cortex-recall

# Recall — Retrieve from Persistent Memory ## Keywords recall, remember, search, find, what did we, do you remember, what was, have we seen, look up, retrieve, past decision, previous fix, history, what do we know, search memory, find memory, related memories ## Overview Retrieve relevant memories using Cortex's 6-signal WRRF (Weighted Reciprocal Rank Fusion) retrieval engine. The system automatically classifies your query intent and adjusts retrieval weights — semantic queries emphasize vector similarity, temporal queries emphasize recency, causal queries traverse the knowledge graph. **Use this skill when:** You need context about past work, decisions, patterns, or fixes. Also use proactively when starting work on a topic that likely has stored context. ## Workflow ### Step 1: Formulate the Query Write a natural language query. The intent classifier handles routing: - **Semantic**: "How does the authentication system work?" - **Temporal**: "What did we work on last week?" - **Causal**: "What caused the deployment failure?" - **Entity**: "Everything about PostgreSQL in this project" - **Multi-hop**: "How does the memory gate relate to consolidation?" ### Step 2: Basic Recall ``` cortex:recall({ "query": "<natural language question or topic>", "limit": 10 }) ``` **Optional filters:** - `"domain"`: Filter to specific project domain - `"tags"`: Filter by tags (e.g. `["bug-fix", "authentication"]`) - `"min_heat"`: Only hot/active memories (0.0-1.0) - `"time_range"...