dify-workflow

name	dify-workflow
description	Dify AI workflow platform — LLM apps, knowledge bases, agents, workflow orchestration, API deployment. Use when working with dify workflow.
domain	core
author	oyi77
license	Apache-2.0
subdomain	core-platform
tags	["ai-agent","api","dify","infrastructure","memory","self-improvement","workflow"]
version	1.0.0

Overview

Dify is an open-source platform for building LLM applications with a visual workflow builder. It supports chatbots, agents, text generators, and complex workflows with knowledge bases, tools, and conditional logic.

Capabilities

• Build LLM apps visually (chatbot, agent, text generator, workflow)
• Create and manage knowledge bases with document upload
• Design workflows with nodes: LLM, knowledge retrieval, code, HTTP, conditional
• Deploy apps via API, embed, or shareable link
• Manage prompts, variables, and conversation memory
• Integrate with OpenAI, Azure, local models via Ollama

When to Use

Trigger phrases:

• "dify workflow"

• "Dify AI workflow platform — LLM apps, knowledge bases, agents, workflow orchestr"

• Building LLM applications without coding

• Needing visual workflow design for AI pipelines

• Wanting knowledge base RAG without infrastructure setup

• Prototyping AI features quickly

• Deploying AI apps as APIs

When NOT to Use

• Task is outside your authorization scope
• You need to implement controls (use implementing-* skills)
• Task is about analysis, not action (use analyzing-* skills)
• You don't have access to target systems
• Task requires compliance expertise (consult professionals)
• Task is about defense, not offense (use defensive skills)

Pseudo Code

# Example workflow for this skill
def execute(input_data):
    # Step 1: Validate input
    if not input_data:
        raise ValueError("Input data is required")

    # Step 2: Process core logic
    result = process(input_data)

    # Step 3: Validate output
    validate_output(result)

    return result

App Types

Type	Use Case
Chatbot	Multi-turn conversation with memory
Agent	Tool-using autonomous assistant
Text Generator	Single input → output transformation
Workflow	Complex multi-step pipeline

Workflow Node Design

Start → Knowledge Retrieval → LLM → Conditional → HTTP Request → End
                                    ↓
                              Code Processing → End

API Deployment

# Get API endpoint from Dify dashboard
# POST /v1/chat-messages
curl -X POST 'https://api.dify.ai/v1/chat-messages' \
  -H 'Authorization: Bearer {api_key}' \
  -H 'Content-Type: application/json' \
  -d '{
    "inputs": {},
    "query": "What is our return policy?",
    "response_mode": "streaming",
    "conversation_id": "",
    "user": "user-123"
  }'

Knowledge Base Setup

# Via API
import requests

# Upload document
files = {'file': open('company_docs.pdf', 'rb')}
response = requests.post(
    'https://api.dify.ai/v1/datasets/{dataset_id}/documents',
    headers={'Authorization': 'Bearer {api_key}'},
    files=files,
)

# Create knowledge base
response = requests.post(
    'https://api.dify.ai/v1/datasets',
    headers={'Authorization': 'Bearer {api_key}'},
    json={
        'name': 'Company Knowledge',
        'indexing_technique': 'high_quality',  # or 'economy'
    },
)

Workflow Variables

# Input variables
user_query: string
user_language: string

# System variables
conversation_id: string
user_id: string

# Output variables
answer: string
sources: array
confidence: number

Self-Hosted Deployment

# Docker
git clone https://github.com/langgenius/dify.git
cd dify/docker
cp .env.example .env
docker-compose up -d

# Access at http://localhost:3000

Integration Patterns

# Python SDK
from dify_client import DifyClient

client = DifyClient(api_key="app-xxx")

# Chat
response = client.chat_messages(
    query="Hello",
    user="user-123",
    conversation_id="",
)

# Completion
response = client.completion_messages(
    inputs={"text": "Summarize this..."},
    user="user-123",
)

Common Patterns

Pattern	When to Use
Knowledge Retrieval node	RAG over uploaded documents
LLM node	Generate text with prompts
Code node	Custom Python/JS logic
HTTP node	Call external APIs
Conditional node	Branch based on variables
Variable Aggregator	Merge outputs from branches
Template Transform	Format output with Jinja2

Error Handling

Error	Cause	Fix
Knowledge retrieval empty	No documents indexed	Upload and process documents
Token limit exceeded	Prompt + context too long	Reduce knowledge retrieval count
Workflow timeout	Long-running HTTP/code node	Increase timeout or optimize
Model not available	API key or model config issue	Check model provider settings

How to Use

1. Invoke the skill when relevant domain keywords appear in the request
2. Provide required inputs as specified in the skill definition
3. Review the output for correctness before delivering to the user
4. Combine with related skills for complex multi-step workflows

Verification

After completing this skill, confirm:

• Output meets the defined quality and completeness requirements
• All prerequisites are verified and documented
• Error handling covers edge cases
• Results are accurate and actionable

Process

1. Analyze the task requirements
2. Apply domain expertise
3. Verify output quality

Anti-Rationalization Table

Rationalization	Reality
"I will add monitoring later"	Without monitoring, you cannot detect failures. Add it from day one.
"One model is enough"	Different tasks need different models. Route intelligently.
"Premature optimization"	Infrastructure decisions are hard to change later. Design for scale early.