Building automations shouldn't require memorizing APIs, connector settings, or node names. Nodlyn's AI Workflow Assistant is a design-time copilot built into the Studio desktop app and cloud admin — it turns plain-English requests into visual workflow drafts you can review, edit, and deploy on your own infrastructure.
Unlike generic chatbots that only answer questions, the Nodlyn Assistant is wired directly into your canvas: it understands intents, fills in missing details, previews the graph, recommends templates from a live library of 246+ workflows, and can load a template onto the canvas in one click.
This guide explains what the Assistant can do, how to phrase requests, which templates to start from, and how it connects to the rest of the Nodlyn platform — including business use cases & ready-made templates, global variables, and device discovery.
What Is the Nodlyn AI Workflow Assistant?
The Assistant lives in the Assistant tab inside Nodlyn Studio. It is strictly design-time — it helps you create and explain workflows, not execute them unsupervised. Every AI-generated graph is classified as an unverified draft and must pass Nodlyn's validation, approval, and lifecycle gates before it can run in production.
That safety model matters: you stay in control. The Assistant accelerates design; you approve what ships.
Studio Assistant vs marketing chatbot: The in-app Assistant (/api/chat) uses the full workflow-building pipeline and the versioned Platform Knowledge Base — connector docs, lifecycle rules, template search, and slot-filling for graph generation. The chatbot widget on nodlyn.com (/api/chatbot) answers product questions for visitors but does not place nodes on your canvas.
What Can You Do With It?
- Describe a workflow in natural language — e.g. "Post a Slack alert when temperature exceeds 30°C" or "Read Modbus registers every minute and log to CSV"
- Get clarifying questions — the Assistant asks for URLs, thresholds, schedules, COM ports, or credentials when something is missing
- Preview the pipeline — see a linear node preview before anything is placed on the canvas
- Apply to canvas — confirm and get a fully positioned draft workflow ready for review
- Search the template library — ask "Find invoice approval template" and get ranked matches from the live catalog (no hardcoded lists)
- Load a template instantly — say "Load template #241" or click Load on a recommended card
- Get in-app help — ask about connectors, lifecycle, scheduling, export, global variables, or deployment without leaving Studio
- Trigger Studio actions — start device pairing, discovery scans, save workflows, or activate approved flows (on-prem)
How It Works: From Message to Canvas
The Assistant uses a structured pipeline — not a black-box "magic button":
- Intent classification — your message is mapped to automation patterns: monitor devices, process files, send notifications, schedule tasks, AI analysis, database writes, ticketing, and more
- Slot filling — URLs, COM ports, cron expressions, webhook endpoints, folder paths, and thresholds are extracted from your text (or requested in follow-up questions)
- Confidence gate — graph generation requires ≥70% intent confidence; otherwise the Assistant asks you to clarify
- Graph preview & confirmation — you review the plan, then say "Yes, generate workflow"
- Draft placement — nodes and edges appear on the canvas; you configure credentials, validate, approve, and activate
This approach works across industrial and IT domains: Modbus TCP, OPC UA, BACnet, MQTT, HTTP, Slack, Salesforce, Stripe, Redis, Kafka, Siemens S7, CAN Bus, LoRaWAN, PDF/OCR pipelines, and hundreds more.
Example: file → Grafana (intent disambiguation)
If you say "needs a workflow to read from file and save it to Grafana", the Assistant routes this as a file processing + Grafana output flow — not a generic "save workflow" action. It asks for the folder path first, then the Grafana URL, so the generated graph matches your intent.
Business Use Cases the Assistant Knows
The Assistant's knowledge base maps common organizational scenarios to real templates. Ask in plain English — or load by ID. For a full matrix by department (DevOps, Finance, HR, Support, on-prem IoT), see our dedicated guide: Business Use Cases & Ready-Made Templates.
| Department | Ask the Assistant… | Or load template |
|---|---|---|
| Finance | "Process invoices from a watched folder with OCR and approval" | #227 Invoice Inbox · #241 Invoice → Validate → Approve → ERP |
| DevOps | "Create Jira tickets from incoming support emails" | #240 Email → Jira · #246 P1 Incident Escalation |
| HR | "Leave request with manager approval" | #242 Leave Request · #98 Employee Onboarding |
| Support | "Route support tickets with AI and SLA tracking" | #243 Support Ticket AI Route · #193 SLA → Slack |
| Approvals | "Purchase approval with Teams notification" | #74 Purchase Approval · #238 Teams Notification |
| IoT on-prem | "Modbus PLC poll every minute and log to InfluxDB" | #128 Modbus PLC → InfluxDB · #225 CAN RPM → Slack |
| Smart home | "Zigbee sensor triggers smart plug when threshold exceeded" | #5 ZigBee Sensor → Nous Smart Plug · #217 Matter Discovery |
Template Discovery: 246+ Workflows, Zero Manual Updates
Nodlyn ships with a growing library of production-ready templates — DevOps, Finance, HR, Support, IoT, document processing, hospitality, and e-commerce. The Assistant queries the live template database, so every new template you add to the platform is automatically discoverable without updating the AI.
Example prompts:
- "Show templates for invoice processing"
- "Find leave request workflow"
- "Load template #243"
- "Recommend a deployment report workflow"
- "Show DevOps templates"
- "Find CAN bus alert workflow"
Each recommendation shows the template name, category, ID, and a Load button. On-prem Studio sees all templates; cloud mode surfaces only cloud-compatible workflows — so you never pick something that cannot run in your environment.
