Models deployed via localAI can be used in LLM functions, but they cannot be used in agents.

[lululu-1997]

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.

Content

Models deployed via localAI can be used in LLM functions, but they cannot be used in agents.

My models:

Use in LLM:

Can't not use in Agent:

[dosubot[bot]]

Hi @lululu-1997! I'm Dosu and I’m helping the dify team.

This is a known limitation in Dify: agent nodes only show models that explicitly support advanced agent strategies like function calling or ReACT. Most local models (including those deployed via LocalAI) do not appear in the agent model selection unless their provider configuration advertises these capabilities, even though they work in LLM nodes. This filtering is enforced in both the backend and UI, which is why you see a difference between LLM and agent model selection. There is no explicit code restriction against LocalAI; the limitation is due to feature filtering for agent compatibility (source, discussion).

If your local model supports function calling, double-check your plugin configuration and try refreshing the page after adding new models—sometimes they don't appear until you do so. If the model still doesn't show up, it likely doesn't support the required agent features. In that case, you can either use it in LLM nodes (which have broader model support) or switch to a model/provider known to work with agents (like Deepseek or GPT-4o) (source).

Full support for arbitrary local models in agent workflows is not yet implemented. Keep an eye on future Dify releases for expanded support.

To reply, just mention @dosu.

---

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