VectorCode
Verified Safeby Davidyz
Overview
VectorCode is a code repository indexing tool that provides Retrieval-Augmented Generation (RAG) context to Large Language Models (LLMs) for coding workflows by vectorizing codebase files and offering query capabilities.
Installation
vectorcode-mcp-serverEnvironment Variables
- USER
- USERNAME
- HOME
- VECTORCODE_LOG_LEVEL
- VECTORCODE_NVIM_LOG_LEVEL
- ANONYMIZED_TELEMETRY
Security Notes
The server's dynamic loading of embedding functions and rerankers from configuration (e.g., `embedding_function`, `reranker`) using `getattr` carries a moderate risk. While intended to resolve trusted internal modules (chromadb.utils.embedding_functions and vectorcode.subcommands.query.reranker), if an attacker could manipulate the configuration or the Python environment's module path, it could potentially lead to arbitrary code execution. The bundled ChromaDB server binds to localhost, which limits network exposure for the database itself. User-configured remote ChromaDB connections via `db_url` and `db_settings` rely on external security (e.g., authentication) which must be properly managed by the user. No hardcoded secrets were identified in the provided source.