dsh-memory-search-plus

In DeepSeek Harness, users often engage in multiple conversations, but the built-in search functionality is limited to the current session, making it difficult to quickly retrieve key information from past chats. The dsh-memory-search-plus plugin addresses this problem by providing local, cross-session full-text search across the entire Harness environment, making all historical content easily accessible.

The plugin's core capability is built on an SQLite FTS5 indexing system that supports Chinese and English tokenization as well as mixed-language queries, ensuring search accuracy and efficiency. It employs an adaptive two-stage search algorithm that first rapidly recalls candidate results and then precisely verifies and highlights message positions. Users can search content across all sessions through a sleek sidebar interface similar to ChatGPT, with role-based filtering for user, assistant, or tool messages. The entire process runs fully locally without network calls, guaranteeing data privacy and security.

dsh-memory-search-plus is particularly suited for users who frequently use DeepSeek Harness for prolonged, multi-session work, such as software developers, researchers, or content creators. If you often need to review past discussions, find specific information, or organize conversation records, this plugin will become a powerful tool to enhance your efficiency, helping you quickly locate needed content amid extensive chat histories.