undefined

undefined

undefined

undefined

helm repo add dify https://langgenius.github.io/dify-helm
helm install dify dify/dify

undefined

**Azure/Google Cloud/Alibaba Cloud:**
See deployment guides in official documentation.

undefined

undefined

undefined

**Application Types:**
1. **Chatbot** - Conversational AI with memory
2. **Text Generator** - Single-turn text generation
3. **Agent** - Task-oriented AI with tools
4. **Workflow** - Custom multi-step processes

You are a helpful customer service assistant for TechCorp.
You can help users with:
- Product information
- Order tracking
- Technical support
- Account management

Be friendly, professional, and concise.

import requests

API_KEY = "your_dify_api_key"
API_URL = "http://localhost/v1"

**Node.js Example:**
```javascript
const fetch = require('node-fetch');

const API_KEY = 'your_dify_api_key';
const API_URL = 'http://localhost/v1';

async function sendMessage(query) {
  const response = await fetch(`${API_URL}/chat-messages`, {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${API_KEY}`,
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      inputs: {},
      query: query,
      response_mode: 'blocking',
      user: 'user-123'
    })
  });

  return await response.json();
}

sendMessage('Hello!').then(console.log);

Workflow Steps:
1. Start → User Message
2. Knowledge Retrieval → Search documentation
3. Conditional Branch:
   - If relevant docs found → Use context
   - If not found → Use general knowledge
4. LLM Node → Generate response
5. Tool Node → Create ticket (if needed)
6. End → Return response

def main(data: dict) -> dict:
    # Access workflow variables
    user_query = data.get('query', '')

    # Custom processing
    processed = user_query.upper()

    # Return results
    return {
        'result': processed,
        'length': len(user_query)
    }

function main(data) {
  const query = data.query || '';

  // Process data
  const words = query.split(' ');

  return {
    word_count: words.length,
    first_word: words[0]
  };
}

Agent Configuration:
  Model: gpt-4
  Reasoning Mode: ReAct
  Tools:
    - Google Search
    - Wikipedia
    - Web Scraper
  System Prompt: |
    You are a research assistant that helps users find
    accurate information from reliable sources.

    Always:
    1. Search for current information
    2. Cite your sources
    3. Verify facts from multiple sources

undefined

undefined

Strategy: fixed_size
Chunk Size: 500 tokens
Overlap: 50 tokens
Use Case: General documents, articles

Strategy: paragraph
Min Size: 100 tokens
Max Size: 800 tokens
Use Case: Well-formatted documents

Strategy: semantic
Model: text-embedding-ada-002
Similarity Threshold: 0.8
Use Case: Complex technical documents

Type: vector
Top K: 5
Score Threshold: 0.7
Embedding Model: text-embedding-3-large

Type: hybrid
Vector Weight: 0.7
Keyword Weight: 0.3
Reranking: enabled
Reranking Model: cross-encoder/ms-marco-MiniLM-L-12-v2

POST /v1/chat-messages
Content-Type: application/json
Authorization: Bearer {api_key}

{
  "inputs": {},
  "query": "Your question here",
  "response_mode": "streaming",
  "user": "user-identifier"
}

POST /v1/completion-messages
Content-Type: application/json
Authorization: Bearer {api_key}

{
  "inputs": {
    "name": "John",
    "topic": "AI"
  },
  "response_mode": "blocking",
  "user": "user-123"
}

POST /v1/messages/{message_id}/feedbacks
Content-Type: application/json
Authorization: Bearer {api_key}

{
  "rating": "like",
  "user": "user-123"
}

from dify_client import DifyClient

client = DifyClient(api_key="your_api_key")

**Streaming Response:**
```python
for chunk in client.chat_stream(
    query="Explain quantum computing",
    user="user-123"
):
    print(chunk.delta, end="", flush=True)

undefined

undefined

undefined

undefined

undefined

undefined

undefined

def invoke(self, model, credentials, prompt, **kwargs):
    # Custom API call logic
    response = your_api_call(prompt)
    return response

undefined

Workflow:
  - Node1: LLM Call
  - Parallel:
      - Node2a: Knowledge Retrieval
      - Node2b: External API Call
  - Node3: Combine Results

undefined

undefined

Services:
  API:
    replicas: 3
    load_balancer: nginx
  Worker:
    replicas: 5
    queue: redis
  Database:
    primary: postgres-main
    replicas: 2
    backup: daily

Monitoring:
  - Prometheus: Metrics collection
  - Grafana: Visualization
  - Loki: Log aggregation
  - Alertmanager: Alerts

undefined

**Security Issues:**
Email: security@dify.ai