Inception Core Server – Modular Memory & Agent Runtime

Inception Core Server is the foundational runtime behind the broader Inception project — a modular AI infrastructure designed around the belief that memory precedes intelligence.

At its heart lies the MCP architecture: Model–Context–Protocol, orchestrating LLMs, contextual memory (short, long, latent), and a structured communication layer for cognitive agents.

Highlights

• Modular Memory System:

  • short_term (in-memory HashMap)
  • long_term (SQLite via sqlx)
  • latent (vector memory via ChromaDB)

• LLM Integration:

  • Pluggable local or remote inference via REST
  • Compatible with Mistral (GGUF / llama.cpp) or remote APIs

• Agent Runtime:

  • One or more agents operate via a shared protocol layer
  • Uses custom REPL DSL: Sentience

• Structured Communication:

  • REST API for memory read/write
  • Latent memory embedding/query endpoints
  • Agent chat + sentience execution

Why It Matters

This isn't just a wrapper around an LLM.
It's a cortex simulator — where agents coordinate over memory, not prompts.

The long-term goal is autonomy — emergent behavior from structured storage, not fine-tuned weights.
Every line of code is part of a larger experiment:

Can intelligence grow from the ground up if memory is structured like biology?

Architecture

• src/icore/ – MCP layer (model.rs, context.rs, protocol.rs)
• src/memory/ – all memory backends
• src/agents/ – agent scaffolding + DSL behavior loader
• src/api/ – REST interface (Axum)
• Docker setup for full LLM + Chroma + SQLite runtime

Project Status

• Core server (Axum + memory)
• REST API routes
• Docker dev + prod config
• ChromaDB support
• Sentience DSL integration
• Multi-agent protocol expansion
• Reflection-driven decision layer

Inception is not just a server — it's the brainstem of a synthetic mind.
Everything else grows from this.