Merge pull request #218 from cita-777/codex/modeltester-attachment-review-fixes

[codex] Fix model tester attachment replay and capability gating

src/server/transformers/anthropic/messages/index.test.ts

@@ -34,7 +34,7 @@ describe('anthropicMessagesTransformer protocol contract', () => {
     });
   });
 
-  it('parses native document blocks into canonical file parts', () => {
+  it('preserves native image and document blocks when parsing into canonical envelopes', () => {
     const result = anthropicMessagesTransformer.parseRequest({
       model: 'claude-sonnet-4-5',
       max_tokens: 256,
@@ -42,15 +42,23 @@ describe('anthropicMessagesTransformer protocol contract', () => {
         {
           role: 'user',
           content: [
-            { type: 'text', text: 'summarize this pdf' },
+            { type: 'text', text: 'inspect both attachments' },
+            {
+              type: 'image',
+              source: {
+                type: 'base64',
+                media_type: 'image/png',
+                data: 'QUFBQQ==',
+              },
+            },
             {
               type: 'document',
-              title: 'brief.pdf',
               source: {
                 type: 'base64',
                 media_type: 'application/pdf',
-                data: 'JVBERi0xLjQK',
+                data: 'JVBERi0x',
               },
+              title: 'brief.pdf',
             },
           ],
         },
@@ -58,20 +66,23 @@ describe('anthropicMessagesTransformer protocol contract', () => {
     });
 
     expect(result.error).toBeUndefined();
-    expect(result.value?.messages).toEqual([
+    expect(result.value).toMatchObject({
-      {
+      messages: [
-        role: 'user',
+        {
-        parts: [
+          role: 'user',
-          { type: 'text', text: 'summarize this pdf' },
+          parts: [
-          {
+            { type: 'text', text: 'inspect both attachments' },
-            type: 'file',
+            { type: 'image', url: 'data:image/png;base64,QUFBQQ==' },
-            filename: 'brief.pdf',
+            {
-            mimeType: 'application/pdf',
+              type: 'file',
-            fileData: 'JVBERi0xLjQK',
+              fileData: 'JVBERi0x',
-          },
+              filename: 'brief.pdf',
-        ],
+              mimeType: 'application/pdf',
-      },
+            },
-    ]);
+          ],
+        },
+      ],
+    });
   });
 
   it('builds native messages requests from canonical envelopes', () => {

src/server/transformers/gemini/generate-content/compatibility.ts

@@ -1,3 +1,4 @@
+import { toOpenAiChatFileBlock } from '../../shared/inputFile.js';
 import type { NormalizedFinalResponse } from '../../shared/normalized.js';
 import { extractReasoningMetadataFromGeminiRequest } from './convert.js';
 
@@ -34,24 +35,58 @@ function buildDataUrl(part: GeminiRecord): string | null {
   return `data:${mimeType};base64,${data}`;
 }
 
-function buildOpenAiFileBlock(input: {
+function buildInlineData(
-  fileData?: string;
+  part: GeminiRecord,
-  fileUrl?: string;
+): { mimeType: string; data: string } | null {
-  mimeType?: string;
+  const inlineData = isRecord(part.inlineData) ? part.inlineData : null;
-}): Record<string, unknown> | null {
+  if (!inlineData) return null;
-  const fileData = asTrimmedString(input.fileData);
+  const mimeType = asTrimmedString(inlineData.mime_type ?? inlineData.mimeType) || 'application/octet-stream';
-  const fileUrl = asTrimmedString(input.fileUrl);
+  const data = asTrimmedString(inlineData.data);
-  const mimeType = asTrimmedString(input.mimeType);
+  if (!data) return null;
-  if (!fileData && !fileUrl) return null;
+  return { mimeType, data };
+}
 
