Crate halldyll_memory_model 

Halldyll Memory Model

Multi-user, multi-model memory system for distributed AI agents. Designed for cloud deployment with PostgreSQL + Redis backend.

Features

• Multi-user isolation: Complete user separation with RBAC
• Multi-model support: Manage memory across different AI models
• Distributed memory: PostgreSQL for persistence, Redis for caching
• Vector search: Semantic search with embedding similarity
• Conversation management: Full conversation history with context
• Fact extraction: Learn and store facts with confidence scoring
• Image metadata: Track generated images and their parameters
• Cloud-ready: Designed for Kubernetes/RunPods deployment

Architecture

├── core          : Types, errors, configuration
├── models        : Database models and entities
├── storage       : PostgreSQL + Redis persistence layer
├── embedding     : Vector generation and search
├── ingest        : Data ingestion and processing
├── retrieval     : Semantic search and ranking
├── context       : Prompt context building
├── user          : User management and isolation
├── engine        : Main orchestration layer
└── utils         : Utilities and helpers

Quick Start

use halldyll_memory_model::{Config, MemoryEngine};
use halldyll_memory_model::models::{Conversation, Message, Role, Memory};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create engine with default config
    let config = Config::default();
    let engine = MemoryEngine::new(config).await?;
    engine.initialize().await?;

    // Store and search memories
    let mut conversation = Conversation::new("Hello");
    conversation.add_message(Message::new(Role::User, "Hello!"));

    // engine.store_memory("user-123", Memory::Conversation(conversation)).await?;
    // let results = engine.search("user-123", "hello", 10).await?;

    Ok(())
}

Re-exports

pub use core::Config;

pub use core::MemoryError;

pub use core::MemoryId;

pub use core::MemoryResult;

pub use core::MemoryType;

pub use engine::HealthStatus;

pub use engine::MemoryEngine;

pub use context::ContextBuilder;

pub use context::ContextConfig;

pub use context::TokenEstimationMethod;

pub use embedding::EmbeddingGenerator;

pub use ingest::MemoryProcessor;

pub use ingest::ProcessingResult;

pub use ingest::ProcessorConfig;

pub use retrieval::MemorySearcher;

pub use retrieval::SearchOptions;

pub use retrieval::SearchResult;

pub use storage::PostgresStorage;

pub use storage::RedisCache;

pub use storage::StoragePool;

pub use user::Permission;

pub use user::PermissionChecker;

pub use user::UserManager;

pub use models::Conversation;

pub use models::Fact;

pub use models::ImageMetadata;

pub use models::Memory;

pub use models::Message;

pub use models::Role;

pub use models::TranscriptionMetadata;

pub use models::UserModel;

pub use models::UserRole;

pub use utils::Validator;

Modules

context

Prompt context building with retrieved memories

core

Core module with fundamental types and error handling

embedding

Embedding generation with ONNX Runtime

engine

Main orchestration layer for the memory system

ingest

Data ingestion and processing

models

Database models and entities

prelude

Prelude module for convenient imports

retrieval

Memory search and retrieval

storage

Storage layer for PostgreSQL and Redis persistence

user

User management and authentication

utils

Utility functions and helpers

Constants

AGENT_CREATOR

Creator

AGENT_NAME

Agent name

VERSION

Library version