How to build a customer chatbot with Dify

Dify lets you build an assistant that knows your company's products and policies, starting from your own documents. Here is the full path.

1. Start Dify

Install Dify with one click, then open /install and create the admin account.

2. Connect a model

In Settings, under Model Provider, you add an API key for a model (OpenAI, Anthropic) or connect a local model through an OpenAI-compatible API, see OpenAI-compatible local API. The local option means fixed cost and data that never leaves the server.

3. Upload the documents (Knowledge)

In the Knowledge section, you upload the files: manuals, FAQs, product descriptions. Dify splits and indexes them automatically, which is exactly the RAG process, see how RAG works.

4. Create the application

You pick the Chatbot type, link the knowledge base from step 3 and write the system instructions: what tone to use, what it must not say, what it does when it does not know the answer.

5. Test and publish

Dify has a test mode where you see which sources it used for each answer, useful for catching wrong replies. When you are happy, you publish the app as a web page or embed it in your site through the API.

6. Put it in production properly

• Add a domain and HTTPS through a reverse proxy.
• Back up the Docker volumes, see container backups.
• Watch resource usage, see resource monitoring.