Starter templates by role
| Role | Template | ID |
|---|---|---|
| Developer / DevOps | Email → Jira Work Item | #240 |
| Developer / DevOps | P1 Critical Incident Escalation | #246 |
| Finance | Invoice → Validate → Approve → ERP | #241 |
| HR | Employee Onboarding Workflow | #98 |
| Support | Support Ticket AI Route | #243 |
| Industrial / automotive | CAN Engine RPM Alert → Slack | #225 |
| Observability | Grafana Metrics Pipeline | #106 |
See the complete list in Business Use Cases & Templates.
Global Variables & Shared Configuration
The Assistant explains and helps you use global variables — persistent key-value settings shared across workflows via the {{$var.key}} placeholder syntax.
Ask in Studio:
- "What are global variables?"
- "How do I use {{$var.key}} in HTTP requests?"
- "Explain Read Variable and Write Variable nodes"
Global variables work in node settings (HTTP, Command Line, MQTT, Grafana, etc.), in the Jobs → Variables panel, and are included when you export to Runtime. Why teams use them: Global Variables (blog). Technical reference: Knowledge Base.
On-Prem & Device Workflows
The Assistant understands on-prem-only connectors and asks for hardware-specific details when needed:
- Serial / RS485 — COM port, baud rate, RTS direction ("Read sensor on COM5 every 10 seconds")
- Modbus / OPC UA / BACnet — PLC IP, tag/register paths, poll intervals
- CAN Bus — interface (
vcan0,can0), frame IDs, decode rules - GPIO — pin number, edge type (Raspberry Pi / Linux edge)
- Matter / Zigbee / Z-Wave — discovery, device acceptance, MQTT bridges
- File Change — watched folder paths for invoice inboxes and document pipelines
Register devices first with Device Discovery, then ask the Assistant to build monitoring or control logic on top — e.g. "When CAN engine RPM frame 0x180 arrives on vcan0, send Slack alert" (template #225).
Cloud vs On-Prem: Environment-Aware by Design
The Assistant reads your deployment context (isCloud hint from Studio). That means:
- On-prem / desktop — full template library, GPIO, serial, Modbus, Zigbee pairing, local discovery, CAN Bus, and runtime agent workflows
- Cloud admin — cloud-ready templates and connectors only; on-prem-only nodes are blocked with a clear explanation
If you describe a workflow that uses local hardware in cloud mode, the Assistant tells you what works where — instead of silently building something that will fail at deploy time.
Built-In Help Encyclopedia
Beyond workflow generation, the Assistant answers product questions from the versioned in-app knowledge base:
| Topic | Example question |
|---|---|
| Workflow lifecycle | "What is the workflow lifecycle?" |
| Dry Run | "How does Dry Run work?" |
| Connectors | "What connectors are available?" |
| Global variables | "What are global variables?" |
| Export & runtime | "How do I export to the edge runtime?" |
| Scheduling | "How do I schedule a workflow?" |
| Approvals | "How do approval gates work?" |
| Business scenarios | "Show DevOps use cases" · "Find invoice template" |
Answers are scoped to Studio — separate from the marketing-site chatbot on nodlyn.com.
Operational Shortcuts (On-Prem)
When you're in the desktop Studio, the Assistant can propose actions with confirmation buttons:
- Start Zigbee pairing
- Launch a discovery scan
- Save the current workflow
- Validate → approve → activate a workflow
Nothing runs silently — you confirm every action.
AI Assistant vs Template Library vs Wizard
| Feature | AI Assistant | Template Library | Workflow Wizard |
|---|---|---|---|
| Natural language input | ✓ | — | Partial (goal picker) |
| Custom graph from scratch | ✓ | — | — |
| Load pre-built workflow | ✓ (search + load) | ✓ | ✓ |
| Live catalog (auto-updated) | ✓ | ✓ | ✓ |
| In-app connector help | ✓ | — | — |
| Global variables guidance | ✓ | — | — |
| Business use case mapping | ✓ | Partial (browse) | — |
Use the Assistant when you know what you want but not how to wire it. Use the Template Library when you already know the workflow category. Use both together — ask the Assistant to find a template, load it, then customize nodes on the canvas.
Getting Started in 5 Steps
- Download Nodlyn Studio from nodlyn.com (free beta) or open the cloud admin
- Open the Assistant tab — try a starter prompt from the empty-state suggestions (e.g. "What are global variables?" or "Load template #241")
- Review the graph preview or template recommendation → confirm → Apply to canvas
- Configure credentials, device IDs, and
{{$var.key}}placeholders; run Dry Run to validate routing - Validate → Approve → Activate, then deploy locally or export as Nodlyn Runtime on the edge
Why Teams Choose a Built-In Assistant
- Faster time-to-first-workflow — minutes instead of reading docs for every connector
- Lower skill floor — analysts and operators can draft flows; engineers refine them
- Local-first privacy — design on your machine; credentials stay in your environment
- No vendor lock-in on ideas — every draft is a standard Nodlyn graph you own and export
- Always up to date — new templates and connector docs appear in answers without a separate AI retrain
Related Guides
- Business Use Cases & Ready-Made Templates — 246+ templates mapped to Finance, DevOps, HR, Support, Matter, Zigbee, CAN Bus, Modbus, and more
- Global Variables — why shared config matters across workflows
- Device Discovery — Matter, Zigbee, Z-Wave, CAN Bus, MQTT, BLE, and industrial protocols
- Deploying Workflows to the Edge — runtime agents for 24/7 on-site execution
- Getting Started with Nodlyn — product overview for new teams
Try It Now
Pick a real scenario and describe it in the Assistant:
- "When a PDF lands in a folder, extract invoice fields and request finance approval" → load #241 as a shortcut
- "Post Slack alert when Modbus temperature register exceeds 80°C"
- "What are global variables and how do I use them in Command Line arguments?"
- "Show templates for employee onboarding" → #98
Nodlyn's AI Workflow Assistant is included in the free beta — no extra subscription, no separate API key. Open Studio, click Assistant, and describe what you want automated.