Mica Space: A DSH Plugin that Separates Follow-up Questions from Answers in Long Conversations

Mica Space is a DSH client plugin maintained by ToiMee; package version 0.99.7, MIT license, 1 star on GitHub. It transforms long conversations into a scalable conversation map: a vertical main thread where follow-up questions branch directly from corresponding answers into sub-threads. The main thread retains full context, while local memory supports main thread summaries, branch summaries, parent-child retrieval, and callbacks upon returning to the main thread. The plugin is zero-configuration, requiring no installation of vector databases, embedding models, Python, or embedding APIs. It offers local OCR (Tesseract, offline) for images and scanned PDFs, a trash can, undo, search, full-database import/export, and Git backup. Data defaults to the `.chat-space/` directory. Lightweight hash vectors and optional local co-occurrence semantic patterns do not access the network. During normal chat, contexts and attachments are sent to the DSH model provider according to user configuration. The plugin runs with the current DSH process permissions. It is a preview version; backups must be preserved before upgrading. The plugin contains no telemetry, and attachment files are not included in exported JSON or Git backups. The Agent mode is not included in the 0.99.7 demo but is planned for exploration in v1.0. Directory: skillhub.cn; Source code: github.com/ToiMee/mica_space.