Add VLM-based diagram validation (#602)

* [Feature] Add VLM-based diagram validation

Add automatic VLM (Vision Language Model) validation after display_diagram
tool execution. The system captures a screenshot of the rendered diagram,
sends it to a VLM for visual analysis, and uses feedback to improve
diagram quality through the existing retry mechanism.

Changes:
- Add /api/validate-diagram endpoint for VLM validation
- Add diagram-validator.ts for client-side validation orchestration
- Add validation-prompts.ts for VLM system prompts
- Add ValidationCard component to display validation status in chat
- Add PNG capture functionality to diagram context
- Integrate validation into tool handlers with retry support (max 3)
- Add "Improve with Suggestions" button for manual regeneration
- Add settings toggle to enable/disable VLM validation
- Add getValidationModel() helper in ai-providers.ts

* refactor(validation): use AI SDK structured outputs and address review feedback

- Replace generateText + manual JSON parsing with generateObject and Zod schema
  for type-safe structured validation output
- Use AbortSignal.timeout() instead of Promise.race for cleaner timeout handling
- Add timeout validation with minimum 1000ms to handle malformed env values
- Remove unused xml parameter from validateRenderedDiagram API
- Remove parseValidationResponse function (now handled by schema)
- Clear validationStates on session switch and new chat to prevent memory leak
- Update 100ms render delay comment to clarify best-effort heuristic
- Remove unused useEffect import from ValidationCard
- Fix optional chaining lint warning in ValidationCard
- Add unit tests for formatValidationFeedback function

* refactor(validation): use AI SDK experimental_useObject hook instead of raw fetch

- Change API endpoint from generateObject to streamObject for useObject compatibility
- Create useValidateDiagram hook using AI SDK's experimental_useObject for reactive validation
- Update useDiagramToolHandlers to accept validation function as parameter
- Update chat-panel to use new useValidateDiagram hook
- Remove validateRenderedDiagram function from lib/diagram-validator.ts (now in hook)
- Export ValidationResultSchema from API route for client-side use

* fix(validation): extract schema to shared file for client/server compatibility

Move ValidationResultSchema to lib/validation-schema.ts to avoid importing
server-side modules (ai-providers) into client-side code. This fixes the
Turbopack build error caused by the hook importing from the API route.

* fix(validation): use 'Valid' instead of 'Complete' for validation success

Change ValidationCard success label from 'Complete' to 'Valid' to avoid
conflicting with ToolCallCard's 'Complete' badge in E2E tests. This fixes
the diagram-generation E2E test that expects a specific count of 'Complete'
badges.

* fix(validation): add aria-hidden to icons to prevent duplicate ID warning

* fix: improve VLM validation with bug fixes and i18n

- Fix race condition in pendingValidationRef (reject previous pending validation)
- Fix response format consistency (use streaming for all responses)
- Remove dead code (unused lastRequestRef and ValidationRequest interface)
- Consolidate duplicate types (re-export from validation-schema.ts)
- Add 'success_with_warnings' status for valid diagrams with warnings
- Fix tool card auto-collapse (only collapse once, respect user toggle)
- Set VLM validation default to disabled
- Add i18n support for diagram validation settings (en/zh/ja)
- Mark feature as experimental in settings UI

* fix: resolve TypeScript errors in electron-standalone

- Add forwardRef support to ChatInput component with ChatInputRef type
- Copy electron.d.ts to electron-standalone/electron folder
- Exclude electron-standalone from root tsconfig type checking

* fix: return empty string for valid result with no issues in formatValidationFeedback

* feat(i18n): add validation strings for ValidationCard component

- Add validation section to en.json, zh.json, ja.json dictionaries
- Update ValidationCard to use useDictionary hook
- Replace all hardcoded English strings with i18n keys

---------

Co-authored-by: dayuan.jiang <jdy.toh@gmail.com>

lib/ai-providers.ts

@@ -1177,3 +1177,27 @@ export function supportsImageInput(modelId: string): boolean {
     // Default: assume model supports images
     return true
 }
+
+/**
+ * Get the AI model for diagram validation.
+ * Uses VALIDATION_MODEL env var if set, otherwise falls back to AI_MODEL.
+ * Throws if the model doesn't support image input.
+ */
+export function getValidationModel(): ReturnType<typeof getAIModel>["model"] {
+    const modelId = process.env.VALIDATION_MODEL || process.env.AI_MODEL
+
+    if (!modelId) {
+        throw new Error(
+            "No validation model configured. Set VALIDATION_MODEL or AI_MODEL.",
+        )
+    }
+
+    if (!supportsImageInput(modelId)) {
+        throw new Error(
+            `Validation requires a vision-capable model. Model "${modelId}" does not support image input.`,
+        )
+    }
+
+    const { model } = getAIModel({ modelId })
+    return model
+}

