fix(tools): correct media provider params and UI fixes

- MiniMax audio: fix invalid params (sample_rate/bitrate as int, auto
  instrumental when no lyrics, pass duration_seconds)
- MiniMax/DashScope image: map aspect_ratio to provider size format
- Gemini video: read person_generation from chain params instead of hardcode
- GetParamInt: add string-to-int coercion for UI select values
- UI: fix combobox portal selection bug, dropdown overflow clipping,
  provider chain form spacing, skill dialog min-height
- Update bitrate options to numeric bps values in params schema

internal/tools/create_audio_minimax.go

@@ -19,8 +19,15 @@ func callMinimaxMusicGen(ctx context.Context, apiKey, apiBase, model, prompt str
 	lyrics := GetParamString(params, "lyrics", "")
 	instrumental := GetParamBool(params, "instrumental", false)
 	lyricsOptimizer := GetParamBool(params, "lyrics_optimizer", false)
-	sampleRate := GetParamString(params, "sample_rate", "44100")
-	bitrate := GetParamString(params, "bitrate", "320k")
+	sampleRate := GetParamInt(params, "sample_rate", 44100)
+	bitrate := GetParamInt(params, "bitrate", 256000)
+
+	// lyrics is required when is_instrumental is false
+	if !instrumental && lyrics == "" {
+		instrumental = true
+	}
+
+	duration := GetParamInt(params, "duration", 0)
 
 	body := map[string]interface{}{
 		"model":            model,
@@ -37,6 +44,9 @@ func callMinimaxMusicGen(ctx context.Context, apiKey, apiBase, model, prompt str
 	if lyrics != "" {
 		body["lyrics"] = lyrics
 	}
+	if duration > 0 {
+		body["duration_seconds"] = duration
+	}
 
 	jsonBody, err := json.Marshal(body)
 	if err != nil {

internal/tools/create_image_dashscope.go

@@ -59,8 +59,25 @@ func dashScopeTaskEndpoint(apiBase, taskID string) string {
 // callDashScopeImageGen calls the DashScope (Alibaba/Bailian) multimodal image generation API.
 // The API is async: an initial POST returns a task_id, which is then polled until done.
 // On completion, output.results[].url contains the image URL to download.
+// aspectRatioToDashScopeSize converts aspect_ratio to DashScope size format.
+// Falls back to explicit "size" param if set, otherwise uses aspect_ratio mapping.
+func aspectRatioToDashScopeSize(params map[string]any) string {
+	if s := GetParamString(params, "size", ""); s != "" {
+		return s
+	}
+	ar := GetParamString(params, "aspect_ratio", "1:1")
+	switch ar {
+	case "16:9":
+		return "1280*720"
+	case "9:16":
+		return "720*1280"
+	default:
+		return "1024*1024"
+	}
+}
+
 func callDashScopeImageGen(ctx context.Context, apiKey, apiBase, model, prompt string, params map[string]any) ([]byte, *providers.Usage, error) {
-	size := GetParamString(params, "size", "1024*1024")
+	size := aspectRatioToDashScopeSize(params)
 	promptExtend := GetParamBool(params, "prompt_extend", true)
 
 	endpoint := dashScopeImageEndpoint(apiBase)