-  const file: Record<string, unknown> = {};
+function buildFileDataSource(
-  if (fileData) file.file_data = fileData;
+  part: GeminiRecord,
-  if (fileUrl && !fileData) file.file_url = fileUrl;
+): { fileUri: string; mimeType: string | null } | null {
-  if (mimeType) file.mime_type = mimeType;
+  const fileData = isRecord(part.fileData) ? part.fileData : null;
-  return {
+  if (!fileData) return null;
-    type: 'file',
+  const fileUri = asTrimmedString(fileData.fileUri ?? fileData.file_uri);
-    file,
+  if (!fileUri) return null;
-  };
+  const mimeType = asTrimmedString(fileData.mimeType ?? fileData.mime_type) || null;
+  return { fileUri, mimeType };
+}
+
+function toOpenAiBlockFromGeminiPart(part: GeminiRecord): Record<string, unknown> | null {
+  const inlineData = buildInlineData(part);
+  if (inlineData) {
+    const normalizedMimeType = inlineData.mimeType.toLowerCase();
+    if (normalizedMimeType.startsWith('image/')) {
+      return {
+        type: 'image_url',
+        image_url: { url: buildDataUrl(part)! },
+      };
+    }
+
+    return toOpenAiChatFileBlock({
+      fileData: inlineData.data,
+      mimeType: inlineData.mimeType,
+    });
+  }
+
+  const fileData = buildFileDataSource(part);
+  if (!fileData) return null;
+  if (fileData.mimeType?.toLowerCase().startsWith('image/')) {
+    return {
+      type: 'image_url',
+      image_url: { url: fileData.fileUri },
+    };
+  }
+
+  return toOpenAiChatFileBlock({
+    fileUrl: fileData.fileUri,
+    mimeType: fileData.mimeType,
+  });
 }
 
 function toOpenAiContent(contentParts: GeminiRecord[]): string | Array<Record<string, unknown>> {
@@ -65,41 +100,9 @@ function toOpenAiContent(contentParts: GeminiRecord[]): string | Array<Record<st
       continue;
     }
 
-    const dataUrl = buildDataUrl(part);
+    const block = toOpenAiBlockFromGeminiPart(part);
-    if (dataUrl) {
+    if (block) {
-      const inlineData = isRecord(part.inlineData) ? part.inlineData : null;
+      blocks.push(block);
-      const mimeType = asTrimmedString(inlineData?.mime_type ?? inlineData?.mimeType) || 'application/octet-stream';
-      if (isImageMimeType(mimeType)) {
-        blocks.push({
-          type: 'image_url',
-          image_url: { url: dataUrl },
-        });
-      } else {
-        const fileBlock = buildOpenAiFileBlock({
-          fileData: asTrimmedString(inlineData?.data),
-          mimeType,
-        });
-        if (fileBlock) blocks.push(fileBlock);
-      }
-      continue;
-    }
-
-    const fileData = isRecord(part.fileData) ? part.fileData : null;
-    const fileUri = asTrimmedString(fileData?.fileUri ?? fileData?.file_uri);
-    if (fileUri) {
-      const mimeType = asTrimmedString(fileData?.mimeType ?? fileData?.mime_type);
-      if (mimeType && isImageMimeType(mimeType)) {
-        blocks.push({
-          type: 'image_url',
-          image_url: { url: fileUri },
-        });
-      } else {
-        const fileBlock = buildOpenAiFileBlock({
-          fileUrl: fileUri,
-          mimeType: mimeType || undefined,
-        });
-        if (fileBlock) blocks.push(fileBlock);
-      }
     }
   }
 

src/server/transformers/gemini/generate-content/index.test.ts

@@ -130,6 +130,49 @@ describe('geminiGenerateContentTransformer.inbound', () => {
     });
   });
 
+  it('compatibility preserves inline document parts as OpenAI file blocks', () => {
+    const body = geminiGenerateContentTransformer.compatibility.buildOpenAiBodyFromGeminiRequest({
+      modelName: 'gpt-5.4',
+      stream: false,
+      body: {
+        contents: [
+          {
+            role: 'user',
+            parts: [
+              { text: 'summarize this file' },
+              {
+                inlineData: {
+                  mimeType: 'application/pdf',
+                  data: 'JVBERi0x',
+                },
+              },
+            ],
+          },
+        ],
+      },
+    });
+
+    expect(body).toEqual({
+      model: 'gpt-5.4',
+      stream: false,
+      messages: [
+        {
+          role: 'user',
+          content: [
+            { type: 'text', text: 'summarize this file' },
+            {
+              type: 'file',
+              file: {
+                file_data: 'JVBERi0x',
+                mime_type: 'application/pdf',
+              },
+            },
+          ],
+        },
+      ],
+    });
+  });
+
   it('preserves native Gemini request fields through normalization', () => {
     const body = geminiGenerateContentTransformer.inbound.normalizeRequest({
       contents: [

src/server/transformers/shared/chatFormatsCore.ts

@@ -1,6 +1,7 @@
 import {
   decodeAnthropicReasoningSignature,
 } from './reasoningTransport.js';
+import { toOpenAiChatFileBlock } from './inputFile.js';
 import {
   consumeThinkTaggedText,
   createThinkTagParserState,
@@ -86,6 +87,10 @@ function isNonEmptyString(value: unknown): value is string {
   return typeof value === 'string' && value.trim().length > 0;
 }
 
