Properly support per-model maxTokens/temperature

pkg/cmd/util/util.go

@@ -19,15 +19,16 @@ import (
 // Prompt prompts the configured the configured model and streams the response
 // to stdout. Returns all model reply messages.
 func Prompt(ctx *lmcli.Context, messages []conversation.Message, callback func(conversation.Message)) (*api.Message, error) {
-	m, _, p, err := ctx.GetModelProvider(*ctx.Config.Defaults.Model, "")
+	modelConfig, err := ctx.GetModelProvider(*ctx.Config.Defaults.Model, "")
 	if err != nil {
 		return nil, err
 	}
+	p := modelConfig.Client
 
 	params := provider.RequestParameters{
-		Model:       m,
-		MaxTokens:   *ctx.Config.Defaults.MaxTokens,
-		Temperature: *ctx.Config.Defaults.Temperature,
+		Model:       modelConfig.Model,
+		MaxTokens:   modelConfig.MaxTokens,
+		Temperature: modelConfig.Temperature,
 	}
 
 	system := ctx.DefaultSystemPrompt()
@@ -206,15 +207,16 @@ Example response:
 		},
 	}
 
-	m, _, p, err := ctx.GetModelProvider(
+	modelConfig, err := ctx.GetModelProvider(
 		*ctx.Config.Conversations.TitleGenerationModel, "",
 	)
 	if err != nil {
 		return "", err
 	}
+	p := modelConfig.Client
 
 	requestParams := provider.RequestParameters{
-		Model:     m,
+		Model:     modelConfig.Model,
 		MaxTokens: 25,
 	}