aelvoxim 1.1.2

Lightweight AI Cognitive Engine Framework — give your AI a memory

Aelvoxim — More than a chatbot. A brain that learns.

A self-hosted AI decision layer with multi-agent orchestration, anti-hallucination, memory, tool calling, desktop control, and autonomous learning.

---

Quick Start

Docker (recommended)

git clone https://github.com/macor24/aelvoxim.git
cd aelvoxim
export LLM_PROVIDER=openai LLM_API_KEY=sk-xxxxx
docker compose up -d
# Open http://localhost:9701

Pip

pip install aelvoxim

# Start the API server
aelvoxim server --port 9701

Windows

pip install aelvoxim
aelvoxim server --port 9701
# Open browser at http://localhost:9701

From source

git clone https://github.com/macor24/aelvoxim.git
cd aelvoxim
pip install -e .
python src/run_server.py 9701

---

Use Cases

Aelvoxim's layered memory, autonomous learning, and meta-cognition unlock scenarios that ordinary LLM wrappers and RAG pipelines cannot handle.

🧑‍🤝‍🧑 Long-Living Virtual Character

Most AI agents lose everything on restart — Aelvoxim doesn't.

• Conversations are stored in episodic memory and never reset
• User preferences and speaking style are extracted into semantic memory (persist across sessions)
• The forgetting curve naturally prioritizes recent interactions while preserving core facts
• After weeks of use, the character grows to genuinely know the user, instead of starting from scratch every session

📚 Living Code Documentation

Hook Aelvoxim into your Git workflow and let it maintain your project docs automatically.

• Connect via GitHub/GitLab webhook — every PR merge triggers re-learning
• Aelvoxim scans new code, extracts module APIs, and updates its knowledge base
• Conflict detection flags when an interface changed but docs weren't updated
• Team members ask "how does this function work?" and get answers grounded in actual code, not stale docs

🔍 Continuous Compliance Monitoring (Medical / Legal)

Meta-cognition + auto-validation as a regulatory watchdog.

• Store regulations, guidelines, or protocols in the knowledge base
• Curiosity-driven discovery periodically searches for new versions or updates
• Conflict detection catches contradictions between new and old knowledge automatically
• Output validation ensures every answer stays within compliance boundaries
• No manual review cycles needed — the system watches itself

🧠 Personal Learning Companion

Unlike ChatGPT where every conversation is isolated, Aelvoxim builds a persistent model of what you know.

• Tracks what you've learned and which concepts you find confusing
• Forgetting curve schedules review reminders at optimal intervals
• Extracts knowledge triples from your notes automatically (knowledge distillation)
• Gap analysis identifies blind spots and suggests what to study next
• Over time it becomes a second brain that knows your knowledge landscape

📊 Automated Business Intelligence

Turn curiosity-driven learning into a competitive radar.

• Define interest directions (competitor moves, industry news, tech trends)
• Curiosity engine autonomously discovers and fetches relevant content
• Extracted insights are stored as structured knowledge
• Conflict alerts fire when new information contradicts existing beliefs (e.g. a competitor changed pricing)
• No manual keyword rules — the AI decides what's worth learning

⚙️ Self-Tuning Service Bot

Deploy a support bot that tunes itself — no DevOps babysitting required.

• Out of the box it works with conservative defaults
• Auto calibration monitors real response quality and adjusts parameters (temperature, expert weights, memory thresholds) based on what actually works
• When traffic patterns shift (e.g. more technical questions at certain hours), the system adapts without a config push
• Knowledge gap analysis spots missing answers and automatically initiates learning
• Over weeks the bot silently improves — your only job is to keep it running

---

Features

🧠 Memory System

Your AI remembers like a human — with layers, decay, and doubt.

Capability	Description
4-layer memory	Working / Episodic / Semantic / Procedural
MemoryFusion	Inverted index + multi-layer retrieval
Forgetting curve	Knowledge decays naturally (Ebbinghaus)
Conflict detection	Finds contradictions, alerts the user
Confidence Matrix	5-dimensional trust score on every response

🔧 Expert System

A team of specialized AI experts working together.

Expert	Role	Community	Pro
LogicExpert	Rule reasoning, contradiction detection	✅	✅
MemoryExpert	Memory retrieval, fusion search	✅	✅
EthicsExpert	15 ethical rules, transparent & auditable	✅	✅
SafetyExpert	Local keyword + regex safety checks	✅	✅
CodeReviewExpert	Code style, complexity, secret scanning	✅	✅
EmotionExpert	Sentiment analysis, empathetic response	❌	✅
CreativeExpert	Creative writing, content generation	❌	✅
IntrospectionExpert	Self-reflection, quality audit	❌	✅
HypothesisEngine	Root cause hypothesis & validation	❌	✅

📚 Learning System

Capability	Community	Pro
Manual learning (API trigger)	✅	✅
7×24 auto-learning loop	❌	✅
Curiosity-driven discovery	❌	✅
Auto parameter tuning	❌	✅
Knowledge gap analysis	❌	✅
Post-validation audit	❌	✅

🔌 API & Integration

• RESTful API — FastAPI, Swagger docs at /docs
• SSE streaming — Real-time token-by-token output
• Multi-user auth — Email/password registration + API Keys
• Knowledge base CRUD — Add, search, update, delete knowledge entries
• Session management — Persistent conversations with history

---

Architecture

aelvoxim/
├── src/aelvoxim/
│   ├── core/         Cognitive engine (belief, metacog, reasoner, judge, DGM-H)
│   ├── learn/        Learning engine (learner, KB, validator, search, curiosity)
│   ├── memory/       Memory system (4 layers, fusion, decay, conflict, scoring)
│   ├── server/       API server (FastAPI, auth, chat pipeline, sessions)
│   ├── cortex/       Brain cortex (intent routing, scheduler)
│   ├── experts/      Expert system (orchestrator, 5+ experts, registry)
│   ├── storage/      Dual storage (SQLite local, PostgreSQL production)
│   ├── utils/        Helpers (i18n, paths, JSON)
│   └── edition.py    Edition gating (community/pro config control)
├── frontend/
│   └── chatael-v2/   ChatAEL — React + Tailwind chat UI
├── tests/            130+ test cases
├── docs/
├── CONTRIBUTING.md
├── SECURITY.md
└── LICENSE (MIT)

---

Community vs Pro

Dimension	Community (free, MIT)	Pro (license key)
Learning	Manual trigger	7×24 auto loop
Discovery	User-specified	Curiosity-driven
Tuning	Static defaults	Dynamic auto-tune
Experts	5 core	12 total (incl. advanced)
Validation	None	Scheduled auto-scan
Security	Local rules	SentriKit deep engine
License	MIT	License key

---

External Dependencies

Only 3 external packages — everything else is Python standard library.

Package	Purpose
bcrypt	Password hashing
fastapi	API framework
uvicorn	ASGI server

---

Why Aelvoxim?

• Lightweight — 3 deps, 34k lines of Python, no Docker required
• Self-contained — SQLite for single-user, PostgreSQL for multi-user
• Private by design — Your data stays on your machine, no telemetry
• Dual storage — SQLite for dev, PostgreSQL for production, same code
• Built-in safety — Ethics rules, safety checks, meta-cognition monitoring

---

License

MIT License — Copyright (c) 2026 macor24

See LICENSE for details.