lib/diagram-validator.ts

@@ -0,0 +1,64 @@
+/**
+ * Types and utilities for VLM-based diagram validation.
+ * The actual validation is performed via useValidateDiagram hook using AI SDK's useObject.
+ */
+
+// Re-export types from the schema file (single source of truth)
+export type { ValidationIssue, ValidationResult } from "./validation-schema"
+
+import type { ValidationResult } from "./validation-schema"
+
+/**
+ * Format validation feedback for display to the AI model.
+ * This creates a human-readable error message that guides the AI to fix issues.
+ *
+ * @param result - The validation result from VLM
+ * @returns Formatted string for tool error output
+ */
+export function formatValidationFeedback(result: ValidationResult): string {
+    // If validation passed with no issues, return empty string
+    if (result.valid && result.issues.length === 0) {
+        return ""
+    }
+
+    const lines: string[] = []
+
+    lines.push("DIAGRAM VISUAL VALIDATION FAILED")
+    lines.push("")
+
+    // Group issues by severity
+    const criticalIssues = result.issues.filter(
+        (i) => i.severity === "critical",
+    )
+    const warnings = result.issues.filter((i) => i.severity === "warning")
+
+    if (criticalIssues.length > 0) {
+        lines.push("Critical Issues (must fix):")
+        for (const issue of criticalIssues) {
+            lines.push(`  - [${issue.type}] ${issue.description}`)
+        }
+        lines.push("")
+    }
+
+    if (warnings.length > 0) {
+        lines.push("Warnings:")
+        for (const issue of warnings) {
+            lines.push(`  - [${issue.type}] ${issue.description}`)
+        }
+        lines.push("")
+    }
+
+    if (result.suggestions.length > 0) {
+        lines.push("Suggestions to fix:")
+        for (const suggestion of result.suggestions) {
+            lines.push(`  - ${suggestion}`)
+        }
+        lines.push("")
+    }
+
+    lines.push(
+        "Please regenerate the diagram with corrected layout to fix these visual issues.",
+    )
+
+    return lines.join("\n")
+}

lib/i18n/dictionaries/en.json

@@ -115,7 +115,11 @@
         "httpProxy": "HTTP Proxy",
         "httpsProxy": "HTTPS Proxy",
         "applyProxy": "Apply",
-        "proxyApplied": "Proxy settings applied"
+        "proxyApplied": "Proxy settings applied",
+        "diagramValidation": "Diagram Validation (Experimental)",
+        "diagramValidationDescription": "Use a vision language model to validate generated diagrams. Requires a VLM like GPT-5.2 or Sonnet-4.5.",
+        "enabled": "Enabled",
+        "disabled": "Disabled"
     },
     "save": {
         "title": "Save Diagram",
@@ -248,6 +252,24 @@
         "searchPlaceholder": "Search chats...",
         "noResults": "No chats found"
     },
+    "validation": {
+        "title": "Validate Diagram",
+        "capturing": "Capturing",
+        "validating": "Validating",
+        "validatingWithAttempt": "Validating ({attempt}/{max})",
+        "valid": "Valid",
+        "validWithWarnings": "Valid with Warnings",
+        "issuesFound": "Issues Found",
+        "error": "Error",
+        "skipped": "Skipped",
+        "capturedScreenshot": "Captured Screenshot:",
+        "issuesFoundLabel": "Issues Found:",
+        "suggestions": "Suggestions:",
+        "passedValidation": "Diagram passed visual validation - no issues detected.",
+        "improvementRequested": "Improvement requested - check the new diagram below",
+        "improveWithSuggestions": "Improve with Suggestions",
+        "regenerateWithFeedback": "Regenerate the diagram using the validation feedback"
+    },
     "modelConfig": {
         "title": "AI Model Configuration",
         "description": "Configure multiple AI providers and models",

lib/i18n/dictionaries/ja.json

@@ -115,7 +115,11 @@
         "httpProxy": "HTTP プロキシ",
         "httpsProxy": "HTTPS プロキシ",
         "applyProxy": "適用",
-        "proxyApplied": "プロキシ設定が適用されました"
+        "proxyApplied": "プロキシ設定が適用されました",
+        "diagramValidation": "ダイアグラム検証（実験的）",
+        "diagramValidationDescription": "視覚言語モデルを使用して生成されたダイアグラムを検証します。GPT-5.2 や Sonnet-4.5 などの VLM が必要です。",
+        "enabled": "有効",
+        "disabled": "無効"
     },
     "save": {
         "title": "ダイアグラムを保存",
@@ -248,6 +252,24 @@
         "searchPlaceholder": "チャットを検索...",
         "noResults": "チャットが見つかりません"
     },
+    "validation": {
+        "title": "ダイアグラムを検証",
+        "capturing": "キャプチャ中",
+        "validating": "検証中",
+        "validatingWithAttempt": "検証中 ({attempt}/{max})",
+        "valid": "有効",
+        "validWithWarnings": "有効（警告あり）",
+        "issuesFound": "問題が見つかりました",
+        "error": "エラー",
+        "skipped": "スキップ",
+        "capturedScreenshot": "キャプチャした画像：",
+        "issuesFoundLabel": "検出された問題：",
+        "suggestions": "提案：",
+        "passedValidation": "ダイアグラムは視覚検証に合格しました - 問題は検出されませんでした。",
+        "improvementRequested": "改善リクエスト済み - 下の新しいダイアグラムを確認してください",
+        "improveWithSuggestions": "提案で改善",
+        "regenerateWithFeedback": "検証フィードバックを使用してダイアグラムを再生成"
+    },
     "modelConfig": {
         "title": "AIモデル設定",
         "description": "複数のAIプロバイダーとモデルを設定",

