Home Network Agent Architecture

Home Network Agent Architecture

Multi-PC Grok Build (grok-build-0.1 / xai-oauth) + Hermes Agent Sync • Primary: C:\Users\Uwe • 2026-06-17

Multi-PC Home Network • GrokSyncMaster / Hermes Gateway Primary PC C:\Users\Uwe Hermes + Grok CLI grok-build-0.1 + X tools Secondary PCs Home Network Sync via Git xAI Auth xai-oauth XAI_API_KEY OIDC + Grüner Mars Git Sync Repo grok-build-sync Rosary-mom • bidirectional Hermes Agent v0.16.0 • default profile Gateway: API connected Skills / Memory / Cron Artifacts / Sessions State: running (pid) Grok Build ~/.grok/ • config.toml grok-build-0.1 (default) x_keyword_search enabled models_cache refreshed X Tools + Grüner Mars @RosaryParish #GrünerMars archive x_search (xai-oauth) Core Skills • reliable-resume-x • artifact-manager • multi-pc-sync-master + resume, grok-sync Artifacts + Memory ~/.grok/artifacts hermes/artifacts MEMORY.md mirror Sessions & Logs state.db / sessions hermes/logs cron / kanban gateway_state.json sync hermes xai-oauth + XAI key x_search orchestrate Legend Grok / Frontend Hermes Core Sync / Artifacts X / Research Auth / Security Auth Flow

Auth & X Status

  • • xai-oauth: active (eurobitz@Jesus.tips)
  • • XAI_API_KEY: referenced
  • • x_search: confirmed (Grüner Mars post)
  • • OIDC expiry: monitored
  • • supports_backend_search: true

Gateway & Hermes

  • • State: running
  • • API Server: connected
  • • Discord/Telegram: retrying
  • • Login item: installed
  • • Config: grok-build-0.1 / xai-oauth

Multi-PC Sync

  • • grok-build-sync: .git present, clean
  • • grok-sync.ps1: available
  • • Artifacts: mirrored dirs
  • • Memory: MEMORY.md align
  • • Handoff: Reliable-Resume-X + Sync

Skills & Artifacts

  • • reliable-resume-x (X + older posts)
  • • artifact-manager (central + version)
  • • multi-pc-sync-master (orchestrator)
  • • + resume, grok-sync, hermes-agent
  • • Architecture artifact: this HTML
Hermes Multi-PC AI Agent Team • Visuals
⚙︎
Hermes Multi-PC
AI Agent Team
GROKbuild 0.1 + 4.5 + KIMI K3
2026-07-22
Slide 01

One overwhelmed orchestrator.
Five calm specialized agents instead.

Stop trying to be Researcher + Writer + Sync + Memory + Handoffs all at once.
😵‍💫
Single GrokSyncMaster
Doing everything
🔬
Research (LOGOS-max)
ON FIRE
✍️
Magnifica Writing
ON FIRE
🔄
Multi-PC Sync + Memory
ON FIRE
🤝
Handoffs + Checkpoints
ON FIRE
🛡️
Gateway + Subagent Health
ON FIRE
„one head, five fires, all burning at once“
Five focused agents
One job each. Nothing left on fire.
G
GrokSyncMaster
Orchestrator • Scheduler • Memory Mirror
✓
R
Researcher
LOGOS-max • Deep exegesis • Sources
✓
W
Writer
Magnifica Humanitas • HTML artifacts
✓
K
KIMI K3 Specialist
Long-context • Careful analysis
✓
S
Subagent Layer
Orchestration • Executors • Health
✓
Slide 02

One workspace.
One project per agent.

YOUR AGENTS
GrokSyncMaster
Researcher — LOGOS-max
Writer — Magnifica
KIMI K3 Specialist
Executor + Subagents
Human (PM Checkpoint)
Cross-PC sync • Memory mirror • Gateway health
GrokSyncMaster
• Gateway running • PID 16356 • Memory: ok
Current Task
Coordinating Psalm 138 blog post: Researcher → Writer → Sync Handoff
Last Sync
2026-07-22 14:32
Handoffs Today
7 successful
Active Profiles
Default + Rosary
Slide 03

Researcher → Writer → GrokSyncMaster → Synced

Three handoffs. Cross-PC memory mirror. Always human checkpoint.
R
Researcher
LOGOS-max finds facts, exegesis, GGG scores, Temple typology
→ Handoff Packet ready
W
Writer
Magnifica Humanitas style + self-contained HTML + GGG simulator
→ Handoff Packet + artifacts
G
GrokSyncMaster
Handoff review • Memory update • grok-sync.ps1 • Index update
→ Sync + Deploy
Synced & Deployed
rosary.health/blog updated
Memory mirrored across PCs
Gateway verified
👤 Human Checkpoint
Human stick figures stand at every handoff. Zero surprises. Run grok-sync.ps1 after each major step.
Slide 04
25 hours a week becomes 5–7.
Before — Manual Multi-PC
~25 hrs
Research + Insight
7h
Writing + HTML
8h
Sync + Memory
5h
Handoffs + Fixes
5h
After — Agent Team
5–7 hrs
Human now only reviews + approves checkpoints. Agents do the rest.
Slide 05
Mental load & sync friction vs structured team
Manual sync errors / drift
High
Context loss between agents
High
With GrokSyncMaster + Handoff Packets
Low
One orchestrator. Explicit packets. Memory mirror.
Slide 06
The old way
✕ One human or single Grok instance trying to hold Research, Writing, Sync, Memory, and Subagents in head.
✕ Context switching. Lost handoffs. Duplicate work across PCs.
✕ Manual grok-sync.ps1 forgotten. Memory drift.
✕ Human becomes the bottleneck for everything.
Same problems, every week. New excuses for delay.
The agent way
✓ Each agent has one narrow, clearly defined job with 5-block description.
✓ Structured Handoff Packets. Explicit checkpoints. Human only reviews.
✓ GrokSyncMaster owns orchestration + memory mirror + sync enforcement.
✓ One clean system. Corrections logs. Measurable improvement.
Four agents + orchestrator. One afternoon to set up. Hours saved every week.
Baseline captured 2026-07-22. Use the metrics tracker file + todo list to measure progress.
Always run grok-sync.ps1 after structural changes.

Schreibe einen Kommentar