Skip to content

HOSTING-AND-DEPENDENCIES.md

Latest commit

 

History

History
109 lines (84 loc) · 3.3 KB

HOSTING-AND-DEPENDENCIES.md

File metadata and controls

109 lines (84 loc) · 3.3 KB

Hosting And Dependencies

This file describes the intended public deployment shape for HyperClaw-Max, based on the live OpenClaw body but trimmed to a public-safe install surface.

Baseline Host

Recommended baseline:

  • Hetzner or equivalent VPS
  • ARM64 or x86_64
  • 8 vCPU
  • 16 GB RAM
  • 160 GB SSD or equivalent room for logs, fixtures, and optional local-model lanes

Why this shape:

  • it already matches a proven reference footprint in the private body
  • it has enough room for gateway, services, tests, and a first adapter lane
  • it can host the public core without forcing a cluster or managed cloud stack

Recommended Network Shape

Recommended:

  • private reachability with Tailscale or equivalent
  • no broad public exposure by default
  • connectors enabled only after configuration and smoke checks

Why:

  • keeps the distro usable while reducing blast radius during bootstrap
  • fits the local-first, owner-controlled operating model

Core Runtime Dependencies

Required for the public core:

  • Python 3.11+
  • Node 20+
  • git
  • bash
  • rg
  • curl
  • systemctl --user

Useful during setup and support:

  • jq
  • gh

Core provider requirement:

  • at least one working provider path configured in the public config

Optional Adapter Dependencies

Local / hybrid model lane:

  • local OpenAI-compatible endpoint or equivalent local runtime
  • optional model assets and healthcheck wrappers

Connector lanes:

  • Telegram bot token
  • optional WhatsApp / webhook / OAuth credentials when those adapters are enabled
  • public connector templates are shipped under install/connectors/

Voice / browser lane:

  • optional voice provider credentials
  • optional RTC/browser assets and service templates

Repo-intel lane:

  • optional GitHub / repo adapter credentials
  • optional local upstream mirror tooling

Dependency Matrix

Surface Required For Public Core Optional Adapter Dependencies
Base runtime Python, Node, git, bash, rg, curl, systemd user services gh, jq
Models one provider path local endpoint, hybrid routing
Default pack public config + workspaces extra overlays
Connectors none for dry local use Telegram, WhatsApp, Gmail/Calendar/Drive hooks
Voice / browser none TTS / STT / RTC providers and service shells
Repo intelligence none GitNexus-style or equivalent adapter
Hosting Linux host Hetzner, Tailscale

Reference Surfaces In The Live Body

The live private system already proves these surfaces:

  • patch-aware gateway and validator flow
  • persistent agents with dedicated workspaces
  • task / delegation / watchdog lanes
  • Telegram plus richer hook ingress
  • local-first and hybrid routing lanes
  • voice/browser services
  • upstream and release advisory sync

Important:

  • that does not mean all of these are already extracted here
  • it means the public distro should describe them honestly and package them incrementally

Deployment Philosophy

HyperClaw-Max should be usable in three increasingly rich modes:

  • cloud-model only
  • local-model assisted
  • hybrid

The public distro should not require every mode or every adapter on day one.

A good first install is:

  • one Linux host
  • one provider path
  • one default pack
  • one baseline memory/context surface
  • one materialized target root produced by hyperclaw-first-run
  • optional channels and adapters enabled later