Skip to main content

Yggdrasil — Nine Realms Ecosystem

Nine Realms. One Tree.

Yggdrasil is a personal project ecosystem organized under the Norse Nine Realms metaphor. At its heart lies Lilith v5.0 — a dark-fantasy CLI agent with hybrid memory, swarm intelligence, MCP protocol, multi-provider LLM, skills framework, batch mode, and a real-time dashboard.

$ git clone https://github.com/BrierAinz/Yggdrasil.git
$ cd Yggdrasil
$ cp Asgard/Lilith/.env.example Asgard/Lilith/.env
$ cd Asgard/Lilith && lilith

What is Yggdrasil?

Yggdrasil is not a monorepo — it is a living ecosystem where every project has a purpose, a lifecycle, and a destination realm. Inspired by Norse cosmology, each realm serves a distinct function in the development lifecycle.

᛭ Key Principles

  1. One Project, One Realm: Every project lives in exactly one realm at any time. No duplicates, no chaos.
  2. Strict Lifecycle: The lifecycle is enforced: Idea → Muspelheim → [Target Realm] → Helheim (if it dies).
  3. Lilith the Orchestrator: A local LLM-powered agent with Telegram control, RAG memory, and sub-agent delegation.
  4. API-First: All communication with the AI flows through a REST Gateway. No cloud dependencies for the core.

The Nine Realms

Each realm has a single, well-defined purpose. Projects migrate between realms as they evolve, following strict rules defined in REGLAS_YGGDRASIL.md.

RealmThemePurpose
AsgardCore TechnologyPermanent core projects — Lilith, Gateway, LLM providers
VanaheimAI AgentsBots, agent profiles, communication protocols
AlfheimUI PrototypesVisual experiments, IDE integrations
SvartalfheimKnowledgeDocumentation, architecture decisions, deployment guides
MuspelheimActive DevelopmentSprint templates, hotfixes, active work
NiflheimResourcesDatasets, LLM checkpoints, training data
MidgardPersonal AppsCompleted, battle-tested applications
JotunheimMassive ProjectsLong-term, large-scope projects
HelheimGraveyardArchived code, legacy versions, regenerable trash

Lilith — The Crown Jewel of Asgard

The crown jewel of Asgard. A local-first AI assistant that runs entirely on your hardware via LM Studio, with no cloud lock-in. Control it from anywhere through Telegram.

Request Flow:
You (Telegram) → Telegram Bot → Gateway (FastAPI) → Lilith (Orchestrator) → LM Studio (Local LLM)

Features

  • ᛒ Hybrid Memory — Vector embeddings + knowledge graph + full-text search (FTS5). Automatic compression, entity extraction, and semantic recall across sessions.
  • ᚹ Swarm Intelligence — LLM-powered specialist agents (researcher, coder, writer, critic) with file locking, conflict resolution, and persistent sessions via SQLite.
  • ᛏ Task Scheduler — Cron-like job scheduling with persistent SQLite backend and REST API exposure.
  • ᛇ Multi-Provider LLM — LM Studio (local) + Kimi (remote) + any OpenAI-compatible provider. Automatic fallback, model auto-detection, and zero-config startup.
  • ᚦ Skills & MCP — Hot-reloadable skill packs with auto-trigger. MCP (Model Context Protocol) connects external tool servers dynamically. 35+ native tools.
  • ᚠ Plugin Architecture — Hot-pluggable tools with dynamic discovery, enable/disable, and custom tool registration.
  • ᛗ Real-time Dashboard — Web dashboard with WebSocket live updates, multi-pane layout, terminal widget, and system monitoring. Dark fantasy aesthetic.
  • ᚨ TOML Config & Resilience — Unified configuration in ~/.lilith/config.toml. Circuit breaker, graceful shutdown, error tracking, and automatic recovery.

Agents of Yggdrasil

Four active agents power the ecosystem, each a specialized LLM personality with a distinct role and context window.

AgentBackend ModelSpecialtyContext
ShalltearVenice AI (llama-3.3-70b)Classification, NL parsing, triage32k
AdánOllama (qwen2.5-coder:7b)Code generation, tests, refactoring8k
EvaGrok (grok-4-fast-reasoning)Long-context analysis, documentation128k
OdínKimi (kimi-for-coding)Deep analysis, research, creative writing262k

Non-agent components that complement the pantheon:

ComponentRealmRole
LilithAsgardCore orchestrator — CLI agent with hybrid memory, swarm intelligence, and command over all sub-agents
BifrostVanaheimTelegram gateway — the rainbow bridge connecting the ecosystem to the mortal plane
ForgeMasterMuspelheimBuild automator — model catalog, GPU detection, VRAM calculator
DashboardAlfheimReal-time terminal UI — system metrics, Git activity, live telemetry

Ecosystem Health

Current state of the major components across all realms.

ComponentRealmStatusDescription
LilithAsgardᛏ ActiveCore orchestrator v5.0 — CLI, memory, swarm, MCP, skills, batch, dashboard
GatewayAsgardᛏ ActiveFastAPI REST bridge exposing Telegram, Scheduler, Agents, Plugins
ForgeMasterMuspelheimᛏ ActiveModel catalog, GPU detection, GGUF/safetensors metadata, download manager
TerminalDashboardAlfheimᛏ ActiveTerminal UI with scanner, Git activity, sidebar search, Rich panels
LLM ProvidersAsgardᛏ ActiveMulti-provider: LM Studio (local), Kimi (remote), OpenAI-compatible, auto-fallback
Hybrid MemoryAsgardᛏ ActiveVector embeddings + knowledge graph + FTS5, auto-compression, entity extraction, session search
Swarm IntelligenceAsgardᛏ ActiveLLM-powered specialist agents with file locking, conflict resolution, persistent SQLite sessions
Skills & MCPAsgardᛏ ActiveHot-reloadable skill packs, MCP protocol, dynamic tool registration, 35+ native tools
Task SchedulerAsgardᛏ ActiveCron-like scheduling with REST API and persistent storage
DashboardAsgardᛏ ActiveReal-time web dashboard with WebSocket, multi-pane layout, terminal widget
TOML Config & ResilienceAsgardᛏ ActiveUnified config, circuit breaker, graceful shutdown, error tracking, auto-recovery
Telegram BotVanaheimᛏ ActiveRemote control interface for the entire ecosystem
VSCode ExtensionAlfheimᚦ PrototypeVisual interface for Lilith inside VS Code
WebsiteSvartalfheimᛏ ActiveGitHub Pages documentation site (this page)