Customizes the model name based on options, creating a temporary model if needed.
When Ollama-only options must be baked into the model (num_ctx, top_k, min_p, typical_p, repeat_last_n, repeat_penalty), returns a temp name under ollmchat-temp/ and creates it via Call.Create if missing. Per-request v1 options (temperature, top_p, seed, stop, presence_penalty, frequency_penalty) are not baked here — ChatCompletions sends those on each request.
When for_ollama is true, returns this.name immediately — native Call.Chat sends options on each /api/chat request; baking would double- apply.
| connection |
The connection to use for API calls |
| options |
Runtime options; -1 / -1.0 means unset for numeric fields |
| for_ollama |
When true, skip temp-model creation (native chat path) |
|
The model name to use (base name or temp variant) |
| Error |
if model creation fails |