How to use Rayven’s Generative AI capabilities to enrich data, power automations, and add LLM-driven intelligence directly into your application logic.
Overview
Rayven’s Generative AI capabilities let you integrate large language models (LLMs) directly into your workflows to drive advanced automation, intelligent data processing, and content generation. This is not about chatbots or conversational agents—it's about using LLMs as powerful tools to transform, summarise, classify, extract, and generate data as part of your end-to-end application logic.
Whether you're structuring unstructured data, transforming documents into datasets, generating dynamic content, or making real-time decisions—Generative AI in Rayven is fully integrated into the platform’s drag-and-drop workflow engine, so you can use it deterministically and reliably, just like any other logic or connector node.
Using Generative AI in Workflows
Generative AI fits seamlessly into Rayven’s workflow engine and behaves like any other logic component—meaning it can receive inputs, produce outputs, and be chained with other nodes to support multi-step processing. This makes it ideal for integrating advanced AI capabilities directly into your application logic.
Here’s how it typically works within a workflow:
▸ Accept dynamic inputs
Generative AI nodes can receive structured or unstructured data—such as uploaded documents, text fields, or values from tables—as part of the workflow payload. This data can be used to dynamically build context for the AI process (e.g. creating a prompt or extracting content).
▸ Perform AI-powered transformation
Once triggered, the Generative AI step can:
-
Extract specific fields from content
-
Summarise or rewrite text
-
Categorise or tag records
-
Generate structured outputs (e.g. JSON rows or formatted summaries)
These operations are typically defined by the logic upstream and refined by prompt structure.
▸ Chain into downstream logic
The AI-generated output can then be passed to:
-
JavaScript nodes for parsing or validation
-
Rule Builder nodes for branching decisions
-
Table update nodes for storage
-
Frontend nodes (such as UI Code widgets) for display
-
Output nodes for external system updates
▸ Support iterative and parallel use
You can use multiple AI steps in a single workflow, either:
-
Sequentially—for multi-stage refinement or comparison
-
In parallel—for evaluating multiple perspectives or options
This lets you build workflows that simulate complex reasoning or generate fallback options.
Working with Structured and Numerical Data
Generative AI in Rayven is not limited to unstructured text—it can also be used to interpret structured datasets and perform high-level analysis on numerical information.
For example, you can:
-
Use workflow logic to generate a dataset (e.g. filtered records, calculated metrics)
-
Format that data into a CSV or tabular format using JavaScript or logic nodes
-
Write the file to an FTP folder using an Output to FTP node
-
Point the AI node to the file path and prompt it to analyse the contents
Typical instructions might include:
-
“Summarise the top trends in this dataset”
-
“Identify anomalies or outliers”
-
“Convert this dataset into a human-readable summary”
This pattern is especially useful when working with complex time series, multi-row reports, or log-based data that benefit from high-level insight generation.
This allows you to inject intelligent interpretation of structured records into your applications—without needing to manually pre-code all possible logic paths.
Example Use Cases
Rayven’s Generative AI capability can support a wide range of application features:
-
Document Processing Pipelines
Extract structured fields (e.g. totals, names, dates) from uploaded PDFs, Excel files, or text-based documents. -
Automated Summary Generation
Create concise summaries of logs, reports, or user-generated content for display, communication, or tracking. -
Content Classification and Tagging
Categorise requests, entries, or descriptions using LLMs to assign types, priorities, or responsible teams. -
Dynamic Message Generation
Automatically generate templated messages (e.g. emails, instructions, alerts) that combine static fields with AI-generated phrasing. -
Cross-Format Conversion
Convert free-form inputs into structured JSON or CSV for downstream logic, storage, or export. -
Structured Data Analysis
Generate insights from numerical datasets (e.g. workflow-generated CSVs) by prompting the AI to identify trends, patterns, or anomalies. -
Decision Support and Recommendations
Ask the AI to recommend next steps, detect risks, or suggest actions based on contextual inputs. -
Data Enrichment
Augment partial records with inferred or summarised details—for example, inferring sentiment, estimating missing values, or generating metadata.
Summary
Rayven’s Generative AI features allow you to embed the power of LLMs into your applications—without compromising on structure, control, or reliability. By treating AI like any other node in the platform, you gain full control over when and how it runs, what data it uses, and where the output goes. This makes Generative AI not just a tool for insights, but a core part of your automation, transformation, and application design strategies.
FAQs
How is Generative AI used in Rayven?
As part of workflows, using AI nodes like Open AI to process or generate data, support automation, and enhance logic.
Is this only for text-based use cases?
No. While LLMs return text, the use cases include parsing documents, classifying data, generating structured records, and more.
Can I control what prompt the LLM sees?
Yes. Prompts can be static or dynamically built from workflow inputs, table rows, or external data sources.
Is Rayven hosting its own LLM?
Not yet, but it's on the roadmap. Currently, the Open AI node integrates with services like OpenAI via API.
Do I need to code to use it?
Not necessarily. Prompts and inputs are set through the node interface. JavaScript nodes can be added if you want to further manipulate outputs.
Can I use this to automate multi-step logic?
Yes. You can use multiple AI nodes in sequence or in combination with conditional or rule nodes to support advanced workflows.