Evancence

A hybrid AI system: C++23 high-speed backend × Common Lisp self-evaluating middleware × React frontend

Overview

Evanescence is built on three layers:

• C++23 Backend — High-speed data processing pipeline with RocksDB storage and zero-copy Flatbuffers serialization
• Common Lisp (SBCL) Middleware — Agent session management with Meta-Circular Evaluator, hot code motility, and CLOS MOP-based self-modification
• React 19 Frontend — SPA dashboard for real-time agent monitoring and configuration

The "Quid of Molting"

Agents in Evanescence can evaluate their own performance, generate improved behavioral code, and replace their own logic at runtime — without downtime.

Project Structure

backend/        → C++23 high-speed data engine
middleware/     → Common Lisp (SBCL) agent middleware
frontend/       → React 19 + TypeScript dashboard
nix/            → Nix build configuration
docker/         → Docker Compose setup

Quick Start

# Using Nix (recommended)
nix develop

# Or with Docker
docker compose up --build

Architecture

See the full architecture document: https://forgejo.pontlevis.ddns.net/bamigos/evanescence/src/branch/main/docs/architecture.md

License

MIT