What problem does it solve?
Provides a structured workflow to locate, evaluate, and retrieve relevant past session observations so users can reuse prior debugging steps, configuration changes, and implementation patterns without re-searching or wasting tokens.
Core Features & Use Cases
- Semantic search (search): Lightweight index queries to surface candidate observations with relevance scores.
- Timeline inspection (timeline): View surrounding chronological context when the order of changes matters.
- Batched retrieval (get_observations): Fetch full, detailed observations in bulk to minimize token cost and latency.
- Privacy and save controls: Tag sensitive content with <private> and manually save important memories via save_memory.
- Use Cases: Reapply a previous bug fix, reproduce a prior PR formatting, or inspect the sequence of changes for a configuration rollout.
Quick Start
Search the memory for a specific error or feature, pick the top one to three high-relevance results, and retrieve their full observations in a single batched request for immediate reuse.