internal/tools/create_image_minimax.go

@@ -16,8 +16,30 @@ import (
 // callMinimaxImageGen calls the MiniMax image generation API.
 // Endpoint: POST {apiBase}/image_generation
 // Response: base64 image data in data.image_list[0].base64_image
+// aspectRatioToMinimaxSize converts aspect_ratio (e.g. "16:9") to MiniMax size format.
+// Falls back to explicit "size" param if set, otherwise uses aspect_ratio mapping.
+func aspectRatioToMinimaxSize(params map[string]any) string {
+	// Explicit size param takes priority (from UI provider settings)
+	if s := GetParamString(params, "size", ""); s != "" {
+		return s
+	}
+	ar := GetParamString(params, "aspect_ratio", "1:1")
+	switch ar {
+	case "16:9":
+		return "1280*720"
+	case "9:16":
+		return "720*1280"
+	case "4:3":
+		return "1024*768"
+	case "3:4":
+		return "768*1024"
+	default:
+		return "1024*1024"
+	}
+}
+
 func callMinimaxImageGen(ctx context.Context, apiKey, apiBase, model, prompt string, params map[string]any) ([]byte, *providers.Usage, error) {
-	size := GetParamString(params, "size", "1024*1024")
+	size := aspectRatioToMinimaxSize(params)
 	promptOptimizer := GetParamBool(params, "prompt_optimizer", true)
 
 	body := map[string]interface{}{

internal/tools/create_video.go

@@ -149,7 +149,7 @@ func (t *CreateVideoTool) callProvider(ctx context.Context, cp credentialProvide
 
 	switch ProviderTypeFromName(providerName) {
 	case "gemini":
-		return t.callGeminiVideoGen(ctx, cp.APIKey(), cp.APIBase(), model, prompt, duration, aspectRatio)
+		return t.callGeminiVideoGen(ctx, cp.APIKey(), cp.APIBase(), model, prompt, duration, aspectRatio, params)
 	case "minimax":
 		return callMinimaxVideoGen(ctx, cp.APIKey(), cp.APIBase(), model, params)
 	default:
@@ -159,7 +159,7 @@ func (t *CreateVideoTool) callProvider(ctx context.Context, cp credentialProvide
 
 // callGeminiVideoGen uses the Gemini predictLongRunning API for Veo video generation.
 // Flow: POST predictLongRunning → poll operation → download video from URI.
-func (t *CreateVideoTool) callGeminiVideoGen(ctx context.Context, apiKey, apiBase, model, prompt string, duration int, aspectRatio string) ([]byte, *providers.Usage, error) {
+func (t *CreateVideoTool) callGeminiVideoGen(ctx context.Context, apiKey, apiBase, model, prompt string, duration int, aspectRatio string, params map[string]any) ([]byte, *providers.Usage, error) {
 	nativeBase := strings.TrimRight(apiBase, "/")
 	nativeBase = strings.TrimSuffix(nativeBase, "/openai")
 
@@ -173,7 +173,7 @@ func (t *CreateVideoTool) callGeminiVideoGen(ctx context.Context, apiKey, apiBas
 		"parameters": map[string]interface{}{
 			"aspectRatio":      aspectRatio,
 			"durationSeconds":  duration,
-			"personGeneration": "allow_all",
+			"personGeneration": GetParamString(params, "person_generation", "allow_all"),
 		},
 	}
 

internal/tools/media_provider_chain.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"io"
 	"log/slog"
+	"strconv"
 	"strings"
 	"time"
 
@@ -297,6 +298,10 @@ func GetParamInt(params map[string]any, key string, fallback int) int {
 		return int(v)
 	case int:
 		return v
+	case string:
+		if n, err := strconv.Atoi(v); err == nil {
+			return n
+		}
 	}
 	return fallback
 }
@@ -319,6 +324,8 @@ func ProviderTypeFromName(name string) string {
 		return "anthropic"
 	case name == "suno" || strings.HasPrefix(name, "suno"):
 		return "suno"
+	case name == "yescale":
+		return "openai"
 	default:
 		return "openai_compat"
 	}

ui/web/src/components/ui/combobox.tsx

@@ -29,6 +29,7 @@ export function Combobox({
   const [open, setOpen] = React.useState(false);
   const [search, setSearch] = React.useState("");
   const containerRef = React.useRef<HTMLDivElement>(null);
+  const dropdownRef = React.useRef<HTMLDivElement>(null);
   const [dropdownStyle, setDropdownStyle] = React.useState<React.CSSProperties>({});
 
