Support Ollama Registry

[kong62]

Can you support ollama models？

[stonezdj]

You can refer https://github.com/stonezdj/harbor/releases/tag/v2.11.0-ollama, left your comment if any feature need to add

[wy65701436]

Can you break down your requirement in details? If the ollama models are OCI compatible, then Harbor can support that.

[kong62]

Can you break down your requirement in details? If the ollama models are OCI compatible, then Harbor can support that.

when we deploy ollama models:
`
ollama pull mistral:7b
...

ollama list
NAME ID SIZE MODIFIED
mistral:7b f974a74358d6 4.1 GB 4 days ago
qwen2.5:32b 9f13ba1299af 19 GB 8 days ago
deepseek-r1:32b 38056bbcbb2d 19 GB 8 days ago
qwen2.5:14b 7cdf5a0187d5 9.0 GB 9 days ago
deepseek-r1:14b ea35dfe18182 9.0 GB 11 days ago
deepseek-r1:70b 0c1615a8ca32 42 GB 11 days ago
deepseek-r1:7b 0a8c26691023 4.7 GB 11 days ago
deepseek-r1:8b 28f8fd6cdc67 4.9 GB 11 days ago
llama3.1:8b 46e0c10c039e 4.9 GB 2 months ago
qwen2.5:7b 845dbda0ea48 4.7 GB 2 months ago
llama3.3:latest a6eb4748fd29 42 GB 2 months ago
`

These models are downloaded from the Internet, which is very slow. And we have many nodes that need to pull these models. So, we want to pull the model through a node, and then push it to the harbor, so that other nodes can point to harbor to pull. This is similar to using docker to pull images.

[bupd]

you can use ramalama to push the models to Harbor.

Currently using ollama push to push model to Harbor returns this error:

❯ ollama push demo.goharbor.io/harborbupd/olulama:latest
retrieving manifest 
Error: Get "harbor?nonce=bZ3Q5RVQd3azeO5mSPqEfA&scope=&service=&ts=1739954090": unsupported protocol scheme ""

But pushing this using ramalama seems to work fine.

[bupd]

Eventhough If we support Ollama in Harbor. Ollama doesn't currently support pull from Private Registry.