Home

Amadeus-AI Wiki

v3.2.2 — Model Manager, Goal Tracking & Transport Refactor A secure autonomous AI operating layer built on Clean Architecture — autonomous tool execution, long-horizon planning, and multi-transport messaging unified under a single service layer.

---

What is Amadeus?

Amadeus is a FastAPI-based autonomous AI backend that provides:

Capability	Implementation
Multi-provider LLM routing	LlamaCpp (local GGUF) → Groq → Gemini with Redis-backed daily quota tracking
53 sandboxed tools	System · filesystem · productivity · information · communication
Long-term semantic memory	Qdrant vector store with all-MiniLM-L6-v2 embeddings
Goal management	GoalRepository tracks multi-session objectives
Multi-transport messaging	FastAPI · Telegram · CLI all share a single AmadeusService
Local model auto-download	ModelManager resolves and fetches models into Model/ on first run

---

What's New in v3.2.2

• Transport Layer — src/api/server.py replaced by discrete transport modules: fastapi_transport.py, telegram_transport.py, cli_transport.py
• Voice Service Removed — STT/TTS pipeline and all speech dependencies purged from the DI container
• ModelManager — src/infra/model_manager.py auto-downloads embed models and GGUF files into Model/ on first run
• Goal Management — create_goal, update_goal, list_active_goals tools; GoalORM backed by PostgreSQL
• Proactive Loop Governance — Per-session rate limiting (PROACTIVE_MESSAGE_LIMIT_PER_HOUR) and dry-run mode (PROACTIVE_DRY_RUN)
• Deployment — docker-compose.yml and deploy/amadeus.service systemd unit added
• Repository cleanup — Removed scratch/, wiki-publish/, stale dev scripts, and empty files

---

Wiki Pages

Section	Description
Architecture	Layer diagram, request lifecycle, LLM routing, model resolution, memory tiers, goal management
Quick-Start	Prerequisites, local installation, Docker
Configuration-Reference	All .env variables — LLM providers, model directory, proactive loop
Core-Systems	Semantic Router, Agent Orchestrator, HITL, Flash Memory Cache
Tool-Registry	All 53 tools organized by category
API-Reference	Chat, messaging, tasks, health endpoints
Redis-Quota-Tracking	Daily LLM quota counters and Redis key schema
Messaging-Integrations	Telegram and Email setup
Security-Model	Auth, prompt injection defence, filesystem sandboxing, tool execution safety
Deployment	Docker Compose, systemd service, environment setup
Development-Guide	Adding LLM providers and tools, testing, coding standards
Observability	Prometheus metrics, health probes, structured logging
Known-Limitations-and-Roadmap	Current gaps and planned improvements
Changelog	Version history

---

Quick Links

• Repository: github.com/adityatawde9699/Amadeus-AI
• API Docs (local): http://localhost:8000/docs (requires DEBUG=true)
• Liveness: GET /api/v1/health/live
• Readiness: GET /api/v1/health/ready
• Issues: GitHub Issues
• Security: SECURITY.md
• Contributing: CONTRIBUTING.md