The pool cli assumes that the model has a 125k context window, but that may not be true (especially when using a locally hosted Ollama model). It would be nice if:
- we could specify a context size, and have
pool tell Ollama to use that size
- failing that, it would be nice if
pool could auto-detect the context size used by the Ollama model
- failing that, it would be nice if we could specify the context size that
pool should use and have it use that context size when reporting usage and auto-compacting the context when it overflows.
The
poolcli assumes that the model has a 125k context window, but that may not be true (especially when using a locally hosted Ollama model). It would be nice if:pooltell Ollama to use that sizepoolcould auto-detect the context size used by the Ollama modelpoolshould use and have it use that context size when reporting usage and auto-compacting the context when it overflows.