diff --git a/cli-config.yaml.example b/cli-config.yaml.example index 4f4b323d6f..dfaeed15d8 100644 --- a/cli-config.yaml.example +++ b/cli-config.yaml.example @@ -113,8 +113,9 @@ model: # # ollama_num_ctx: Ollama only — the num_ctx sent on every chat request. # Hermes auto-detects the model's window and sends it (Ollama otherwise - # defaults to 2048); set this to cap VRAM use. context_length, if set, - # caps it further. + # defaults to 2048); context_length, if set, caps the detected value. + # An explicit ollama_num_ctx is sent as-is (never capped) — set it to + # pin VRAM use. # # ollama_num_ctx: 32768 #