How does ASTra integrate into a developer's workflow?

ASTra functions as an MCP (Multi-Code Protocol) server, allowing seamless integration with compatible editors and IDEs (like Claude Desktop/Cursor). It also persists index data for fast warm restarts, enhancing developer efficiency.

What is ASTra and its primary function?

ASTra (Abstract Syntax Tree retrieval augmentation) is a local-only semantic code search server. It parses source code to build an AST call graph and uses A* traversal biased by semantic similarity to find relevant code snippets.

Which programming languages does ASTra support?

ASTra offers multi-language support, currently providing robust semantic code search capabilities for Rust, Python, and JavaScript/TypeScript projects.

How does ASTra perform semantic search without external APIs?

ASTra performs all search operations locally. It uses tree-sitter for source parsing and `fastembed` via ONNX Runtime to generate embeddings, ensuring that queries and data processing remain on your machine without external API calls.

ASTra

Name: ASTra
Author: modpotatodotdev

bymodpotatodotdev

0•

개발자 도구

데이터 과학 및 ML

Performs local semantic code search by parsing source code, building an AST call graph, and traversing it with A* biased by semantic similarity.

ASTra is a local-only semantic code search server designed to run as an MCP process over stdio, offering advanced code understanding for developers. It leverages tree-sitter to parse source files, constructs an Abstract Syntax Tree (AST) call graph using `petgraph`, and executes A* graph traversal, prioritizing paths based on semantic similarity derived from embeddings. Supporting Rust, Python, and JavaScript/TypeScript, ASTra enhances code navigation and retrieval augmentation without relying on external APIs for search queries after an initial indexing, which can be configured for fully local operation or with a lightweight cloud model. It provides tools for semantic RAG, semantic path search, and structured path search, making it ideal for integrating into LLM-powered development environments like Claude Desktop or Cursor.

주요 기능

01Multi-language support for Rust, Python, and JavaScript/TypeScript

02Persists index data for fast warm restarts and efficient subsequent runs

03MCP server integration for seamless use with compatible editors/IDEs

040 GitHub stars

05AST-driven RAG with vector-biased A* graph traversal

06Local-only semantic code search without external APIs for queries

사용 사례

01Traversing and understanding execution paths through complex codebases

02Performing semantic code search within local development workspaces

03Enhancing Retrieval Augmented Generation (RAG) for LLMs in code-related tasks