+function asTrimmedString(value: unknown): string {
+  return typeof value === 'string' ? value.trim() : '';
+}
+
 function pickFiniteNumber(value: unknown): number | undefined {
   return typeof value === 'number' && Number.isFinite(value) ? value : undefined;
 }
@@ -548,6 +553,111 @@ function collectToolCallsFromResponsesPayload(payload: Record<string, unknown>):
   return toolCalls;
 }
 
+function formatAnthropicBase64DataUrl(mimeType: string, data: string): string {
+  return `data:${mimeType};base64,${data}`;
+}
+
+function parseAnthropicBase64Source(
+  source: unknown,
+): { mimeType: string; data: string } | null {
+  if (!isRecord(source)) return null;
+  const sourceType = asTrimmedString(source.type).toLowerCase();
+  if (sourceType !== 'base64') return null;
+  const mimeType = (
+    asTrimmedString(source.media_type)
+    || asTrimmedString(source.mime_type)
+    || asTrimmedString(source.mediaType)
+    || asTrimmedString(source.mimeType)
+    || 'application/octet-stream'
+  );
+  const data = asTrimmedString(source.data);
+  if (!data) return null;
+  return { mimeType, data };
+}
+
+function parseAnthropicUrlSource(
+  source: unknown,
+): { url: string; mimeType: string | null } | null {
+  if (!isRecord(source)) return null;
+  const sourceType = asTrimmedString(source.type).toLowerCase();
+  if (sourceType !== 'url') return null;
+  const url = asTrimmedString(source.url);
+  if (!url) return null;
+  const mimeType = (
+    asTrimmedString(source.media_type)
+    || asTrimmedString(source.mime_type)
+    || asTrimmedString(source.mediaType)
+    || asTrimmedString(source.mimeType)
+    || null
+  );
+  return { url, mimeType };
+}
+
+function toOpenAiContentBlockFromClaudeBlock(
+  block: Record<string, unknown>,
+): Record<string, unknown> | null {
+  const blockType = asTrimmedString(block.type).toLowerCase();
+
+  if (blockType === 'text') {
+    const text = asTrimmedString(block.text);
+    return text ? { type: 'text', text } : null;
+  }
+
+  if (blockType !== 'image' && blockType !== 'document') {
+    return null;
+  }
+
+  const title = asTrimmedString(block.title);
+  const base64Source = parseAnthropicBase64Source(block.source);
+  const urlSource = parseAnthropicUrlSource(block.source);
+  const mimeType = (base64Source?.mimeType || urlSource?.mimeType || '').toLowerCase();
+  const treatAsImage = blockType === 'image' || mimeType.startsWith('image/');
+
+  if (treatAsImage) {
+    const imageUrl = base64Source
+      ? formatAnthropicBase64DataUrl(base64Source.mimeType, base64Source.data)
+      : (urlSource?.url || '');
+    return imageUrl
+      ? {
+        type: 'image_url',
+        image_url: { url: imageUrl },
+      }
+      : null;
+  }
+
+  if (base64Source) {
+    return toOpenAiChatFileBlock({
+      fileData: base64Source.data,
+      filename: title || undefined,
+      mimeType: base64Source.mimeType,
+    });
+  }
+
+  if (urlSource) {
+    return toOpenAiChatFileBlock({
+      fileUrl: urlSource.url,
+      filename: title || undefined,
+      mimeType: urlSource.mimeType,
+    });
+  }
+
+  return null;
+}
+
+function collapseOpenAiContentBlocks(
+  blocks: Array<Record<string, unknown>>,
+): string | Array<Record<string, unknown>> | null {
+  if (blocks.length <= 0) return null;
+  const textOnly = blocks.every((block) => block.type === 'text' && typeof block.text === 'string');
+  if (!textOnly) return blocks;
+
+  const text = blocks
+    .map((block) => (typeof block.text === 'string' ? block.text : ''))
+    .join('\n\n')
+    .trim();
+  return text || null;
+}
+
 function convertClaudeRequestToOpenAiBody(body: Record<string, unknown>): {
   model: string;
   stream: boolean;
@@ -609,10 +719,10 @@ function convertClaudeRequestToOpenAiBody(body: Record<string, unknown>): {
     const toolCalls: Array<Record<string, unknown>> = [];
 
     const flushContentAsMessage = () => {
-      const contentValue = buildOpenAiMessageContent(contentBlocks);
+      const contentPayload = collapseOpenAiContentBlocks(contentBlocks);
       contentBlocks.length = 0;
-      if (contentValue === undefined || contentValue === '') return;
+      if (contentPayload === null) return;
-      messages.push({ role: mappedRole, content: contentValue });
+      messages.push({ role: mappedRole, content: contentPayload });
     };
 
     for (const block of content) {
@@ -648,22 +758,33 @@ function convertClaudeRequestToOpenAiBody(body: Record<string, unknown>): {
         continue;
       }
 
