nexo_cognitive_retrieve
Retrieve relevant memories from STM and LTM using natural language queries, then trigger rehearsal to reinforce retention.
Instructions
RAG query over cognitive memory (STM+LTM). Triggers rehearsal on retrieved results.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | Natural language query to search for | |
| top_k | No | Maximum number of results to return (default 10) | |
| min_score | No | Minimum cosine similarity score (default 0.5) | |
| stores | No | Which store to search — "both", "stm", or "ltm" (default "both") | both |
| source_type | No | Filter by source type e.g. "change", "learning", "diary" (default: all) | |
| domain | No | Filter by domain e.g. "project-a", "shopify" (default: all) | |
| include_archived | No | If True, also search archived memories (default False) | |
| use_hyde | No | If True/False, force HyDE on/off. If omitted, NEXO auto-enables it for conceptual queries. | |
| spreading_depth | No | If >0, boost co-activated neighbors directly. If omitted, NEXO may auto-enable shallow spreading for multi-hop queries. | |
| hybrid_alpha | No | Weight for vector vs BM25 fusion (default 0.6) | |
| decompose | No | If True, decompose complex queries into sub-queries (default True) | |
| exclude_dreams | No | If True, exclude dream insights from retrieval by default | |
| exclude_dormant | No | If True, keep dormant LTM out of results unless explicitly requested |