Dify + Nbility: From Workflow Builder to Published AI App

The previous articles covered Hermes, OpenClaw, and token cost. This one moves into practical application deployment: connect Dify to Nbility and turn a model API into a usable product workflow.

By the end, you should understand:

• what Dify is useful for;
• where Base URL, API Key, and model name belong;
• how to avoid unnecessary token consumption.

What Is Dify Good For?

Dify turns model capabilities into an application surface: knowledge bases, workflows, chat panels, automation, support bots, or group assistants.

A typical architecture is:

user / trigger -> Dify -> OpenAI-compatible API -> Nbility -> model -> result

The app manages UX and workflow. Nbility manages a unified model endpoint, API keys, and token usage.

Step 1: Prepare Nbility API Details

Open:

https://nbility.dev

Prepare:

Base URL: https://api.nbility.dev/v1
API Key:  [REDACTED]
Model:    a model available in your account

Never publish real keys in articles, screenshots, or logs.

Step 2: Deploy or Open Dify

If self-hosting, start from the official Docker or installation guide. The important idea is: if the app supports OpenAI-compatible or custom OpenAI endpoints, it can usually connect to Nbility.

Step 3: Configure the Model Provider

Look for model/provider settings and fill in:

Provider: OpenAI / OpenAI-compatible / Custom
Base URL: https://api.nbility.dev/v1
API Key: [REDACTED]
Model: your model name

If the UI has no Base URL field, check Docker environment variables, server config, admin settings, or “custom provider” options.

Step 4: Run the Smallest End-to-End Test

Start with one simple prompt:

Reply in one sentence and confirm the current model endpoint is working.

Check:

• response works;
• no 401 error;
• model name is valid;
• logs do not expose the key;
• token usage looks reasonable.

Step 5: Add a Real Workflow

Dify becomes valuable when it handles repeatable work:

• personal knowledge-base Q&A;
• business FAQ support;
• daily summaries and reports;
• group-chat Q&A;
• document summarization;
• automation notifications;
• API and data workflows.

Token Cost Controls

Once Dify becomes a daily tool, token usage becomes continuous. Recommended controls:

1. create a separate Nbility API Key for Dify;
2. use cost-effective models for daily tasks;
3. reserve stronger models for hard tasks;
4. control RAG chunking and retrieval count;
5. add triggers and cooldowns for group/automation flows;
6. avoid logging full prompts and keys;
7. review usage weekly.

Common Issues

1. 401 unauthorized

Usually a wrong API key, extra whitespace, or a web login token instead of an API key.

2. model not found

Use a model name available in your Nbility account. Do not blindly copy another platform’s model name.

3. Should Base URL include /v1?

Most OpenAI-compatible apps use:

https://api.nbility.dev/v1

Some apps append /v1 automatically, in which case use:

https://api.nbility.dev

If you see 404 errors, check this first.

4. Usage is higher than expected

Check knowledge base retrieval, long history, auto summaries, plugins, tool calls, and group-chat triggers.