Skip to main content
Normalized for Mintlify from knowledge-base/neurigraph-memory-architecture/hyperthyme-memory-framework/legacy-memory-recall-overview.mdx.

​
Recall: Persistent Conversational Memory System

​
Overview

Recall is a memory persistence layer for the AI brain that solves two fundamental limitations in current AI systems:
  1. Context window limits — Conversations eventually exceed what the AI can “see” at once
  2. Session persistence — Information is lost when a chat ends or a new session begins

​
How It Works

Recall continuously captures conversation content into simple markdown files at configurable intervals (e.g., every N tokens or based on other metrics). These files serve as a searchable memory archive that exists outside any single conversation.

​
The Flow

Conversation happens
       ↓
Every [configured interval], save conversation chunk to .md file
       ↓
Files accumulate over time as persistent memory
       ↓
Later: "Do you remember X?"
       ↓
AI checks current context → Not found
       ↓
AI searches recall files → Finds relevant file
       ↓
AI reads file → Now has full context
       ↓
AI responds with remembered information

​
Key Characteristics

  • Format: Plain markdown files (simple, readable, portable)
  • Trigger: Configurable intervals (token count, time, or custom metric)
  • Scope: Works across any chat session—not tied to a single conversation
  • Retrieval: Search-based lookup when current context lacks needed information

​
Why This Works

Traditional AI memory approaches often involve:
  • Complex vector databases
  • Embedding-based semantic search
  • Summarization that loses detail
Recall takes a simpler path: just keep the actual text. When you need it, read it. The AI can process natural language natively, so there’s no need to transform the memory into a different format—markdown files are already in the language the AI understands.

​
Use Cases

  • Recalling project decisions made weeks ago
  • Picking up a topic from a previous session
  • Cross-referencing information discussed in different chats
  • Building continuity in long-running projects
Part of the AI Brain architecture
Last modified on April 18, 2026