Inception-Core

! Work in progress

---

Modular Memory Server for Cognitive AI Systems

Inception is a local memory server (cortex) designed for use in agent-based and cognitive systems such as synthamind and sentience. It provides structured memory, a query engine, STM/LTM hierarchy, and REST/gRPC/WebSocket interfaces.

---

Key Features

• Short-Term and Long-Term Memory model
• Query engine (simple DSL: subject=ball AND object=red)
• Pluggable storage backends: in-memory, file-based, RocksDB
• REST API with optional gRPC and WebSocket support
• Easy integration into other systems (via rpc_client)
• Fully local, no external dependencies

---

Getting Started

1. Clone and run

git clone https://github.com/nbursa/inception
cd inception
cargo run

2. Configure .env

API_ADDRESS=127.0.0.1:8080
MEMORY_BACKEND=inmemory

---

Usage Examples

Store a memory token

curl -X POST http://localhost:8080/memory/store \
  -H "Content-Type: application/json" \
  -d '{"subject":"ball","relation":"is","object":"red","tags":["toy"]}'

Recall memory

curl -X POST http://localhost:8080/memory/recall \
  -H "Content-Type: application/json" \
  -d '{"query":"subject=ball AND object=red"}'

---

Integrating from Other Projects

let client = InceptionClient::new("http://localhost:8080");

client.store_token(&MemoryToken {
    subject: "door".into(),
    relation: "is".into(),
    object: "closed".into(),
    tags: vec!["sensor".into()],
}).await?;

---

License

MIT © Nenad Bursać
Inspired by biological memory and the architecture of higher cognitive systems.