I’m trying to connect Hugging Face models to n8n, but I keep getting the error: “No Inference Provider available for model x.” I’ve tested several models and receive the same issue each time.
My goal is to integrate a model trained specifically for medical image analysis (e.g., [MedGemma]). Is there a supported way to connect it to n8n?
I also tried using a medical VLM via Ollama, but I get an error of: unable to analyze medical images (only answers medical question, which should not be the case as the model supports both text and image inputs).
Hi @Rana_Benmahmoud The error message “No Inference Provider Available” is not a bug of MedGemma but a limitation. The n8n node’s Hugging Face inference service does not support many specialized models like MedGemma.
Two main way forward, supported by us.
It is recommended to pull and run MedGemma locally using Ollama. Then connect n8n to MedGemma using the OpenAPI-Compatible API credentials and set the endpoint to http://localhost:11434/v1. It goes directly against Hugging Face service.
Try checking if your model is available on any services such as Fal.ai, Replicate, and Novita. If that’s the case, you can use the HTTP Request node to link n8n to them. Larger and known models work with hugging face but some specially trained or different small models like this one gives errors, hope this helps.
To add some clarity, n8n doesn’t run models by name alone. Every model must be accessed through a supported inference provider (such as OpenAI-compatible APIs, Ollama, or external inference services).
When a model is not exposed by one of these providers, n8n cannot execute it directly, which is why the “No Inference Provider available” message appears. In these cases, the supported approach is to either use a provider that already serves the model or connect to a custom inference endpoint via the HTTP Request node, as outlined in the documentation.
So, for a normal Hugging Face account, it won’t be available unless you deploy it,
If you’ve already deployed it, you should see it in the list of available models on your side,
here, in my case it’s not available, so I get this error:
Hi @mohamed3nan Sorry if that felt unprofessional or incorrect, in my college project i have used mlgawd/navarasa-2b-2.0-cyberdost · Hugging Face this model and hosted this on AWS it works fine for me, like i understand your concern to be more professional here. I would consider this from now on, my bad. I will delete that for now.
For the context of the asked question: If you want you can self host that model on AWS which is a bit technical but would get the job done, i do not recommend that because your self hosted model would not be updated and will not get any future release, so for now i recommend using OpenAI’s model cause they are very largely trained and with right system prompt and AI prompt would really get the job done without any much overhead of updating in future. Hope this helps.
@tamy.santos@Anshul_Namdev@mohamed3nan Thank you everyone for your suggestions! I have managed to connect MedGemma through replicate (via HTTP Request Node).
Now, I want to connect it to an AI agent to manage tool calls. I’ve tried various methods without success.
I’m considering using MedGemma with OLLAMA locally. Will this setup work, and can I avoid issues like “no inference provider available”? As using it locally requires more RAM to my laptop (so extra cost).
@Rana_Benmahmoud First, that model is not actually capable for tool calling, and this is really not a good practice that you are running models locally, this process is really slow and time consuming for things you can get for free, just use Google’s AI Studio to grab a pair of Free API keys for all Gemini models:
Understood you want a very specific model and FREE, so consider going with Gemini as it is capable of tool calling at certain level and will fulfill most of your use cases.
Sorry if that felt unprofessional or incorrect, in my college project i have used