added parameter option for ollama

cmd/mcp.go

@@ -258,10 +258,18 @@ func handleHelpCommand() {
 		"- **Anthropic Claude**: `anthropic:claude-3-5-sonnet-latest`\n",
 	)
 	markdown.WriteString("- **Ollama Models**: `ollama:modelname`\n")
-	markdown.WriteString("\nExamples:\n")
+
+	markdown.WriteString("\n## Ollama Parameters\n\n")
+	markdown.WriteString("For Ollama models, you can control generation parameters:\n\n")
+	markdown.WriteString("- **--temperature**: Controls randomness (0.0-1.0, default 0.7)\n")
+	markdown.WriteString("- **--max-tokens**: Maximum tokens in response (default uses model setting)\n")
+	markdown.WriteString("- **--num-ctx**: Context window size (default uses model setting)\n")
+
+	markdown.WriteString("\n## Examples\n")
 	markdown.WriteString("```\n")
 	markdown.WriteString("mcphost -m anthropic:claude-3-5-sonnet-latest\n")
 	markdown.WriteString("mcphost -m ollama:qwen2.5:3b\n")
+	markdown.WriteString("mcphost -m ollama:llama3 --temperature 0.5 --max-tokens 2048 --num-ctx 4096\n")
 	markdown.WriteString("```\n")
 
 	rendered, err := renderer.Render(markdown.String())

cmd/root.go

@@ -34,6 +34,10 @@ var (
 	anthropicBaseURL string // Base URL for Anthropic API
 	openaiAPIKey     string
 	anthropicAPIKey  string
+	// Parameters for Ollama
+	temperature      float32 // Temperature for Ollama models
+	maxTokens        int     // Max tokens for Ollama models
+	numCtx           int     // Context size for Ollama models
 )
 
 const (
@@ -54,8 +58,14 @@ Available models can be specified using the --model flag:
 - OpenAI: openai:gpt-4
 - Ollama models: ollama:modelname
 
+For Ollama models, you can control generation parameters:
+- --temperature: Controls randomness (0.0-1.0, default 0.7)
+- --max-tokens: Maximum tokens in response (default uses model setting)
+- --num-ctx: Context window size (default uses model setting)
+
 Example:
   mcphost -m ollama:qwen2.5:3b
+  mcphost -m ollama:llama3 --temperature 0.5 --max-tokens 2048 --num-ctx 4096
   mcphost -m openai:gpt-4`,
 	RunE: func(cmd *cobra.Command, args []string) error {
 		return runMCPHost()
@@ -78,6 +88,13 @@ func init() {
 	rootCmd.PersistentFlags().
 		StringVarP(&modelFlag, "model", "m", "anthropic:claude-3-5-sonnet-latest",
 			"model to use (format: provider:model, e.g. anthropic:claude-3-5-sonnet-latest or ollama:qwen2.5:3b)")
+	// Flags for Ollama parameters
+	rootCmd.PersistentFlags().
+		Float32Var(&temperature, "temperature", 0.7, "temperature setting for Ollama models (0.0-1.0)")
+	rootCmd.PersistentFlags().
+		IntVar(&maxTokens, "max-tokens", 0, "maximum tokens in response for Ollama models (0 for default)")
+	rootCmd.PersistentFlags().
+		IntVar(&numCtx, "num-ctx", 0, "context window size for Ollama models (0 for default)")
 
 	// Add debug flag
 	rootCmd.PersistentFlags().
@@ -118,7 +135,8 @@ func createProvider(modelString string) (llm.Provider, error) {
 		return anthropic.NewProvider(apiKey, anthropicBaseURL, model), nil
 
 	case "ollama":
-		return ollama.NewProvider(model)
+		// Pass temperature, maxTokens, and numCtx to the Ollama provider
+		return ollama.NewProvider(model, temperature, maxTokens, numCtx)
 
 	case "openai":
 		apiKey := openaiAPIKey

pkg/llm/ollama/provider.go

@@ -18,19 +18,25 @@ func boolPtr(b bool) *bool {
 
 // Provider implements the Provider interface for Ollama
 type Provider struct {
-	client *api.Client
-	model  string
+	client      *api.Client
+	model       string
+	temperature float32
+	maxTokens   int
+	numCtx      int
 }
 
 // NewProvider creates a new Ollama provider
-func NewProvider(model string) (*Provider, error) {
+func NewProvider(model string, temperature float32, maxTokens int, numCtx int) (*Provider, error) {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
 		return nil, err
 	}
 	return &Provider{
-		client: client,
-		model:  model,
+		client:      client,
+		model:       model,
+		temperature: temperature,
+		maxTokens:   maxTokens,
+		numCtx:      numCtx,
 	}, nil
 }
 
@@ -43,7 +49,10 @@ func (p *Provider) CreateMessage(
 	log.Debug("creating message",
 		"prompt", prompt,
 		"num_messages", len(messages),
-		"num_tools", len(tools))
+		"num_tools", len(tools),
+		"temperature", p.temperature,
+		"max_tokens", p.maxTokens,
+		"num_ctx", p.numCtx)
 
 	// Convert generic messages to Ollama format
 	ollamaMessages := make([]api.Message, 0, len(messages)+1)
@@ -153,14 +162,36 @@ func (p *Provider) CreateMessage(
 
 	log.Debug("sending messages to Ollama", 
 		"messages", ollamaMessages,
-		"num_tools", len(tools))
+		"num_tools", len(tools),
+		"temperature", p.temperature,
+		"max_tokens", p.maxTokens,
+		"num_ctx", p.numCtx)
 
-	err := p.client.Chat(ctx, &api.ChatRequest{
+	// Create chat request with options for temperature, max_tokens, and num_ctx
+	chatRequest := &api.ChatRequest{
 		Model:    p.model,
 		Messages: ollamaMessages,
 		Tools:    ollamaTools,
 		Stream:   boolPtr(false),
-	}, func(r api.ChatResponse) error {
+		Options:  make(map[string]interface{}),
+	}
+
+	// Only set temperature if it's provided (non-zero)
+	if p.temperature > 0 {
+		chatRequest.Options["temperature"] = p.temperature
+	}
+
+	// Only set max tokens if it's provided (non-zero)
+	if p.maxTokens > 0 {
+		chatRequest.Options["num_predict"] = p.maxTokens
+	}
+
+	// Only set context window size if it's provided (non-zero)
+	if p.numCtx > 0 {
+		chatRequest.Options["num_ctx"] = p.numCtx
+	}
+
+	err := p.client.Chat(ctx, chatRequest, func(r api.ChatResponse) error {
 		if r.Done {
 			response = r.Message
 		}