fix(ollama): improve tool-calling support and restore gemma/llama context limits

- Explicitly set tool_choice: auto when tools are present to aid gemma/llama models
- Sync stop sequences into the options map for broader compatibility
- Restore gemma/llama to the high-context (32k) optimization list

internal/providers/ollama.go

@@ -121,13 +121,14 @@ func BuildOllamaBody(request *models.UnifiedRequest, messagesJSON []interface{},
 
 	// Context window size (default 8k for all, 32k+ for modern large-context models)
 	ctxSize := 8192
-	if strings.Contains(modelLower, "llama3") || 
+	if strings.Contains(modelLower, "llama") || 
+	   strings.Contains(modelLower, "gemma") || 
 	   strings.Contains(modelLower, "mistral") || 
 	   strings.Contains(modelLower, "mixtral") ||
 	   strings.Contains(modelLower, "qwen") || 
 	   strings.Contains(modelLower, "deepseek") ||
 	   strings.Contains(modelLower, "command-r") ||
-	   strings.Contains(modelLower, "phi3") {
+	   strings.Contains(modelLower, "phi") {
 		ctxSize = 32768
 	}
 	options["num_ctx"] = ctxSize
@@ -157,15 +158,21 @@ func BuildOllamaBody(request *models.UnifiedRequest, messagesJSON []interface{},
 		options["top_k"] = *request.TopK
 	}
 
+	if len(request.Stop) > 0 {
+		body["stop"] = request.Stop
+		options["stop"] = request.Stop
+	}
+
 	if len(options) > 0 {
 		body["options"] = options
 	}
 
-	if len(request.Stop) > 0 {
-		body["stop"] = request.Stop
-	}
 	if len(request.Tools) > 0 {
 		body["tools"] = request.Tools
+		// Explicitly set tool_choice to auto if tools are present but choice is not specified
+		if request.ToolChoice == nil {
+			body["tool_choice"] = "auto"
+		}
 	}
 	if request.ToolChoice != nil {
 		var toolChoice interface{}