Michael Yang

@BowenKwan your issue appears different so I'll respond in #188

Setting seed alone won't achieve reproducible outputs. The [fill-in-middle](https://github.com/ollama/ollama-python/blob/main/examples/fill-in-middle/main.py#L16) example sets `temperature=0` which is sufficient for generating reproducible outputs.

this is now available https://ollama.com/library/phi3:mini-128k

Additionally `dd20bb891979d25aebc8bec07b2b3bbc.r2.cloudflarestorage.com` should be allowed as well. In summary - `.ollama.com` - `registry.ollama.ai` - `dd20bb891979d25aebc8bec07b2b3bbc.r2.cloudflarestorage.com`

reordered this series of prs so push is updated before pull

This is tricky. discollama uses the mention to identify when it's supposed to respond. Without a check like this, it'll try to respond to every single message. For example, if...