Remembering Conversations

You can search past conversations for context using a local CLI tool that indexes historical chats with embeddings and a SQLite database. It supports semantic search based on query meaning and text search for exact phrases like error codes.

Yes, you can locate exact phrases in your chat history using text search. This allows you to find specific error codes, variable names, or other precise strings from previous conversations stored in the local database.

Semantic search works alongside date filtering to narrow down results to specific time ranges. This combination lets you retrieve context and patterns from conversations that occurred during a defined historical period.

Local conversation indexing operates entirely on your machine using SQLite and local embeddings, ensuring privacy. It provides detailed source attribution for synthesized findings, helping you trace retrieved context back to the original chats.

The best way to retrieve patterns without manual searching is using semantic search across indexed history. This approach finds conversations based on the meaning of your query, preventing you from having to ask your partner or scroll through logs.

Yes, the tool requires SQLite dependencies including better-sqlite3 and sqlite-vec, along with @xenova/transformers for embeddings and the @anthropic-ai/claude-agent-sdk. These dependencies power the local indexing and retrieval database.