What problem does it solve? Knowledge gained during development sessions—bugfixes, architectural decisions, unexpected discoveries—is often lost between sessions, forcing agents and developers to rediscover the same context repeatedly. ## Core Features & Use Cases - Structured Observation Format: Records every finding with a searchable title, typed category (bugfix, decision, architecture, discovery, pattern, config, preference), and What/Why/Where/Learned content fields. - Topic Key Management: Uses stable topic_key identifiers so evolving topics update in place instead of creating conflicting duplicates. - Retrieval Workflow: Recovers knowledge through mem_context for recent history, mem_search for full-text FTS5 queries, and mem_get_observation for complete untruncated content. - Use Case: After fixing a race condition in a task queue, the agent immediately saves an observation documenting the mutex fix, the affected files, and the lesson learned, so future sessions can retrieve it via search. ## Quick Start Ask the agent to save a memory observation documenting the bugfix or decision just completed, including the affected file paths and any lessons learned.
