LogoLogo
Documentation
DocumentationUptiq Agents
Documentation
DocumentationUptiq Agents
  • Getting Started
    • Introduction
    • Sign up to Developer Edition
    • Build Your First Agent
    • Developer Support
  • Core Concepts
    • Agent
      • Knowledge
      • Webhook
    • PII Masking
    • Sub-Agent
    • Intent
    • Workflow
      • Node
        • Input
        • Output
        • Loader
        • Display
        • API Node
        • Web Crawler
        • Table Write
        • Table Read
        • Ruleset
        • Upload Document
        • Javascript
        • Workflow
        • Loop
        • Document To Image
        • External Database
        • Storage Write
        • Storage Read
        • Fetch Document
        • Prompt
        • RAG Query
        • Vector Search
        • Emit Event
    • RAG
    • Model Hub
      • Entity Recognizers
    • Data Gateway
    • Rulesets
    • Code Snippets
    • Tables
    • Storage
    • Widget
  • Overview of GenAI
    • Introduction
    • Key concepts
      • Intent Classification
      • Inference
      • Generative AI Models
      • Large Language Models (LLMs)
      • Prompt Engineering
      • AI Agents
      • RAG (Retrieval Augmented Generation)
      • AI Workflow Automation
      • AI Agents vs LLM-based APPs
Powered by GitBook
On this page
  • What is a workflow ?
  • How to create a workflow ?
Export as PDF
  1. Core Concepts

Workflow

What is a workflow ?

Understanding UPTIQ AI Workbench Workflows

Workflows in UPTIQ AI Workbench represent low-code to no-code solutions designed to help developers implement business logic for specific AI agents. These workflows act as the backbone of AI agents, enabling them to process user queries and deliver desired responses or actions. Each workflow is essentially a sequence of interconnected nodes, with each node representing a specific action, capability, or logic. To effectively create workflows, developers must understand the capabilities of these nodes, as they are the building blocks for crafting the behavior of AI agents.

Key Features of Workflows:

  1. Visual Interface: Workflows are created using a visual interface, as seen in the image above. Developers can drag and drop nodes to design the logic without requiring extensive coding knowledge.

  2. Node-Centric Structure: Each node in a workflow represents a predefined capability, such as fetching data, transforming documents, or integrating external APIs. Developers need to learn these node capabilities, similar to learning functions in a programming language like Python.

  3. Execution Logic: The workflow begins when the reasoning engine interprets a user query and identifies the relevant intent. It then executes the corresponding workflow to process the query and return results.

  4. Pre-Built Capabilities: The system includes pre-built nodes for various actions, such as:

    • Data Operations: Fetching data from external databases, reading or writing to tables, filtering data, and querying graph databases.

    • Integrations: API calls, webhooks, CRM integrations, and notifications.

    • AI-Specific Operations: OCR processing, document conversion, and invoking large language models (LLMs) for text-based responses.

Analogy for Developers:

Think of workflows as programs or functions, and nodes as the syntax or commands you use to write them. Just as developers must learn Python syntax to write effective Python code, developers working with UPTIQ AI Workbench must understand the functionality and configuration of each node to build workflows efficiently. Mastering these nodes allows for the creation of sophisticated and tailored AI agent behaviors.

Example Workflow (Based on Image):

In the workflow depicted in the image:

  1. External Database Node: Fetches data from an external source.

  2. Data Processing Nodes:

    • Fetch Document: Retrieves a specific document.

    • Document to Image: Converts a document into an image format if it’s a PDF.

    • Pass Through: Directly passes the data if it's not a PDF.

  3. AI Logic Nodes:

    • Prompt: Sends a query to the LLM for generating intelligent responses.

    • Display: Presents the results to the end user.

  4. Custom JavaScript Nodes:

    • Adds flexibility by allowing developers to execute custom logic when needed.

Key takeaway for developers:

To leverage the full potential of UPTIQ AI Workbench workflows, developers should:

✅ Explore and understand the purpose and configuration of each node.

✅ Experiment with different workflows to see how nodes interact.

✅ Treat workflows as modular, reusable components of an AI agent's behavior.

By mastering workflows, developers can create powerful, efficient, and intelligent AI agents to meet specific business needs with minimal coding effort.

How to create a workflow ?

PreviousIntentNextNode

Last updated 1 month ago