recall

Search memory stores via SQLite FTS5 index and return top 5 cited results.

Updated Dec 17, 2025

Search Tags:#indexing #retrieval #evidence #confidence #semantic-search #memory-search #memory-stores

Ready to run

npx skills add https://github.com/eprouveze/HealthPulse --skill recall-eprouveze

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: recall
Source: https://github.com/eprouveze/HealthPulse/tree/main/.claude/skills/recall
Command: npx skills add https://github.com/eprouveze/HealthPulse --skill recall-eprouveze

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill enables fast retrieval of information across all memory stores, surfacing relevant decisions, feedback, solutions, and context with evidence paths and confidence scores.

Core Features & Use Cases

Cross-store search: Query multiple memory stores and retrieve top matches with citations and scoring.
Structured results: Each result includes evidence path, store type, project, and a breakdown of the scoring factors.
Fallback & stability: If the primary index is unavailable, the tool automatically falls back to a grep-based search across configured stores for resilience.

Quick Start

Run a query using the memory search interface to retrieve the most relevant past items and their evidence.

Frequently Asked Questions about recall

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I perform a semantic memory search across multiple stores?▼

Semantic memory search queries multiple stores using an SQLite FTS5 index with Gemini embeddings, returning the top 5 results with source citations and confidence scores.

What is the best way to retrieve past decisions and context from memory stores?▼

Retrieve past decisions and context using cross-store search, which surfaces relevant items with evidence paths, store type, project, and a breakdown of scoring factors.

How does the search handle abbreviations and evidence scoring?▼

The search preprocesses queries by expanding common abbreviations before retrieval, then presents structured results with evidence paths and a breakdown of confidence scoring factors.

What happens if the primary SQLite FTS5 index is unavailable?▼

If the primary index is unavailable, the search automatically falls back to a parallel grep-based search across designated stores, surfacing results with weighted scoring and deduplication.

Do I need a specific database setup to retrieve evidence paths with confidence scores?▼

Retrieval relies on an Anamnesis SQLite index combining FTS5 and Gemini embeddings to generate evidence paths and confidence scores, requiring no additional dependencies for standard operation.