Integrations

Dify

Install the Qubrid AI plugin for Dify and use OpenAI-compatible serverless models in apps, agents, and workflows

Dify is an open-source platform for building LLM apps, agents, and workflows. The Qubrid AI Dify plugin registers Qubrid as a model provider so you can use Qubrid serverless models anywhere Dify supports LLMs.

OpenAI-compatible serverless models from the Qubrid AI Platform are available through the model picker, or you can add any Qubrid model ID as a custom model.

Prerequisites:

  • A running Dify instance (self-hosted or cloud) with plugin support
  • Admin access to install plugins and configure model providers
  • A Qubrid account on the Qubrid AI Platform
  • A Qubrid API key (how to generate an API key)
  • Outbound HTTPS access from your Dify instance to https://platform.qubrid.com

Install the plugin

Install from the Dify Marketplace or upload the package locally.

Qubrid AI on Dify Marketplace

Install the official Qubrid AI plugin directly from the Dify Marketplace.

qubridai-qubridai_0.1.0.difypkg

Download

Download the plugin package above for local installation, then follow the setup steps below.

Setup

Install the plugin in Dify

In Dify, go to Plugins → Install plugin → Install from local package.

Upload qubridai-qubridai_0.1.0.difypkg (the file you downloaded above), or install from the Dify Marketplace once published.

Open the Qubrid AI provider settings

Go to Settings → Model Provider → Qubrid AI → Set up.

Enter your Qubrid API key

Enter your API key from platform.qubrid.com/api-keys.

Your API key is shown only once for security. If you lose it, generate a new key from Manage API Keys.

Configure the API base URL (optional)

Leave API Base URL empty to use the default (https://platform.qubrid.com/v1), or set it explicitly if you use a custom endpoint.

Save and select a model

Click Save, then select a Qubrid model in any Dify app or workflow.

Usage

The plugin registers Qubrid AI as a model provider. Use it anywhere Dify supports LLMs:

  • Chat apps: conversational interfaces backed by Qubrid models
  • Agents: autonomous agents with tool use and reasoning
  • Workflows: multi-step pipelines with branching, code nodes, and LLM steps

Predefined models appear in the model picker. You can also add any other Qubrid model ID as a custom model.

Browse the full catalog and copy exact model strings from Serverless Models. Model IDs are case-sensitive.

Required credentials

CredentialRequiredDescription
API KeyYesQubrid API key from platform.qubrid.com/api-keys
API Base URLNoDefaults to https://platform.qubrid.com/v1

Connection requirements

Your Dify instance must be able to reach:

  • https://platform.qubrid.com: model inference and catalog API

Outbound HTTPS (port 443) is required. No inbound ports or webhooks are needed.

Included models

The plugin ships 43 predefined chat models synced from the Qubrid catalog (/api/v1/models), including:

  • openai/gpt-oss-120b
  • Qwen/Qwen3.7-Max
  • meta-llama/Llama-3.3-70B-Instruct
  • deepseek-ai/DeepSeek-V3.2

You can also add any other Qubrid model ID as a custom model. See Serverless Models for the full list.

Monitor usage on Qubrid

Keep an eye on your usage and costs on the Qubrid platform.

  • Review Inference Logs to see request history and debug issues.
  • Check your credit balance before long-running or high-volume sessions.
  • Rotate API keys if you suspect a key was exposed.

Configuration reference

SettingValue
API KeyYour Qubrid API key
API Base URLhttps://platform.qubrid.com/v1 (default if left empty)
Example modelsopenai/gpt-oss-120b, Qwen/Qwen3.7-Max
Catalog endpointhttps://platform.qubrid.com/api/v1/models

Frequently Asked Questions (FAQ)

Developer guide

These steps are for maintaining the plugin source, not for end-user installation.

Source repository: github.com/QubridAI-Inc/Dify-Plugin

Project layout

PathPurpose
manifest.yamlPlugin metadata
provider/Model provider definition
models/llm/Predefined model YAMLs and LLM implementation
catalog_utils.pyCatalog → Dify model mapping
scripts/Build, sync, and E2E test scripts
qubridai.difypkgPackaged plugin for local install

Refresh models from catalog

cd qubridai
source .venv/bin/activate
python scripts/sync_models_from_catalog.py              # local catalog file
python scripts/sync_models_from_catalog.py --source api # live API
bash scripts/build_plugin.sh                            # sync + package

E2E tests (local Dify)

Requires local Dify at http://localhost and API keys configured for testing.

cd qubridai
source .venv/bin/activate
 
# Smoke test: install plugin + single LLM workflow
python scripts/e2e_local_dify_test.py
 
# Complex workflows + agent-chat (branching, code→LLM, chains, tools, ReAct agents)
python scripts/e2e_complex_workflows_test.py

Rebuild the package

cd qubridai
bash scripts/build_plugin.sh

Go to Dify Docs

Official Dify documentation for plugins, model providers, and workflows.

PreviousOpen WebUINextClaude Code