-      const contentBlock = convertClaudeContentBlockToOpenAi(block);
+      const contentBlock = toOpenAiContentBlockFromClaudeBlock(block);
-      if (contentBlock) contentBlocks.push(contentBlock);
+      if (contentBlock) {
+        contentBlocks.push(contentBlock);
+        continue;
+      }
+
+      const text = parseClaudeMessageContent(block);
+      if (text) {
+        contentBlocks.push({
+          type: 'text',
+          text,
+        });
+      }
     }
 
-    const contentValue = buildOpenAiMessageContent(contentBlocks);
+    const merged = collapseOpenAiContentBlocks(contentBlocks);
     if (toolCalls.length > 0) {
       const assistantMessage: Record<string, unknown> = {
         role: 'assistant',
         tool_calls: toolCalls,
       };
-      assistantMessage.content = contentValue ?? '';
+      assistantMessage.content = merged || '';
       messages.push(assistantMessage);
-    } else if (contentValue !== undefined && contentValue !== '') {
+    } else if (merged !== null) {
       messages.push({
         role: mappedRole,
-        content: contentValue,
+        content: merged,
       });
     }
   }

src/web/pages/ModelTester.tsx

@@ -45,7 +45,12 @@ import {
   type TestTargetFormat,
   type TestChatPayload,
 } from './helpers/modelTesterSession.js';
-import { resolveConversationFileCapability } from './helpers/conversationFileCapabilities.js';
+import {
+  buildConversationFileAccept,
+  buildConversationFileHint,
+  isConversationUploadedFileSupported,
+  resolveConversationFileCapability,
+} from './helpers/conversationFileCapabilities.js';
 import ModernSelect from '../components/ModernSelect.js';
 import { useAnimatedVisibility } from '../components/useAnimatedVisibility.js';
 import { tr } from '../i18n.js';
@@ -77,7 +82,6 @@ type ConversationFileState = UploadState & {
 };
 
 const POLL_INTERVAL_MS = 1200;
-const CONVERSATION_FILE_ACCEPT = '.pdf,.txt,.md,.markdown,.json,image/*,audio/*';
 const wait = (ms: number) => new Promise<void>((resolve) => setTimeout(resolve, ms));
 
 const createConversationFileLocalId = () =>
@@ -690,6 +694,19 @@ export default function ModelTester() {
   const restoredSessionRef = useRef<ReturnType<typeof parseModelTesterSession>>(null);
   const streamAbortRef = useRef<AbortController | null>(null);
   const streamStopRequestedRef = useRef(false);
+  const conversationFileCapability = useMemo(
+    () => resolveConversationFileCapability(inputs.protocol),
+    [inputs.protocol],
+  );
+  const conversationFileSupported = conversationFileCapability.supported;
+  const conversationFileAccept = useMemo(
+    () => buildConversationFileAccept(conversationFileCapability),
+    [conversationFileCapability],
+  );
+  const conversationFileHint = useMemo(
+    () => buildConversationFileHint(conversationFileCapability),
+    [conversationFileCapability],
+  );
 
   const pushDebug = useCallback((level: DebugTimelineEntry['level'], text: string) => {
     const now = new Date().toISOString();
@@ -881,18 +898,35 @@ export default function ModelTester() {
         name: file.name,
         mimeType: file.type || 'application/octet-stream',
         dataUrl: await readFileAsDataUrl(file),
-        fileId: null,
+          fileId: null,
-        status: 'pending' as const,
+          status: 'pending' as const,
-        errorMessage: null,
+          errorMessage: null,
-      })));
+        })));
-      setConversationFiles((prev) => [...prev, ...nextFiles]);
+      const acceptedFiles = nextFiles.filter((file) => isConversationUploadedFileSupported(
-      pushDebug('info', `已添加 ${nextFiles.length} 个会话附件。`);
+        conversationFileCapability,
+        { filename: file.name, mimeType: file.mimeType },
+      ));
+      const rejectedFiles = nextFiles.filter((file) => !isConversationUploadedFileSupported(
+        conversationFileCapability,
+        { filename: file.name, mimeType: file.mimeType },
+      ));
+
+      if (acceptedFiles.length > 0) {
+        setConversationFiles((prev) => [...prev, ...acceptedFiles]);
+        pushDebug('info', `已添加 ${acceptedFiles.length} 个会话附件。`);
+      }
+
+      if (rejectedFiles.length > 0) {
+        const message = `当前协议不支持这些会话附件：${rejectedFiles.map((file) => file.name).join('、')}。${conversationFileHint}`;
+        setError(message);
+        pushDebug('warn', message);
+      }
     } catch (readError: any) {
       const message = readError?.message || '读取附件失败';
       setError(message);
       pushDebug('error', message);
     }
-  }, [pushDebug]);
+  }, [conversationFileCapability, conversationFileHint, pushDebug]);
 
