Inception Core Server – Modular Agent Platform with Memory, Context, and Protocols

Design Philosophy

I believe nature is the ultimate architect.
Humanity didn’t invent intelligence — we’ve just learned to copy its patterns.

Inception Core Server isn’t engineered in the traditional sense. It’s grown — modular, layered, and memory-first.
It models cognition from the ground up using:

• Short-term memory (in-memory)
• Long-term memory (SQLite)
• Latent memory (ChromaDB, vector store)

What It Is

A memory-integrated agent runtime written in Rust.
It exposes a clean HTTP API (via Axum) for setting/getting memory, embedding vectors, querying latent space, and running agents powered by the Sentience DSL.

Architecture Highlights

• Memory Layers:
  • mem.short: volatile context
  • mem.long: persistent knowledge (via SQLite)
  • mem.latent: semantic vector memory (via ChromaDB)
• Modular Protocol Layer:
  • Agents communicate over HTTP
  • Each agent can embed, recall, and reason via structured inputs
• Sentience Integration:
  • DSL for memory-aware agent behavior
  • Agents can reflect, embed, and trigger conditionals based on memory state
• LLM optional: fallback generation is supported (e.g. via Mistral/GGUF)

Why It Exists

Modern AI stacks often hard-code intelligence into weights.
But without memory, there is no continuity — no identity, no reasoning, no self.

Inception flips this: memory is the system.

Everything is memory-driven. Agents have no “model” beyond their interaction with structured memory.
This aligns more closely with how intelligence emerges in biological systems — via stratified memory and recursive feedback.

Project Status

• HTTP memory API (Axum)
• Sentience REPL support
• SQLite + ChromaDB integration
• Agent loading from agent.sent
• WIP: semantic reflection and multi-agent loops

Usage

git clone https://github.com/nbursa/inception-core-server
git clone https://github.com/nbursa/sentience ../sentience

docker compose up

Server will auto-load agent.sent if present and expose memory APIs at /api/*

Final Thought

This isn’t a framework — it’s the foundation of a brain.

You don’t use it to build apps. You grow agents with it — layer by layer, memory by memory.

— Nenad