developers who need to navigate large codebases by intent and logic. By leveraging vector embeddings through backends like llama.cpp, ONNX, or Ollama, it allows for high-performance code discovery without ever sending your private source code to an external API.
As one of the essential Openclaw Skills for AI-native development, it bridges the gap between traditional grep and modern LLM-powered context retrieval. It automatically handles the heavy lifting of project indexing and vector storage, ensuring that finding complex logic—like authentication flows or payment processing—is as simple as describing what the code does in plain English.
semgrepll - Local Semantic Code Search Use Cases
- Finding specific functional logic across massive mono-repos without knowing exact variable names.
- Indexing local projects to provide high-quality context for AI coding agents.
- Performing semantic audits of codebases for security or architectural patterns.
- Searching for code implementations when only the high-level concept is known.
How semgrepll - Local Semantic Code Search Works
- The user triggers an index command on a specific directory path.
- semgrepll parses the source files and generates vector embeddings using the configured local backend (llama.cpp, ONNX, or Ollama).
- These embeddings are stored in a local database (SQLite or LanceDB) for persistence.
- When a search query is issued, the tool converts the query into a vector and performs a similarity search against the indexed codebase.
- The most relevant code snippets are returned based on their semantic meaning.
semgrepll - Local Semantic Code Search Setup
Install the core package via pip. To ensure full compatibility with various local backends, you can specify optional dependencies.
# Basic installation
pip install semgrepll
# Installation with ONNX support for faster CPU inference
pip install semgrepll[onnx]
Ensure you have Python 3.10+ and a local provider like Ollama running or a GGUF model path defined in your environment variables.
semgrepll - Local Semantic Code Search Data Schema & Taxonomy
semgrepll organizes its data to balance speed and local privacy. It utilizes a pluggable storage backend for its vector index.
| Component | Type | Description |
|---|---|---|
| Vector Store | SQLite / LanceDB | Stores the document embeddings and file metadata. |
| Embeddings | mxbai-embed-large-v1 | Default model used to represent code logic numerically. |
| Cache | Local Filesystem | Caches embeddings to allow for rapid re-indexing of modified files. |
semgrepll - Local Semantic Code Search Advanced Features
- Multi-backend support allowing seamless switching between llama.cpp, ONNX, and Ollama.
- Auto-detection of the fastest available compute backend for your specific hardware.
- Support for LanceDB to handle high-performance vector operations on larger projects.
- Flexible environment configuration for integrating with custom GGUF models via LLM_MODEL_PATH.
- Native integration capability with AI agents via Openclaw Skills protocols.
SKILL.md
Loading
Related Openclaw Skills
ClawHub Skill Manager
A comprehensive command-line management tool for discovering, installing, and updating skills from public registries.
Virtual Board of Advisors
A strategic advisory engine that emulates the decision frameworks of 10 world-class business and industry leaders to guide complex professional decisions.
Patrick Emotional Companion
A whimsical emotional support agent inspired by Patrick Star that offers companionship, humor, and integrated weather reporting.
Autoresearch Skill
Autoresearch is an autonomous, goal-directed skill designed to systematically optimize system metrics through iterative make-change-verify-keep/revert cycles.
Parma
A powerful integration skill for Parma that allows AI agents to manage restaurant orders, CRM data, and project documentation through the Membrane platform.
bkash-nagad-tracker
A conversational financial tracking tool specifically designed for the Bangladesh market to monitor mobile financial services and cash spending.
Featured*
