Dify Workflow Basics

What you'll learn

• Explain what Dify is, its app types, and how to choose between workflow and chatflow apps.
• Build workflows from start, LLM, knowledge retrieval, logic, code and end nodes with clean variables.
• Set up a knowledge base with sensible chunking and retrieval settings, and recognise retrieval failures.
• Route and loop with if/else, question classifiers, iteration and variable aggregation, and test routing accuracy.
• Call published workflows through the API in blocking and streaming modes from your own code.
• Debug, cost, secure and maintain Dify apps in production.

Syllabus

Getting to Know Dify

1. What Dify Is
2. App Types: Chatbot, Agent, Chatflow and Workflow
3. Setting Up: Cloud or Self-Hosted, and Model Providers

Workflow Building Blocks

1. Start Node and Variables
2. LLM Nodes and Prompt Templates
3. End Node and Outputs

Knowledge Bases and Retrieval

1. Creating a Knowledge Base and Chunking
2. Retrieval Settings: Search Mode, Top K and Threshold
3. Grounded Answers From Retrieved Context

Routing, Classification and Iteration

1. If/Else Branches
2. Question Classifier Node
3. Iteration Over Lists
4. Variable Aggregator and Merging Branches

Code, Extraction and External Calls

1. Code Nodes
2. Parameter Extraction
3. HTTP Requests, Tools and Plugins

Chatflows, Memory and Agents

1. Chatflows Versus Workflows
2. Memory and Context Length
3. Agent Nodes and Tool Use

Publishing and Calling Apps

1. Calling a Workflow Through the API
2. Streaming Responses
3. Publishing, Web Apps and Versions

Debugging, Cost and Production Readiness

1. Logs, Tracing and Debugging
2. Cost per Run
3. A Production Readiness Checklist