Upload 62 files

ai-routes.js

@@ -29,7 +29,7 @@ async function recordUsage(model, provider) {
     } catch (e) { console.error("Failed to record AI usage stats:", e); }
 }
 
-// Fallback Speech-to-Text using Hugging Face Inference API
+// Fallback Speech-to-Text using Hugging Face (Fixed 410 Error)
 async function transcribeAudioWithHF(audioBase64) {
     const token = await getHFToken();
     if (!token) {
@@ -38,42 +38,33 @@ async function transcribeAudioWithHF(audioBase64) {
     }
 
     try {
-        console.log("[AI] 🎤 Using Hugging Face Fallback STT...");
-        const response = await fetch(
-            "https://api-inference.huggingface.co/models/openai/whisper-large-v3",
-            {
-                headers: {
-                    Authorization: `Bearer ${token}`,
-                    "Content-Type": "application/json",
-                },
-                method: "POST",
-                body: JSON.stringify({
-                    inputs: audioBase64, // HF Inference often accepts base64 direct in inputs or raw bytes. For stability with free tier, simple base64 JSON payload often works for some wrappers, but raw bytes are safer for 'audio/...' types.
-                    // Actually, let's convert to buffer then send raw bytes for better compatibility.
-                }),
-            }
-        );
-        
-        // Retry with raw bytes if JSON failed or for standard audio models
+        console.log("[AI] 🎤 Using Hugging Face ASR (Whisper v3)...");
         const buffer = Buffer.from(audioBase64, 'base64');
-        const rawResponse = await fetch(
+        
+        // Using the standard inference endpoint which is more stable
+        const response = await fetch(
             "https://api-inference.huggingface.co/models/openai/whisper-large-v3",
             {
                 headers: {
                     Authorization: `Bearer ${token}`,
-                    "Content-Type": "audio/webm", // Assuming webm from frontend
                 },
                 method: "POST",
                 body: buffer,
             }
         );
 
-        if (!rawResponse.ok) {
-            const errText = await rawResponse.text();
-            throw new Error(`HF API Error: ${rawResponse.status} ${errText}`);
+        if (!response.ok) {
+            const errText = await response.text();
+            if (response.status === 503) {
+                console.log("[AI] HF Model loading, retrying...");
+                await new Promise(r => setTimeout(r, 3000));
+                return transcribeAudioWithHF(audioBase64);
+            }
+            throw new Error(`HF API Error: ${response.status} ${errText}`);
         }
 
-        const result = await rawResponse.json();
+        const result = await response.json();
+        console.log("[AI] HF Transcribed:", result.text);
         return result.text;
     } catch (e) {
         console.error("[AI] HF STT Failed:", e.message);
@@ -84,16 +75,12 @@ async function transcribeAudioWithHF(audioBase64) {
 function convertGeminiToOpenAI(baseParams) {
     const messages = [];
     if (baseParams.config?.systemInstruction) messages.push({ role: 'system', content: baseParams.config.systemInstruction });
-    
     let contents = baseParams.contents;
-    if (contents && !Array.isArray(contents)) {
-        contents = [contents];
-    }
+    if (contents && !Array.isArray(contents)) contents = [contents];
 
