Concepts

Concepts are the generalized knowledge that emerges from consolidation. They're the persistent, meaningful output of Remind's memory system — what episodes become after the LLM "sleeps on" them.

Structure

Each concept has:

Field	Description
id	Unique identifier
title	Short descriptive title
summary	Natural language description of the generalized knowledge
concept_type	Type of concept: pattern (generalizations), fact_cluster (grouped facts), or legacy (pre-existing)
confidence	How certain (0.0–1.0), based on supporting evidence
instance_count	How many episodes support this concept
relations	Typed edges to other concepts
conditions	When/where this concept applies
exceptions	Known cases where it doesn't hold
source_episodes	Episode IDs this was derived from
specifics	For fact_clusters: list of verbatim facts
evidence	Key quotes from source episodes
conflicts	For fact_clusters: detected conflicting facts
entity_ids	Entities linked to this concept (union of source episodes' entities)
embedding	Dense vector for similarity retrieval
tags	Searchable tags
topic	Knowledge area this concept belongs to (e.g., "architecture", "product")
decay_factor	Retrieval priority weight (affected by memory decay)

How concepts differ from episodes

	Episodes	Concepts
Nature	Specific, raw	Generalized, processed
Created by	remember()	Consolidation (LLM)
Lifespan	Temporary (consumed)	Persistent
Structure	Plain text + type	Text + confidence + relations + conditions
Retrieval	Not directly queried	Primary retrieval target

Concept Types

Remind distinguishes between two types of knowledge:

Pattern Concepts

Generalizations extracted from observations, decisions, and outcomes. These capture abstract understanding:

Pattern: "The team favors statically typed languages for backend services"

• Derived from multiple language preference observations
• Generalizes across specific instances

Fact Clusters

Groups of related concrete facts, preserved verbatim:

Fact Cluster: "Redis Configuration"

• Redis TTL: 300s
• Redis connection pool: 10
• Redis max memory: 2GB

Fact clusters never generalize away specific values. When facts conflict (e.g., different TTL values from different times), both are kept with conflict markers.

Legacy Concepts

Concepts created before dual-track was introduced. They continue to work normally and don't have a type badge.

Standalone Facts

Single fact episodes that don't share entities with other facts are not clustered. They remain as first-class retrieval targets via direct episode search. This avoids creating single-fact "clusters."

Concept quality

Good concepts are:

• Specific and falsifiable — Making concrete claims, not abstract platitudes
• Generalized (for patterns) — Not just restating a single episode, but abstracting across multiple
• Verbatim (for facts) — Preserving exact values, configs, and technical details
• Conditional — Stating when they apply, not claiming universal truth
• Connected — Linked to related concepts via typed relations
• Grounded — Traceable back to source episodes

Topics

Concepts can belong to a topic — a knowledge area like "architecture", "product", or "infra". Topics scope consolidation (episodes are grouped by topic before processing) and retrieval (queries can be filtered to a specific topic).

Concepts without a topic are treated as general knowledge accessible from any topic context.

Managing concepts

# List all concepts
remind inspect

# View a specific concept with details
remind inspect <concept-id>

# Update a concept's summary
remind update-concept <id> -s "Refined understanding"

# Update confidence
remind update-concept <id> --confidence 0.9

# Reassign or clear topic (ID or name, same as remember)
remind update-concept <id> --topic product
remind update-concept <id> --clear-topic

# Soft delete
remind delete-concept <id>

# Restore
remind restore-concept <id>

TIP

Updating a concept's summary clears its embedding vector. The embedding is regenerated on the next recall query.