   const removeConversationFile = useCallback((localId: string) => {
     if (sending) return;
@@ -966,6 +1000,20 @@ export default function ModelTester() {
       data: item.dataUrl,
     })), [conversationFiles]);
 
+  const ensureSupportedConversationFiles = useCallback((files: ConversationUploadedFile[]): boolean => {
+    const unsupported = files.filter((file) => !isConversationUploadedFileSupported(conversationFileCapability, file));
+    if (unsupported.length <= 0) return true;
+
+    const names = unsupported.map((file, index) => {
+      const filename = typeof file.filename === 'string' ? file.filename.trim() : '';
+      return filename || `附件${index + 1}`;
+    });
+    const message = `当前协议不支持这些会话附件：${names.join('、')}。${conversationFileHint}`;
+    setError(message);
+    pushDebug('warn', message);
+    return false;
+  }, [conversationFileCapability, conversationFileHint, pushDebug]);
+
   const buildConversationMessagesWithSystem = useCallback((baseMessages: ChatMessage[]) => {
     if (!inputs.systemPrompt.trim()) return baseMessages;
     return [
@@ -1335,11 +1383,6 @@ export default function ModelTester() {
     () => filteredModels.map((item) => ({ value: item, label: item })),
     [filteredModels],
   );
-  const conversationFileCapability = useMemo(
-    () => resolveConversationFileCapability(inputs.protocol),
-    [inputs.protocol],
-  );
-  const conversationFileSupported = conversationFileCapability.supported;
   const canSend = useMemo(() => {
     if (sending || pendingJobId || !inputs.model) return false;
     if (inputs.mode !== 'conversation') {
@@ -1762,6 +1805,9 @@ export default function ModelTester() {
     baseMessages: ChatMessage[],
     files: ConversationUploadedFile[] = [],
   ) => {
+    if (!ensureSupportedConversationFiles(files)) {
+      return;
+    }
     const userMessage = createConversationUserMessage(prompt, files);
     const loadingAssistant = createLoadingAssistantMessage();
     const nextMessages = [...baseMessages, userMessage, loadingAssistant];
@@ -1780,7 +1826,7 @@ export default function ModelTester() {
     }
 
     await dispatchPayload(nextMessages, payload, { syncedCustomBody });
-  }, [buildConversationProxyEnvelope, buildPayloadWithMessages, createConversationUserMessage, customRequestMode, dispatchPayload, dispatchProxyEnvelope, inputs.protocol, pushDebug]);
+  }, [buildConversationProxyEnvelope, buildPayloadWithMessages, createConversationUserMessage, customRequestMode, dispatchPayload, dispatchProxyEnvelope, ensureSupportedConversationFiles, inputs.protocol, pushDebug]);
 
   const sendModeRequest = useCallback(async () => {
     const envelope = buildModeProxyEnvelope();
@@ -1817,8 +1863,13 @@ export default function ModelTester() {
     if (!customRequestMode && conversationFileSupported && conversationFiles.length > 0) {
       setSending(true);
       try {
+        const draftFiles = inlineConversationFiles();
+        if (!ensureSupportedConversationFiles(draftFiles)) {
+          setSending(false);
+          return;
+        }
         const uploadedFiles = conversationFileCapability.documentMode === 'inline_only'
-          ? inlineConversationFiles()
+          ? draftFiles
           : await uploadConversationFiles();
         setInput('');
         setConversationFiles([]);
@@ -1857,7 +1908,7 @@ export default function ModelTester() {
         { stream: inputs.stream, jobMode: !inputs.stream },
       ),
     );
-  }, [canSend, conversationFileCapability, conversationFileSupported, conversationFiles.length, customRequestBody, customRequestMode, dispatchPayload, inlineConversationFiles, input, inputs.mode, messages, pushDebug, sendModeRequest, sendWithPrompt, uploadConversationFiles]);
+  }, [canSend, conversationFileCapability, conversationFileSupported, conversationFiles.length, customRequestBody, customRequestMode, dispatchPayload, ensureSupportedConversationFiles, inlineConversationFiles, input, inputs.mode, messages, pushDebug, sendModeRequest, sendWithPrompt, uploadConversationFiles]);
 
