How to Create LLM Apps and Agent Workflows with Dify

What you need

A Dify instance (cloud.dify.ai or self-hosted via Docker).

API keys for one or more LLM providers.

Documents or data sources for the knowledge base.

Step 1: Deploy or sign up

Self-host: `docker compose up -d` from the Dify repository. Access at `http://localhost:3000`.

Or use the cloud version at cloud.dify.ai for immediate access.

Create a workspace and configure your model providers in Settings.

Step 2: Create a knowledge base

Upload documents (PDF, DOCX, TXT, web pages) that the AI will use to answer questions.

Dify automatically chunks, embeds, and indexes the content.

Configure the retrieval settings: top-k results, similarity threshold, and reranking.

Step 3: Build a RAG pipeline

Create a new 'Chatbot' or 'Agent' app type.

Connect the knowledge base as the primary context source.

Configure the prompt with retrieval instructions: 'Answer based on the knowledge base. If the answer is not in the knowledge base, say you do not know.'

Step 4: Add agent capabilities

Enable the agent mode to give the app tool-using capabilities.

Add tools: web search, code interpreter, image generation, or custom API tools.

Configure which tools are available and any usage limits.

Step 5: Publish the app

Dify generates a shareable chat URL, embeddable iframe, and REST API endpoint.

Configure authentication: public, API key, or OAuth.

Monitor usage in the Dify dashboard — logs show every query, retrieval, and agent action.

FAQ

Create Dify LLM Apps

1. Deploy Dify or sign up for cloud
2. Upload documents to a knowledge base
3. Build a RAG pipeline with retrieval config
4. Enable agent mode with tool access
5. Configure multi-model support
6. Publish as chat, API, or embed