GhostGoat

Universal NLM Orchestrator

GhostGoat is a self-evolving multi-agent AI system that can handle any domain — coding, research, creative work, trading, analysis, planning, and more.

Core Architecture

ghostgoat_core.py          ← Main orchestrator (GhostGoat, RecursiveMemory, Task)
config/unified_config.py   ← Centralized configuration
frameworks/
  llm/                     ← Multi-LLM adapters (OpenAI, Anthropic, Gemini, Mock)
  monitoring/              ← Metrics, health checks, performance tracking
  api/                     ← REST API gateway
telegram_ghostgoat_universal.py  ← Telegram bot interface

Key Features

• Multi-LLM support — Claude, OpenAI, Gemini via unified adapter layer
• Recursive memory — Hierarchical memory with optional semantic search (sentence-transformers) and graph reasoning (networkx)
• Domain-agnostic agents — Specialist pools for coding, research, creative, trading, analysis, planning
• Self-evolution — LLM-generated code with human approval gates
• Graceful degradation — Works with zero optional deps (mock LLM + simple memory) up to full stack

Quick Start

# Minimal (no API keys needed)
export LLM_PROVIDER=mock
python ghostgoat_core.py

# With Claude
export ANTHROPIC_API_KEY=your-key
python ghostgoat_core.py

# Telegram bot
export TELEGRAM_BOT_TOKEN=your-token
python telegram_ghostgoat_universal.py

# Run tests
python tests/smoke_test.py

Configuration

Set via environment variables or ~/.ghostgoat/config.json:

Variable	Default	Description
LLM_PROVIDER	openai	openai, anthropic, mock
LLM_MODEL	gpt-4	Model name
MEMORY_BACKEND	chromadb	chromadb, memory, knowledge_tank
ANTHROPIC_API_KEY	—	Claude API key
OPENAI_API_KEY	—	OpenAI API key
TELEGRAM_BOT_TOKEN	—	Telegram bot token

Project Structure

GhostGoat/
├── ghostgoat_core.py              # Core orchestrator
├── telegram_ghostgoat_universal.py # Telegram interface
├── unified_integration.py          # Multi-system integration layer
├── unified_demo.py                 # System demo
├── ml_vault.py                     # ML training resources vault
├── setup_knowledge_tank.py         # Knowledge tank setup
├── config/
│   └── unified_config.py           # Configuration system
├── frameworks/
│   ├── llm/                        # LLM adapters
│   ├── monitoring/                 # Observability
│   └── api/                        # API gateway
├── core/                           # Extended subsystems
│   ├── orchestrator/               # LLM orchestrator
│   ├── memory/                     # Memory backends
│   ├── agents/                     # Agent frameworks
│   ├── ACS_SYSTEM/                 # Advanced Cryptographic System
│   └── modules/                    # Governance, learning, diagnostics
├── integrations/                   # External integrations
├── applications/                   # Domain applications
├── tests/
│   └── smoke_test.py               # Quick validation
├── scripts/
│   └── start.sh                    # Startup script
└── docs/                           # Architecture docs

Requirements

• Python 3.8+
• Optional: anthropic, openai, google-genai (for LLM providers)
• Optional: sentence-transformers, networkx, numpy (for semantic memory + graph)
• Optional: python-telegram-bot (for Telegram interface)
• Optional: chromadb (for persistent vector memory)