   const retryPending = useCallback(async () => {
     if (sending || pendingJobId || !pendingPayload) return;
@@ -2793,7 +2844,7 @@ export default function ModelTester() {
                       ref={conversationFileInputRef}
                       type="file"
                       multiple
-                      accept={CONVERSATION_FILE_ACCEPT}
+                      accept={conversationFileAccept}
                       style={{ display: 'none' }}
                       onChange={(event) => {
                         void handleConversationFilesChange(event.target.files);
@@ -2815,9 +2866,7 @@ export default function ModelTester() {
                           ? '自定义请求模式不会自动上传这些附件；关闭自定义模式后可走标准 /v1/files 链路。'
                           : !conversationFileSupported
                             ? (conversationFileCapability.reason || '当前协议暂不支持会话附件注入。')
-                            : conversationFileCapability.documentMode === 'inline_only'
+                            : conversationFileHint}
-                              ? '支持 PDF / TXT / Markdown / JSON / 图片 / 音频；发送时会以内联文档方式注入。'
-                              : '支持 PDF / TXT / Markdown / JSON / 图片 / 音频；发送前会先上传到 /v1/files。'}
                       </span>
                     </div>
                     {conversationFiles.length > 0 && (

src/web/pages/helpers/conversationFileCapabilities.test.ts

@@ -1,31 +1,79 @@
 import { describe, expect, it } from 'vitest';
 
-import { resolveConversationFileCapability } from './conversationFileCapabilities.js';
+import {
+  buildConversationFileAccept,
+  buildConversationFileHint,
+  isConversationUploadedFileSupported,
+  resolveConversationFileCapability,
+} from './conversationFileCapabilities.js';
 
 describe('resolveConversationFileCapability', () => {
   it('keeps OpenAI and Responses on native file references', () => {
     expect(resolveConversationFileCapability('openai')).toEqual({
       supported: true,
+      imageMode: 'native',
+      audioMode: 'native',
       documentMode: 'native',
       reason: '',
     });
     expect(resolveConversationFileCapability('responses')).toEqual({
       supported: true,
+      imageMode: 'native',
+      audioMode: 'native',
       documentMode: 'native',
       reason: '',
     });
   });
 
-  it('marks Claude and Gemini picker flows as inline-only document transports', () => {
+  it('maps Claude and Gemini media support separately', () => {
     expect(resolveConversationFileCapability('claude')).toEqual({
       supported: true,
+      imageMode: 'native',
+      audioMode: 'unsupported',
       documentMode: 'inline_only',
       reason: '当前界面的会话附件会以内联文档方式发送。',
     });
     expect(resolveConversationFileCapability('gemini')).toEqual({
       supported: true,
+      imageMode: 'native',
+      audioMode: 'native',
       documentMode: 'inline_only',
       reason: '当前界面的会话附件会以内联文档方式发送。',
     });
   });
+
+  it('builds the accept list from the supported file kinds', () => {
+    expect(buildConversationFileAccept(resolveConversationFileCapability('claude'))).toBe(
+      '.pdf,.txt,.md,.markdown,.json,image/*',
+    );
+    expect(buildConversationFileAccept(resolveConversationFileCapability('gemini'))).toBe(
+      '.pdf,.txt,.md,.markdown,.json,image/*,audio/*',
+    );
+  });
+
+  it('describes the actual per-kind transport behavior', () => {
+    expect(buildConversationFileHint(resolveConversationFileCapability('claude'))).toBe(
+      '支持 PDF / TXT / Markdown / JSON / 图片；文档会以内联数据注入，图片会按图片部件发送。',
+    );
+    expect(buildConversationFileHint(resolveConversationFileCapability('gemini'))).toBe(
+      '支持 PDF / TXT / Markdown / JSON / 图片 / 音频；文档会以内联数据注入，图片会按图片部件发送，音频会按音频部件发送。',
+    );
+  });
+
+  it('rejects unsupported file kinds before send', () => {
+    const claudeCapability = resolveConversationFileCapability('claude');
+
+    expect(isConversationUploadedFileSupported(claudeCapability, {
+      filename: 'paper.pdf',
+      mimeType: 'application/pdf',
+    })).toBe(true);
+    expect(isConversationUploadedFileSupported(claudeCapability, {
+      filename: 'chart.png',
+      mimeType: 'image/png',
+    })).toBe(true);
+    expect(isConversationUploadedFileSupported(claudeCapability, {
+      filename: 'voice.mp3',
+      mimeType: 'audio/mpeg',
+    })).toBe(false);
+  });
 });

src/web/pages/helpers/conversationFileCapabilities.ts

@@ -1,25 +1,101 @@
 import type { PlaygroundProtocol } from './modelTesterSession.js';
 