lib/i18n/dictionaries/zh.json

@@ -115,7 +115,11 @@
         "httpProxy": "HTTP 代理",
         "httpsProxy": "HTTPS 代理",
         "applyProxy": "应用",
-        "proxyApplied": "代理设置已应用"
+        "proxyApplied": "代理设置已应用",
+        "diagramValidation": "图表验证（实验性）",
+        "diagramValidationDescription": "使用视觉语言模型验证生成的图表。需要支持视觉的模型，如 GPT-5.2 或 Sonnet-4.5。",
+        "enabled": "已启用",
+        "disabled": "已禁用"
     },
     "save": {
         "title": "保存图表",
@@ -248,6 +252,24 @@
         "searchPlaceholder": "搜索对话...",
         "noResults": "未找到对话"
     },
+    "validation": {
+        "title": "验证图表",
+        "capturing": "截图中",
+        "validating": "验证中",
+        "validatingWithAttempt": "验证中 ({attempt}/{max})",
+        "valid": "通过",
+        "validWithWarnings": "通过（有警告）",
+        "issuesFound": "发现问题",
+        "error": "错误",
+        "skipped": "已跳过",
+        "capturedScreenshot": "截图预览：",
+        "issuesFoundLabel": "发现的问题：",
+        "suggestions": "建议：",
+        "passedValidation": "图表通过视觉验证 - 未发现问题。",
+        "improvementRequested": "改进请求已发送 - 请查看下方新图表",
+        "improveWithSuggestions": "根据建议改进",
+        "regenerateWithFeedback": "使用验证反馈重新生成图表"
+    },
     "modelConfig": {
         "title": "AI 模型配置",
         "description": "配置多个 AI 提供商和模型",

lib/storage.ts

@@ -24,4 +24,7 @@ export const STORAGE_KEYS = {
 
     // Chat input preferences
     sendShortcut: "next-ai-draw-io-send-shortcut",
+
+    // Diagram validation
+    vlmValidationEnabled: "next-ai-draw-io-vlm-validation-enabled",
 } as const

lib/validation-prompts.ts

@@ -0,0 +1,22 @@
+/**
+ * VLM system prompt for diagram validation.
+ * Note: Response parsing is now handled via AI SDK's structured outputs (generateObject with schema).
+ */
+
+export const VALIDATION_SYSTEM_PROMPT = `You are a diagram quality validator. Analyze the rendered diagram image for visual issues.
+
+Evaluate the diagram for the following issues:
+
+1. **Overlapping elements** (critical): Shapes covering each other inappropriately, making content unreadable
+2. **Edge routing issues** (critical): Lines/arrows crossing through shapes that are not their source or target
+3. **Text readability** (warning): Labels cut off, overlapping, or too small to read
+4. **Layout quality** (warning): Poor spacing, misalignment, or cramped elements
+5. **Rendering errors** (critical): Incomplete, corrupted, or missing visual elements
+
+Rules:
+- Set "valid" to true ONLY if there are no critical issues
+- Be specific about which elements have problems (e.g., "The 'Login' box overlaps with 'Register' box")
+- Provide actionable suggestions (e.g., "Move the Login box 50 pixels to the left")
+- Minor cosmetic issues (slight misalignment, non-uniform spacing) should be warnings, not critical
+- Empty diagrams or diagrams with only 1-2 elements should pass unless they have obvious errors
+- If the diagram looks generally acceptable, set valid to true even with minor warnings`

lib/validation-schema.ts

@@ -0,0 +1,38 @@
+/**
+ * Shared validation schema for VLM-based diagram validation.
+ * This file can be safely imported on both client and server.
+ */
+
+import { z } from "zod"
+
+// Schema for structured validation output
+export const ValidationResultSchema = z.object({
+    valid: z.boolean().describe("True if there are no critical issues"),
+    issues: z
+        .array(
+            z.object({
+                type: z
+                    .enum([
+                        "overlap",
+                        "edge_routing",
+                        "text",
+                        "layout",
+                        "rendering",
+                    ])
+                    .describe("Type of visual issue"),
+                severity: z
+                    .enum(["critical", "warning"])
+                    .describe("Severity level"),
+                description: z
+                    .string()
+                    .describe("Clear description of the issue"),
+            }),
+        )
+        .describe("List of visual issues found"),
+    suggestions: z
+        .array(z.string())
+        .describe("Actionable suggestions to fix issues"),
+})
+
+export type ValidationResult = z.infer<typeof ValidationResultSchema>
+export type ValidationIssue = ValidationResult["issues"][number]