The live dashboard is the command surface for Kaizen's local AI stack: private chat, model routing, service lifecycle controls, memory, search governance, audit history, Communications Hub, Max SMS, voice-call alerts, reminders, Beacon workflows, image workflows, HomeAuto, users, groups, notifications, and system prompts. This public page shows what that console does without exposing private controls or internal data.
Kaizen's dashboard is built as a working operator console, not a marketing screen.
First-party Kaizen Chat with project sync, conversation archive, local Max models, cloud proxy routing, streaming jobs, and text-to-speech playback.
Start, stop, restart, inspect, and log the registered stack: Gateway, Runtime, Search, Memory, Voice, IMGEN, Communications, voice-call alerts, Reminders, Beacon workflows, HomeAuto, cloud proxies, and support services.
Queue-Lite audit runs, model performance cards, report history, email delivery, audit schedule controls, and stack health signals.
Database-backed module registry for dashboard routes, pages, links, health checks, dependencies, operator actions, and launch surfaces.
Local-first provider routing — a private metasearch layer first, a governed paid provider as fallback — with trigger planning, cache controls, quota tracking, dry-run tests, decisions, and search logs.
A full planning workspace: Gantt views with saved layouts, schedule-quality audits, staged change sets with human-only approval, structural change planning, and a schedule-aware assistant.
A standing health collector, live service probes, and a health panel — the same lane the assistant reads when an authorized user asks how the platform is doing.
Rebuilt access and tier management: who exists, what group they belong to, which models, tools, and lanes each tier receives, enforced centrally at the gateway.
A backup center with scheduled sync and retention policy, plus an encrypted local secret vault that replaced flat env files across the stack.
Private memory review and management through the Orchestrator, keeping the memory API behind the dashboard boundary instead of exposing it publicly.
SMS contact sync, conversations, voice-call alerts, inbound/outbound message history, Telnyx delivery status, deterministic STOP/START/HELP handling, and Max SMS queue state.
Known-user reminder targets, pending and due counters, natural-language reminder parsing, and delivery through SMS, email, or Pushover.
Scheduled source checks, local transcription, classification, replay handling, notification state, watchdog settings, and delivery-receipt verification.
The operator console manages the stack. The iOS app is the private client that carries the same model, memory, voice, image, and connector controls into daily use.
The console gives the owner a private operating surface for service health, runtime lifecycle, audits, module registry, user tiers, memory controls, public edge limits, Communications Hub, voice-call alerts, reminders, Beacon workflows, HomeAuto safety, IMGEN, and chat history diagnostics.
The app signs into Kaizen, syncs chat history and projects, selects local or governed cloud models, uses memory, routes image generation, manages MCP connectors, and exposes Advanced Voice without handing private tools to the public web.
The dashboard ties together model execution, operator safety, public access, and real-world automation.
Owns auth, profiles, archive sync, local chat jobs, search, memory injection, HomeAuto routing, and the OpenAI-compatible dispatch surface.
Visible model filtering separates local Max modes, private admin tools, public-safe models, and optional cloud model proxies.
KaizenAI Open has a kill switch, queue controls, daily limits, voice toggles, model/release status, and bounded activity logs.
Inbound SMS, operational call alerts, outbound delivery receipts, contact sync, conversation state, provider IDs, STOP/START/HELP handling, and Max SMS worker visibility.
Real-time inbound calling is live: approved users talk to Max on a real phone line with local speech processing, per-call logging, and an operator kill switch beside the existing call-alert workflows.
Schedule quality, logic audits, critical-path reads, staged typed changes, structural simulation, and an AI planner that is server-side barred from ever applying its own proposals.
Durable SQLite scheduling, known-user targets, channel selection, and Max-created reminder commands across chat and SMS.
Run log, saved-audio audit modes, source recordings, local transcription, classification, watchdog settings, and final SMS delivery proof.
IMGEN tracks workflow registry, intent routing, Fast/Balanced/Quality profiles, model and node presence, recent routing, and safe test runs.
Operator-gated Hubitat and UniFi automation status with safety boundaries. Public chat cannot control devices or security systems.
Kaizen profile management, default models, group policy, dashboard/chat/search/memory/tool access, and per-user visibility controls.
Email and Pushover alert routing for audits, stack health, security events, service monitor state, quiet windows, and delivery history.
Dashboard-managed model prompt files, assigned local brains, sampling settings, and cloud prompt configuration for controlled behavior changes.
The public site shows the shape of the system. The dashboard operates the live internals behind authentication.
The point is not to expose the dashboard. The point is to expose a narrow, safe experience while the dashboard controls the real platform.
Visitors can use KaizenAI Open through the constrained public chat lane with queueing, daily limits, public-safe behavior, and no internal tool access.
The dashboard requires Kaizen authentication and group authorization. Dashboard access and private chat access are separate policy gates.
Private memory, home controls, stack lifecycle actions, prompt edits, user management, and internal diagnostics are never published on the public site.
Project Kaizen and KaizenAI: cybersecurity, infrastructure intelligence, local-first AI orchestration, the AI Orchestrator Console, KaizenAI iOS app, private Max assistant workflows, public KaizenAI Open chat, search, memory, voice, image generation, and MCP connectors.
INTRAC.NET est. 1998