+export type ConversationFileTransportMode = 'native' | 'inline_only' | 'unsupported';
+
 export type ConversationFileCapability = {
   supported: boolean;
-  documentMode: 'native' | 'inline_only' | 'unsupported';
+  imageMode: ConversationFileTransportMode;
+  audioMode: ConversationFileTransportMode;
+  documentMode: ConversationFileTransportMode;
   reason: string;
 };
 
+const DOCUMENT_ACCEPT_PARTS = ['.pdf', '.txt', '.md', '.markdown', '.json'] as const;
+const DOCUMENT_MIME_TYPES = new Set([
+  'application/json',
+  'application/pdf',
+  'text/markdown',
+  'text/plain',
+]);
+const DOCUMENT_EXTENSIONS = ['.json', '.md', '.markdown', '.pdf', '.txt'] as const;
+const IMAGE_EXTENSIONS = ['.avif', '.bmp', '.gif', '.jpeg', '.jpg', '.png', '.svg', '.webp'] as const;
+const AUDIO_EXTENSIONS = ['.aac', '.flac', '.m4a', '.mp3', '.ogg', '.wav', '.weba'] as const;
+
+type ConversationFileDescriptor = {
+  filename?: string | null;
+  mimeType?: string | null;
+};
+
+type ConversationFileKind = 'document' | 'image' | 'audio' | 'unknown';
+
+const isSupportedMode = (mode: ConversationFileTransportMode): boolean => mode !== 'unsupported';
+
+const normalizeName = (value: string | null | undefined): string => (value || '').trim().toLowerCase();
+
+function detectKindFromFilename(filename: string): ConversationFileKind {
+  if (!filename) return 'unknown';
+  if (DOCUMENT_EXTENSIONS.some((extension) => filename.endsWith(extension))) return 'document';
+  if (IMAGE_EXTENSIONS.some((extension) => filename.endsWith(extension))) return 'image';
+  if (AUDIO_EXTENSIONS.some((extension) => filename.endsWith(extension))) return 'audio';
+  return 'unknown';
+}
+
+function resolveConversationFileKind(file: ConversationFileDescriptor): ConversationFileKind {
+  const mimeType = normalizeName(file.mimeType);
+  if (mimeType.startsWith('image/')) return 'image';
+  if (mimeType.startsWith('audio/')) return 'audio';
+  if (DOCUMENT_MIME_TYPES.has(mimeType)) return 'document';
+  return detectKindFromFilename(normalizeName(file.filename));
+}
+
+function buildSupportedTypeLabel(capability: ConversationFileCapability): string {
+  const labels: string[] = [];
+  if (isSupportedMode(capability.documentMode)) labels.push('PDF / TXT / Markdown / JSON');
+  if (isSupportedMode(capability.imageMode)) labels.push('图片');
+  if (isSupportedMode(capability.audioMode)) labels.push('音频');
+  return labels.join(' / ');
+}
+
+function buildTransportNotes(capability: ConversationFileCapability): string[] {
+  const notes: string[] = [];
+  if (capability.documentMode === 'inline_only') notes.push('文档会以内联数据注入');
+  if (capability.imageMode === 'native' && capability.documentMode === 'inline_only') {
+    notes.push('图片会按图片部件发送');
+  }
+  if (capability.audioMode === 'native' && capability.documentMode === 'inline_only') {
+    notes.push('音频会按音频部件发送');
+  }
+  return notes;
+}
+
 export function resolveConversationFileCapability(
   protocol: PlaygroundProtocol,
 ): ConversationFileCapability {
   if (protocol === 'openai' || protocol === 'responses') {
     return {
       supported: true,
+      imageMode: 'native',
+      audioMode: 'native',
       documentMode: 'native',
       reason: '',
     };
   }
 
