llama-stack
llama-stack copied to clipboard
feat(responses): implement full multi-turn support
I think the implementation needs more simplification. Spent way too much time trying to get the tests pass with models not co-operating :( Finally had to switch claude-sonnet to get things to pass reliably.