remembering-conversations

Search past conversations using natural language queries to retrieve relevant context from archived chats. The skill applies semantic similarity and exact-text matching over your conversation history, enabling you to quickly locate discussions about specific topics like architecture decisions or code patterns.

Yes, retrieve conversation history across projects for debugging, reviewing architectural decisions, and reusing patterns. The skill supports project-wide archiving with configurable concurrency, allowing you to access indexed conversations relevant to your current work.

Conversations are automatically indexed using 384-dim embeddings via Xenova all-MiniLM-L6-v2 and stored in sqlite-vec for vector search. The skill tracks per-exchange indexing and generates automated summaries, enabling offline local indexing and fast retrieval without external dependencies.

The skill requires npm for installation. It handles semantic indexing, vector storage, and summary generation internally, so you don't need to configure external APIs or language models beyond the built-in indexing pipeline.

Semantic similarity matches the meaning of your search query against archived conversations, not just exact keywords. Combined with exact-text matching, this dual approach ensures you find contextually relevant discussions even when phrased differently from your original conversation.

Search retrieves past Claude Code conversations using both semantic and exact-text matching, pulling relevant context with automated summaries. Per-exchange last-indexed tracking ensures efficient updates, letting you instantly access weeks-old discussions without manual organization.