Offline deployment configuration model error

[dayulxl]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

In a networked environment, the full image of Dify was downloaded using Docker-compose, but the packaged image was not submitted.
Then, in a CentOS 64 system environment without internet access, after deploying using an uncommitted Docker full image, when entering the page to configure the model, the model plugin kept reporting errors that it could not connect to the internet.
Is it because the plugin wasn't packaged with the image, or can we set up an offline mode?

2. Additional context or comments

[dosubot[bot]]

Hi @dayulxl! I'm Dosu and I’m helping the dify team.

The error "PluginDaemonInternalServerError: no available node, plugin runtime not found" in your offline environment is caused by the plugin runtime needing to download Python dependencies, which fails without internet access.

Key issue: Even with Docker images transferred, plugins still need Python packages at runtime.

Solutions:

1. Set up an internal PyPI mirror (recommended) and configure:

   PIP_MIRROR_URL=http://your-internal-pypi-mirror/simple

2. Add these environment variables to your .env:

   MARKETPLACE_ENABLED=false
   FORCE_VERIFYING_SIGNATURE=false
   PYTHON_ENV_INIT_TIMEOUT=640
   PLUGIN_MAX_EXECUTION_TIMEOUT=2400

3. Pre-package model plugins in an online environment using dify plugin package ./plugin_project, then transfer the .difypkg files and install via Plugins → Install Plugin → Via Local File in the UI.

4. Ensure you have the langgenius/dify-plugin-daemon image loaded - this is required for model plugins to work.

For local models like Ollama, you'd also need to pre-download the model files and transfer them to your offline environment.

See this related discussion on air-gapped deployment and this thread on handling Python packages offline for more details.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[aniruddhaadak80]

From my point of view, this only becomes a real offline deployment once plugin dependency resolution is removed from the runtime path entirely.

Moving images into an air gapped environment is the easy part. The harder part is making sure every Python dependency, plugin artifact, and verification step is already materialized before first boot. If the plugin daemon still expects to resolve packages dynamically, the deployment is not really offline in an operational sense.

I would be interested to know whether anyone here has built a reproducible flow around a local wheelhouse or prebuilt plugin runtime image. That seems like the cleanest long term answer for CentOS style environments.