Ollama configuration in .env: correct base URL and model name (/api, /v1, llama3 vs llama3:latest)

[sebastien-libbrecht]

Hello,

I am trying to configure Ollama as the LLM backend using environment variables, but I am having trouble understanding the correct values to put in the .env file.

In particular, I am confused about:

Which base URL should be used

• With or without /api?
• With or without /v1?

Which model name is expected

• llama3 vs llama3:latest

How to avoid common errors such as:

• 400 Bad Request
• 404 Not Found

What I have verified so far:

• Ollama is running correctly
• The model is present (checked with ollama list)
• The OpenAI-compatible endpoint responds at /v1/models

Despite this, depending on how the .env variables are configured, the application still returns Bad Request or Not Found errors when performing chat completions.

My questions:

• What are the recommended environment variables to configure Ollama correctly?
• Should the base URL include /api, /v1, or only the root URL?
• Is llama3 a valid model name, or should llama3:latest always be used?
• Is there an official example of a working .env configuration?

The documentation for Ollama is not very clear on these points, and a concise example would really help.

Thank you in advance for your help,
and happy end-of-year holidays.

[elie222]

The setup CLI / example .env doesn't help here? I haven't used ollama a lot, you'd need to research a bit online. This may help:
https://ai-sdk.dev/providers/community-providers/ollama

We use AI SDK.

[elie222]

If you get it working I'd love to update the docs/example/cli with better values so others aren't confused about it in the future.