NMQ.Architect v3.0

Secure Provisioner

Creates a folder in projects/, writes API keys into .env, and locks them via .gitignore.

Git Automator

Pick an existing project from projects/ and automatically run git init + initial commit.

Deploy

Sends projects/<name> to a VPS (rsync/scp) · remote npm · PM2. Safe: projects/ only, admin auth, rate limit.

DB Manager

07 AI Market & Idea Tracker

Researcher gathers trends → Filter categorizes → Architect proposes programs and dashboard improvements.

08 AI Trend Analyzer

Scan · Analyze · Propose — parallel on daily AI news (HN, Reddit, TechCrunch, arXiv).

System Health

Last 20 errors from the database · refresh on click

🧠 Vector Memory

Index a project or remember facts — on-disk vector search (SQLite + embeddings).

🐝 Swarm Agents

Run multiple AI tasks in parallel. Concurrency is limited by the hardware profile (low=1, medium=2, high=3).

⚙️ Advanced options

Feature flags — all off by default. Enable in production via env vars or admin toggles (in-memory + app_settings).

AI Memory

Structured per-project memory (.ai-memory.json) — tasks, decisions, errors, and next step. Injected into AI only when a project is selected.

Domain Scanner & Auditor

Paste a domain list — checks Wayback history and live signals. Candidates: have an archive but look abandoned (NXDOMAIN, parking, soft-404…). DA/PA/DR are best-effort (Moz optional; otherwise heuristics).

Ghost Responder

IMAP → AI draft → you approve → optional SMTP. No automatic sending. Newsletters are skipped heuristically.

Skill Store

Deep Code–compatible SKILL.md packs. Enable skills the agent may use, preview procedures, or Run in Agent Studio. Vertical packs (finance / healthcare) are workflow skill packs — not certifications. Ascend: DeepSeek V4 models can run on Huawei Ascend (DeepSeek’s stack); NMQ is compatible via DeepSeek API/models — not Ascend-optimised hardware.

Automations

Interval jobs (off by default). Enable only what you need.

Ghost Agent · Automations

Triggers: onPush · onPR · onSchedule (cron) · onError

MCP Gateway

Central bridge: DeepSeek V4 ↔ Agent Studio via Model Context Protocol. Catalog + custom stdio/URL servers · Discover · tools · multi-session. Call tools directly or via skill mcp:<id>:<tool>. Secrets stay in .env as ${VAR} refs. External servers (DV4-MCP, deepseek-mcp-server) are community links — NMQ product, not affiliated. Positioning: MCP Gateway bridges DeepSeek V4 ↔ agents. Partner claims stay out of UI until accepted — see POSITIONING.

Dashboard Generator

Describe the dashboard you want — AI generates a full HTML/JS page saved under /generated/<slug>/ and open it for preview immediately.

Orchestration

Run multi-agent orchestration. Defaults to the local Agent Team. Set ORCHESTRATOR=openai-agents or ORCHESTRATOR=crewai in .env for alternate adapters.

Background Agents

Scheduled (recurring) and one-shot AI jobs that run while you are away. Results are stored in the database.

AI Usage

AI call history — by provider, errors, latency. Live builds use ai_usage_events; DEMO shows sample sandbox metrics.

NMQ Care Compass

Wherever you are — bills, policy rights, local health & customs, and project insurance orientation. Uses DeepSeek V4 when available. Assistive only — not medical, legal, or insurance advice. Prefer redacted documents (no full national ID / SSN).