Self-hosted Ollama with ssl

[kapsner]

I am trying to connect deepagents-cli to a self-hosted ollama instance that uses ssl / https.
My "~/.deepagents/config.toml" looks as follows:

[models.providers.ollama]
models = ["llama3.3", "gpt-oss", "mistral-large"]
api_key_env = "OLLAMA_API_KEY"
base_url = "https://123.45.67.89"

I am starting the cli with deepagents-cli --model ollama:llama3.3, however, not getting beyond the following error:

How can I suppress ssl-verifciation using deepagents-cli?

(In plain python, I am using verify=False parameter when calling requests.post(), which works perfectly using the above mentioned config)

Edit:
I found a solution and another issue:
One needs to install pip install certifi first and append the content of the self-signed certificate (download *.pem-file via browser) to the file located at python -c "import certifi; print(certifi.where()) to address the ssl-verification issue.

However, now I am facing the following issue: when specifying the base-url as base_url = "https://123.45.67.89" I get this error:

Looking into the POST-HTTP packages, I see that deepagents automatically tries to access api-endpoint "/api/chat".
But on our hosted ollama-instance, only "/api/chat/completions" endpoint is exposed.
Question: can I configure the endpoint-path somehow?

[kapsner]

Solution: to get the "/api/chat/completions" endpoint working with ollama-hosted models, one need to configure the config.toml as if it was an openai-model-provider:

[models.providers.openai]
models = ["llama3.3", "gpt-oss", "mistral-large"]
api_key_env = "OPENAI_API_KEY"
base_url = "https://123.45.67.89/api"

[Mason Daugherty (mdrxy)]

good to see you found a workaround kapsner -- can you let me know more about how your ollama instance is set up/hosted?

[kapsner]

Sure, actually I am querying openwebui (https://openwebui.com/), which uses ollama for model handling. Those are different pods in a kubernetes stack. However, API key and base-URL refer to our openwebui instance (I myself was using 'ollama' and 'openwebui' interchangeably, which was totally wrong and I just realized it by finding the solution to this issue) - thus, using the "openai" style config here for the local setup makes sense.

(Probably it could be helpful to add openwebui as a separate model provider to make it even more clear how to configure it)

[Mason Daugherty (mdrxy)]

got it, is openwebui openai compatible?

[kapsner]

yes: https://github.com/open-webui/open-webui?tab=readme-ov-file#open-webui-

Open WebUI is an extensible, feature-rich, and user-friendly self-hosted AI platform designed to operate entirely offline. It supports various LLM runners like Ollama and OpenAI-compatible APIs, with built-in inference engine for RAG, making it a powerful AI deployment solution.

Sorry for the messy discussion here. Initially, I struggled with the ssl-verification thingy and just found out in the process, that there is another issue, which, however, is completely unrelated to the ssl-verification