-  if (protocol === 'claude' || protocol === 'gemini') {
+  if (protocol === 'claude') {
     return {
       supported: true,
+      imageMode: 'native',
+      audioMode: 'unsupported',
+      documentMode: 'inline_only',
+      reason: '当前界面的会话附件会以内联文档方式发送。',
+    };
+  }
+
+  if (protocol === 'gemini') {
+    return {
+      supported: true,
+      imageMode: 'native',
+      audioMode: 'native',
       documentMode: 'inline_only',
       reason: '当前界面的会话附件会以内联文档方式发送。',
     };
@@ -27,7 +103,53 @@ export function resolveConversationFileCapability(
 
   return {
     supported: false,
+    imageMode: 'unsupported',
+    audioMode: 'unsupported',
     documentMode: 'unsupported',
     reason: '当前协议暂不支持会话附件。',
   };
 }
+
+export function buildConversationFileAccept(capability: ConversationFileCapability): string {
+  const parts: string[] = [];
+  if (isSupportedMode(capability.documentMode)) parts.push(...DOCUMENT_ACCEPT_PARTS);
+  if (isSupportedMode(capability.imageMode)) parts.push('image/*');
+  if (isSupportedMode(capability.audioMode)) parts.push('audio/*');
+  return parts.join(',');
+}
+
+export function buildConversationFileHint(capability: ConversationFileCapability): string {
+  if (!capability.supported) {
+    return capability.reason || '当前协议暂不支持会话附件。';
+  }
+
+  const typeLabel = buildSupportedTypeLabel(capability);
+  if (!typeLabel) {
+    return capability.reason || '当前协议暂不支持会话附件。';
+  }
+
+  if (
+    capability.documentMode === 'native'
+    && capability.imageMode === 'native'
+    && capability.audioMode === 'native'
+  ) {
+    return `支持 ${typeLabel}；发送前会先上传到 /v1/files。`;
+  }
+
+  const notes = buildTransportNotes(capability);
+  if (notes.length <= 0) {
+    return `支持 ${typeLabel}。`;
+  }
+  return `支持 ${typeLabel}；${notes.join('，')}。`;
+}
+
+export function isConversationUploadedFileSupported(
+  capability: ConversationFileCapability,
+  file: ConversationFileDescriptor,
+): boolean {
+  const kind = resolveConversationFileKind(file);
+  if (kind === 'document') return isSupportedMode(capability.documentMode);
+  if (kind === 'image') return isSupportedMode(capability.imageMode);
+  if (kind === 'audio') return isSupportedMode(capability.audioMode);
+  return false;
+}

src/web/pages/helpers/modelTesterSession.test.ts

@@ -535,6 +535,54 @@ describe('modelTesterSession', () => {
     });
   });
 
+  it('builds Claude conversation payloads with image blocks for inline image attachments', () => {
+    const payload = buildApiPayload(
+      [{
+        id: 'u1',
+        role: 'user',
+        content: 'describe this image',
+        createAt: 1,
+        parts: [
+          {
+            type: 'input_file',
+            filename: 'chart.png',
+            mimeType: 'image/png',
+            data: 'data:image/png;base64,QUFBQQ==',
+          },
+        ],
+      } as ChatMessage],
+      {
+        ...DEFAULT_INPUTS,
+        model: 'claude-opus-4-6',
+        protocol: 'claude',
+      },
+      DEFAULT_PARAMETER_ENABLED,
+    );
+
+    expect(payload.jsonBody).toEqual({
+      model: 'claude-opus-4-6',
+      stream: false,
+      max_tokens: 4096,
+      messages: [
+        {
+          role: 'user',
+          content: [
+            { type: 'text', text: 'describe this image' },
+            {
+              type: 'image',
+              source: {
+                type: 'base64',
+                media_type: 'image/png',
+                data: 'QUFBQQ==',
+              },
+            },
+          ],
+        },
+      ],
+      temperature: 0.7,
+    });
+  });
+
   it('builds Gemini conversation payloads with inlineData document parts for conversation files', () => {
     const payload = buildApiPayload(
       [{

src/web/pages/helpers/modelTesterSession.ts

@@ -355,11 +355,22 @@ const toClaudeContentPart = (part: ConversationContentPart): Record<string, unkn
   if (part.type !== 'input_file' || typeof part.data !== 'string' || !part.data.trim()) return null;
   const parsed = parseDataUrl(part.data);
   if (!parsed) return null;
+  const mimeType = part.mimeType || parsed.mimeType || 'application/octet-stream';
+  if (mimeType.toLowerCase().startsWith('image/')) {
+    return {
+      type: 'image',
+      source: {
+        type: 'base64',
+        media_type: mimeType,
+        data: parsed.data,
+      },
+    };
+  }
   return {
     type: 'document',
     source: {
       type: 'base64',
-      media_type: part.mimeType || parsed.mimeType || 'application/octet-stream',
+      media_type: mimeType,
       data: parsed.data,
     },
     ...(typeof part.filename === 'string' && part.filename.trim()