Advanced Workflow Patterns
Once you're comfortable with basic workflows, you can design more sophisticated automation using advanced patterns.
Logic: branching, loops, and memory
Modern workflows aren't just straight lines. You can:
- Loop a step or a whole branch once per item in a list (iteration)
- Branch down different paths based on a value (switch), and cap how many items continue (limit)
- Remember values between runs with the cross-run data store
- Compute inline with expressions, and ask the runner for values up front with declared run inputs
These get full treatment in Looping, Branching & Logic Nodes and Expressions, Run Inputs & the Data Store.
Message-based workflows
For quick automations, you can create a workflow from a single prompt. Just describe a multi-step task in natural language — "Research AI trends, summarize the findings, and format as an HTML report" — and Nuveki will break it into steps, assign agents, and execute the pipeline automatically. This is great for one-off tasks where you don't need the precision of the visual editor.
Scheduling
Workflows can run on a schedule — daily, weekly, or on custom cron expressions with timezone support. This is useful for:
- Daily news digests
- Weekly report generation
- Recurring data analysis
- Automated monitoring and alerting
Set up scheduling from the workflow settings. Each scheduled run executes independently and logs results to the execution history.
Saving outputs to the Library
You can configure a workflow to automatically save its final output to the Library. This is useful for content pipelines — the finished article, report, or formatted document goes straight to your Library where it's indexed for RAG and available for future reference.
Best practices
Keep steps focused
Each step should do one thing well. A step that tries to research AND summarize AND format will produce worse results than three focused steps.
Use the right agent for each step
Don't use the same agent for every step. Match the agent's specialty to the task — use a Data Processor for extraction, a Content Summarizer for summaries, and an HTML Expert for formatting.
Test incrementally
Build your workflow one step at a time. Run after each addition to verify the output is what you expect before adding the next step.
Handle failures
If a step fails, the workflow stops. Design your prompts to handle edge cases — include fallback instructions like "If no data is found, output 'No results' instead of failing."
Subscription requirements
Every paid tier — and the Free tier — can build workflows; the caps scale with your plan:
- Free: up to 3 workflows, 20 steps each
- Starter: up to 25 workflows, 50 steps each
- Pro: up to 100 workflows, 200 steps each
- Team: unlimited workflows, up to 1,000 steps each
Higher tiers also raise execution concurrency, loop-iteration limits, data-store size, and how long run history is kept. Guests can't run workflows.