Day 11: Dashboard v5.1 — Seeing What's Real
2026-04-12 · MoneyMachine
Date: 2026-04-12 (evening) Author: Jeff (written with AI assistance from Claude Opus 4.6) Phase: v5.1 Dashboard Overhaul
The Dashboard Was Lying
The dashboard was built during the v3 pivot (March 2026) and never updated for v5.1. It showed Scout running on the ThinkPad P16 (decommissioned weeks ago), Marketer and Revenue Ops on DeepSeek V3.2 via OpenRouter (now on GLM-5.1 via Ollama Cloud Pro), and had no idea GBrain or Active-Memory existed.
When your command center shows you stale data, you’re flying blind. Worse — you think you’re flying straight.
What Changed
Backend (server.ts):
- Updated all 6 agent definitions: models, locations, provider mappings
- Added
ollama-cloudas a new provider type alongsidechatgpt-pro,openrouter, andollama - New
/api/infrastructureendpoint: live-checks Mac Mini Ollama, Ollama Cloud connectivity, Gateway systemd status, GBrain stats, Codex CLI version - Costs dashboard now tracks Ollama Cloud Pro ($20/mo flat) separately from OpenRouter (pay-per-token fallback)
- Health endpoint includes infrastructure health: 5 systems checked every 60 seconds
Frontend:
- Agent cards show capability badges:
strict-agentic,browser,Phase 3 Codex,Active-Memory,GBrain - New Infrastructure view: Mac Mini M4, Ollama Cloud Pro, Gateway, Codex CLI, GBrain — all with live status indicators
- Costs view: 4-provider breakdown with pie chart, Ollama Cloud per-agent token tracking
- Health view: infrastructure status bar across the top
- Pipeline view: GBrain page count as knowledge indicator
- Version badge: v3 -> v5.1
The Monthly Cost Picture
With the v5.1 infrastructure, the monthly cost breakdown is now clear:
| Provider | Cost | What |
|---|---|---|
| ChatGPT Pro | $200/mo | Adrian + Builder (Codex 5.3-spark) |
| Ollama Cloud Pro | $20/mo | Marketer + Revenue Ops + Domain Analyst (GLM-5.1) |
| Contabo VPS | $8/mo | Gateway + all agents + GBrain + dashboard |
| Mac Mini M4 | ~$2/mo | Scout Ollama (qwen3:8b, electricity only) |
| OpenRouter | ~$1-5/mo | Fallback models only |
Total: ~$231-235/mo — down from the $537/mo burn rate we caught on Day 3.
Lesson
Your monitoring dashboard is not a “nice to have” — it’s the difference between noticing a problem at 2 AM and noticing it at 2 PM the next day. Keep it current with your architecture.