Thunder Forge

Infrastructure/process layer for running self-hosted AI capabilities to follow Shared Goals.

Thunder Forge focuses on operating the compute and automation stack that can work with private data (finance, healthcare, family/child privacy) without relying on third-party hosted agents.

• GitHub org: https://github.com/shared-goals/
• This repo: https://github.com/shared-goals/thunder-forge

Shared Goals (idea)

A personal image of Joy and Happiness is treated as the base source of motives.

Details (RU): Shared Goals — use case and concept

• A motive is a reason to act (rooted in what brings joy/happiness).
• A goal is a direction or outcome shaped by one or more motives.
• When goals are shared among coauthors, motives combine and the overall dynamics increase.

“Text” as a forkable source of goals

Shared Goals is developed as a living Text that anyone can fork and rewrite into their own.

• Concept Text (evolving): https://github.com/bongiozzo/whattodo
• Common build submodule: https://github.com/shared-goals/text-forge

text-forge transforms a Text repository into:

• a website (with link-sharing functionality in the publishing format)
• an EPUB book
• a combined Markdown corpus suitable for AI usage (RAG/MCP agents and skills)

What Thunder Forge manages

Thunder Forge is the infrastructure/process layer for self-hosted execution of agents and skills.

Typical managed parts:

• Nodes: machines in a self-hosted cluster (e.g., several Mac Studios)
• LLMs on nodes: models served locally via Ollama (https://github.com/ollama/ollama)
• Automation/agents runtime: workflows and long-running processes (often via n8n)
  • n8n: https://github.com/n8n-io/n8n
• Skills: reusable tool capabilities agents can invoke
  • github_repo skills (agent skills): https://github.com/agentskills/agentskills

Note: n8n and skills catalogs are intended integration points. This repo is the operational “glue” and runbooks/specs layer to run them self-hosted.

Ecosystem map

flowchart LR
  SG["Shared Goals<br/>joy/happiness -> motives"] --> G["Goals<br/>(shared among coauthors)"]

  T["Text<br/>(forkable Markdown)"] --> TF["text-forge<br/>site + EPUB + combined corpus"]
  TF --> K["AI-ready corpus<br/>(RAG/MCP input)"]

  G --> A["Agents<br/>(plan and act)"]
  K --> A

  A --> S["Skills<br/>(callable capabilities)"]
  S --> N["Self-hosted nodes<br/>(cluster machines)"]

  N --> O["Local LLMs<br/>(Ollama)"]

  W["Orchestrator (optional)<br/>(n8n)"] --> A
  W --> S

Loading

Privacy & self-hosting principles

Shared Goals activities can involve highly sensitive data.

• Prefer self-hosted nodes and self-hosted agents for private domains.
• Keep data access least-privilege (skills should request only what they need).
• Treat secrets and tokens as production-grade (no plaintext in repos).
• Make agent activity auditable (logs, runs, and permissions).

Status

This repository is currently at an early scaffolding stage (see LICENSE).

Near-term intended contents include:

• node inventory/specs and bootstrap runbooks
• model/LLM deployment conventions (Ollama-managed)
• agent/workflow conventions (including optional n8n patterns)
• a skills registry format + examples

Localhost Mini App (restricted, no DB)

This repo currently contains a minimal vertical slice:

• GET /health (public)
• GET /mini-app/ (static Mini App)
• POST /api/mini-app/me (Telegram initData auth + admin allowlist)
• POST /api/mini-app/status (live reachability checks from inventory)

Quickstart (localhost)

1. Create tf.yml (kept out of git; it’s ignored by default):

server:
  bind: 127.0.0.1
  port: 8000
  reload: true

telegram:
  bot_token: "..."

access:
  admin_telegram_ids:
    - 123

mini_app_url: http://127.0.0.1:8000/mini-app/

settings:
  ssh:
    connect_timeout_seconds: 1.0
    batch_mode: true
  monitor:
    ssh_port: 22
    ollama_port: 11434
  hosts_sync:
    managed_block_start: "# BEGIN thunder-forge"
    managed_block_end: "# END thunder-forge"

nodes:
  defaults:
    ssh_user: you
    service_manager: brew
  items:
    - name: msm1
      mgmt_ip: 192.168.1.101

fabricnet:
  service_name: "Thunderbolt Bridge"
  ipv4_defaults:
    netmask: 255.255.255.252
    router: ""
  nodes:
    - name: msm1
      address: 172.16.10.2

2. Run:

• make sync
• make serve

The server binds to 127.0.0.1:8000 by default.

Notes

• This is intentionally restricted: if your Telegram user ID is not listed in tf.yml (access.admin_telegram_ids), the Mini App API returns 403.
• No DB is used; all state is computed on-demand from tf.yml.

Fabric + hosts setup (script)

This repo includes a KISS setup script that uses tf.yml to:

• configure fabric network IPv4 on each node (via SSH + networksetup on macOS; typically "Thunderbolt Bridge")
• generate a managed /etc/hosts block and push it to all nodes

Commands:

• make setup-env (configures fabric IPs; requires fabricnet.nodes entries)
• make hosts (writes artifacts/hosts.block)
• make push-hosts (writes artifacts/hosts.block and updates /etc/hosts on all nodes)