diff --git a/html/docs-clawdie-si/docs/index.html b/html/docs-clawdie-si/docs/index.html index 4494470..4f24ec0 100644 --- a/html/docs-clawdie-si/docs/index.html +++ b/html/docs-clawdie-si/docs/index.html @@ -87,13 +87,13 @@
- v0.7.2 — Docs Redesign + Skills Artifact Plan
- Introduces clawdie-hostd: a root daemon on a Unix socket that
- handles all privileged host operations (bastille, zfs, pf, service) via
- whitelisted op handlers. PF RDR is now the canonical web edge — nginx runs
- inside the cms jail, traffic arrives via rdr pass
- rules written by setup/pf.ts. Self-healing control plane checks
- at startup and every 5 minutes via src/controlplane.ts.
+ v0.8.0 Warden — Wayland-First Display, Full Subnet Layout, Security Docs
+ Worker jails now use cage + Chromium for agent browser automation —
+ no bhyve VM required. Full subnet layout defined (.1–.10 + .101+ workers).
+ Shared ZFS pkg cache wired into all jail steps. Wayland stack and bhyve
+ prerequisites in host baseline by default, no feature flags. New architecture
+ docs: jail vs Docker IPC, Wayland display model, prompt injection research
+ with hardening checklist.
Clawdie is a Self Hosted AI System built on top of FreeBSD. It follows the NanoClaw upstream line, but runs with native jails, PF, ZFS, and a - local built-in knowledge layer designed to reduce first-install LLM + local built-in knowledge layer designed to reduce first-install LLM setup friction.
@@ -229,7 +229,7 @@- Clawdie is moving beyond chat: setup, maintenance, docs lookup, and - machine-related workflows. The CNC module is currently a concept. Split - brain keeps stable operator knowledge separate from changing user context. - No public roadmap is promised yet. + The long-term goal is a CNC-enabled AI assistant — a system that understands + shop workflows, machine state, and operator context well enough to be genuinely + useful at a workbench, not just a chat window. That is the design target. + We are not there yet, and we are not pretending otherwise.
- Agent System Skills carries the stable operating playbook. User/Agent Memory - carries shop, operator, or agent-specific context. That separation makes the - system easier to trust, audit, and grow. + What we are building now is the foundation that makes it possible: a stable + agent runtime, jail isolation, a privilege model that can be trusted in a real + shop environment, and Stripe built into core so the agent can eventually manage + its own hardware development lifecycle — from sponsorship intake to prototype + procurement. Code has little cost. Hardware is a different story.
+ ++ When the sponsorship path is ready, it will live inside the agent itself — + not on a separate campaign page. Early sponsors will be the first to access + the CNC prototype. If this direction matters to you, the simplest thing you + can do right now is run the software, give feedback, and + become a sponsor. + That is how a prototype becomes viable. +
+