Back to Home

ODAM Documentation

API Reference & Guides

Memory Types

Understanding how ODAM organizes and stores different types of memories

Memory Architecture

Intelligent Memory Classification

ODAM automatically categorizes memories based on content type, importance, and relationships. This enables efficient retrieval and contextual understanding across different conversation types.

Core Memory Types

Personal Facts

Information about users: names, preferences, relationships, goals, and personal details.

"Alex is a developer from Kyiv who loves TypeScript"

Conversational Context

Key discussion points, decisions made, and important topics covered in conversations.

"Decided to use React for the new project"

Knowledge Base

Domain-specific information, procedures, and factual knowledge shared in conversations.

"API endpoint: /api/v1/users for user management"

Actions & Events

Record of actions taken, events occurred, and significant milestones in user interactions.

"Completed project setup on March 15th"

Memory Operations

+
Memory Creation

ODAM automatically extracts and stores relevant information from conversations using AI-powered entity recognition and fact extraction.

Process Flow
Input → Entity Extraction → Fact Validation → Memory Storage → Indexing

Memory Retrieval

Intelligent search across stored memories using semantic similarity, keyword matching, and contextual relevance scoring.

Retrieval Methods
• Semantic Search (Vector Similarity)
• Keyword Matching
• Temporal Filtering
• Entity-based Queries

Memory Updates

Dynamic memory updates when new information conflicts with or enhances existing memories, maintaining accuracy and relevance.

Update Strategies
• Conflict Resolution
• Information Merging
• Confidence Scoring
• Temporal Versioning

Advanced Memory Features

Graph Memory

Advanced relationship mapping between entities, concepts, and events for deeper contextual understanding.

Pro Feature
Available in Pro and Enterprise plans

Privacy Controls

Granular privacy settings, data retention policies, and user consent management for sensitive information.

Standard Feature
Available in all plans

Memory Lifecycle

1

Capture

Information is extracted from conversations and events

2

Process

AI analyzes content for entities, relationships, and importance

3

Store

Memories are indexed and stored with metadata and embeddings

4

Retrieve

Relevant memories are fetched based on current context

5

Update/Expire

Memories are updated with new information or expire based on policies

Learn More