Library

The library is simse's persistent memory system. It stores knowledge across sessions — project context, code patterns, decisions, notes — and retrieves what is relevant when you need it. The more you use simse, the more useful the library becomes.

What the library does

Every time you work in a session, simse learns from the conversation. When a session ends, its context flows into the library and is organized automatically. The next time you start a session, simse pulls in what is relevant based on what you are working on.

You can also ask simse to store specific content — notes, decisions, reference material — so simse has access to it alongside the context it gathers from your sessions.

Search

The library supports search by meaning, not just keywords. Ask simse "how did I handle authentication" and it retrieves relevant entries even if those exact words do not appear.

simse searches the library automatically at each turn to surface relevant context. You can also ask it to search explicitly:

Search the library for notes on API rate limiting.

simse uses the library_search tool internally, which accepts a query string and an optional maximum number of results.

Organization

The library organizes entries by topic automatically. When content enters the library, simse classifies it into the topic hierarchy. You can browse entries by topic using library_catalog or ask simse to show you what is stored under a particular topic.

Automatic cataloging

You do not need to manually organize entries. When new content enters the library — whether from a session, an import, or a direct write — simse classifies it, assigns it to appropriate stacks and shelf categories, and connects it to related entries.

Different types of content are handled with different cataloging strategies:

• Code files are indexed by function names, class definitions, and module structure. simse extracts the key symbols so you can later search by function or class name rather than remembering which file something was in.
• Markdown and documentation are indexed by heading structure. Each section becomes a separately retrievable entry, so a search for "deployment steps" finds the right section of a long document without returning the entire file.
• Conversation turns are indexed by topic and intent. A session where you discussed error handling strategy becomes findable by that topic, even if the words "error handling" only appeared once in passing.

This means retrieval stays precise as the library grows — you get the specific section, function, or decision you need, not a wall of loosely related content.

Adaptive retrieval

The library learns what is relevant to you over time. It tracks which entries you access, which searches produce useful results, and which context actually helps during sessions. This feedback loop means retrieval gets more accurate the longer you use simse.

Library tools

The library is managed through built-in tools that simse uses during sessions. You interact with these by asking simse in natural language.

Tool	Parameters	Description
library_search	query (string), maxResults (number)	Search the library by query
library_shelve	text (string), topic (string)	Store an entry in the library
library_withdraw	id (string)	Remove an entry by ID
library_catalog	topic (string, optional)	List entries, optionally filtered by topic
library_compact	topic (string)	Consolidate entries within a topic

Relationship to sessions

Sessions and the library work together in a continuous cycle:

1. A session starts and simse retrieves relevant context from the library.
2. You work, generating new context — code changes, decisions, patterns.
3. The session ends and its context flows into the library.
4. The library organizes and catalogs the new entries automatically.
5. The next session benefits from everything that came before.

This cycle is what makes simse more useful the more you use it. Context is never lost — it is organized, indexed, and ready when you need it.