   // Sync search text when value changes externally — show label if available
@@ -41,7 +42,11 @@ export function Combobox({
   React.useEffect(() => {
     if (!open) return;
     const handler = (e: MouseEvent) => {
-      if (containerRef.current && !containerRef.current.contains(e.target as Node)) {
+      const target = e.target as Node;
+      if (
+        containerRef.current && !containerRef.current.contains(target) &&
+        (!dropdownRef.current || !dropdownRef.current.contains(target))
+      ) {
         setOpen(false);
       }
     };
@@ -54,11 +59,14 @@ export function Combobox({
     if (!open || !portalContainer?.current || !containerRef.current) return;
     const inputRect = containerRef.current.getBoundingClientRect();
     const portalRect = portalContainer.current.getBoundingClientRect();
+    const left = inputRect.left - portalRect.left;
+    const maxWidth = portalRect.width - left;
     setDropdownStyle({
       position: "absolute",
       top: inputRect.bottom - portalRect.top + 4,
-      left: inputRect.left - portalRect.left,
+      left,
       width: inputRect.width,
+      maxWidth,
       zIndex: 50,
     });
   }, [open, search, portalContainer]);
@@ -89,10 +97,11 @@ export function Combobox({
 
   const dropdownContent = open && filtered.length > 0 && (
     <div
+      ref={dropdownRef}
       style={portalContainer ? dropdownStyle : undefined}
       className={cn(
-        "bg-popover text-popover-foreground max-h-60 min-w-full overflow-y-auto rounded-md border p-1 shadow-md",
-        !portalContainer && "absolute top-full left-0 z-50 mt-1",
+        "bg-popover text-popover-foreground max-h-60 overflow-y-auto rounded-md border p-1 shadow-md",
+        portalContainer ? "overflow-hidden" : "absolute top-full left-0 z-50 mt-1 min-w-full",
       )}
     >
       {filtered.map((o) => (

ui/web/src/pages/builtin-tools/media-provider-chain-form.tsx

@@ -425,7 +425,7 @@ export function MediaProviderChainForm({
         </DialogDescription>
       </DialogHeader>
 
-      <div className="space-y-2 max-h-[60vh] overflow-y-auto pr-1">
+      <div className="space-y-2 max-h-[60vh] overflow-y-auto pr-1 my-4">
         <DndContext sensors={sensors} collisionDetection={closestCenter} onDragEnd={handleDragEnd}>
           <SortableContext items={entries.map((e) => e.id)} strategy={verticalListSortingStrategy}>
             {entries.map((entry, index) => (

ui/web/src/pages/builtin-tools/media-provider-params-schema.ts

@@ -101,11 +101,12 @@ export const MEDIA_PARAMS_SCHEMA: Record<string, Record<string, ParamField[]>> =
         key: "bitrate",
         label: "Bitrate",
         type: "select",
-        default: "320k",
+        default: "256000",
         options: [
-          { value: "128k", label: "128k" },
-          { value: "256k", label: "256k" },
-          { value: "320k", label: "320k" },
+          { value: "32000", label: "32kbps" },
+          { value: "64000", label: "64kbps" },
+          { value: "128000", label: "128kbps" },
+          { value: "256000", label: "256kbps" },
         ],
       },
     ],

ui/web/src/pages/skills/skill-detail-dialog.tsx

@@ -100,7 +100,7 @@ export function SkillDetailDialog({
 
   return (
     <Dialog open onOpenChange={() => onClose()}>
-      <DialogContent className="max-h-[85vh] overflow-hidden flex flex-col sm:max-w-2xl md:max-w-4xl lg:max-w-5xl xl:max-w-6xl 2xl:max-w-7xl">
+      <DialogContent className="max-h-[85vh] md:min-h-[60vh] overflow-hidden flex flex-col sm:max-w-2xl md:max-w-4xl lg:max-w-5xl xl:max-w-6xl 2xl:max-w-7xl">
         <DialogHeader>
           <DialogTitle className="flex items-center gap-2 flex-wrap">
             {skill.name}