     if (contents && Array.isArray(contents)) {
         contents.forEach(content => {
             let role = (content.role === 'model' || content.role === 'assistant') ? 'assistant' : 'user';
-            
             const messageContent = [];
             if (content.parts) {
                 content.parts.forEach(p => {
@@ -102,13 +89,11 @@ function convertGeminiToOpenAI(baseParams) {
                         if (p.inlineData.mimeType.startsWith('image/')) {
                             messageContent.push({ type: 'image_url', image_url: { url: `data:${p.inlineData.mimeType};base64,${p.inlineData.data}` } });
                         } else if (p.inlineData.mimeType.startsWith('audio/')) {
-                            // Temporary marker for audio, will be resolved before sending if possible
                             messageContent.push({ type: 'audio_base64', data: p.inlineData.data });
                         }
                     }
                 });
             }
-            
             if (messageContent.length > 0) {
                 if (messageContent.length === 1 && messageContent[0].type === 'text') {
                     messages.push({ role: role, content: messageContent[0].text });
@@ -124,14 +109,11 @@ function convertGeminiToOpenAI(baseParams) {
 const PROVIDERS = { GEMINI: 'GEMINI', OPENROUTER: 'OPENROUTER', GEMMA: 'GEMMA' };
 const DEFAULT_OPENROUTER_MODELS = ['qwen/qwen3-coder:free', 'openai/gpt-oss-120b:free', 'qwen/qwen3-235b-a22b:free', 'tngtech/deepseek-r1t-chimera:free'];
 
-// Runtime override logic
 let runtimeProviderOrder = [];
 
 function deprioritizeProvider(providerName) {
     if (runtimeProviderOrder.length > 0 && runtimeProviderOrder[runtimeProviderOrder.length - 1] === providerName) return;
-    console.log(`[AI System] ⚠️ Deprioritizing ${providerName} due to errors. Moving to end of queue.`);
     runtimeProviderOrder = runtimeProviderOrder.filter(p => p !== providerName).concat(providerName);
-    console.log(`[AI System] 🔄 New Priority Order: ${runtimeProviderOrder.join(' -> ')}`);
 }
 
 function isQuotaError(e) {
@@ -139,7 +121,6 @@ function isQuotaError(e) {
     return e.status === 429 || e.status === 503 || msg.includes('quota') || msg.includes('overloaded') || msg.includes('resource_exhausted') || msg.includes('rate limit') || msg.includes('credits');
 }
 
-// Streaming Helpers
 async function streamGemini(baseParams, res) {
     const { GoogleGenAI } = await import("@google/genai");
     const models = ['gemini-2.5-flash', 'gemini-2.5-flash-lite'];
@@ -150,35 +131,21 @@ async function streamGemini(baseParams, res) {
         const client = new GoogleGenAI({ apiKey });
         for (const modelName of models) {
             try {
-                console.log(`[AI] 🚀 Attempting Gemini Model: ${modelName} (Key ends with ...${apiKey.slice(-4)})`);
                 const result = await client.models.generateContentStream({ ...baseParams, model: modelName });
-                
                 let hasStarted = false;
                 let fullText = "";
-                
                 for await (const chunk of result) {
-                    if (!hasStarted) {
-                        console.log(`[AI] ✅ Connected to Gemini: ${modelName}`);
-                        recordUsage(modelName, PROVIDERS.GEMINI);
-                        hasStarted = true;
-                    }
+                    if (!hasStarted) { recordUsage(modelName, PROVIDERS.GEMINI); hasStarted = true; }
                     if (chunk.text) {
                         fullText += chunk.text;
                         res.write(`data: ${JSON.stringify({ text: chunk.text })}\n\n`);
-                        if (res.flush) res.flush();
                     }
                 }
                 return fullText;
-            } catch (e) {
-                console.warn(`[AI] ⚠️ Gemini ${modelName} Error: ${e.message}`);
-                if (isQuotaError(e)) {
-                    continue; 
-                } 
-                throw e; 
-            }
+            } catch (e) { if (isQuotaError(e)) continue; throw e; }
         }
     }
-    throw new Error("Gemini streaming failed (All keys/models exhausted)");
+    throw new Error("Gemini exhausted");
 }
 
 async function streamOpenRouter(baseParams, res) {
@@ -188,22 +155,17 @@ async function streamOpenRouter(baseParams, res) {
     const keys = await getKeyPool('openrouter');
     if (keys.length === 0) throw new Error("No OpenRouter API keys");
     
-    // --- FALLBACK STT LOGIC ---
-    // Check if there are any pending audio parts in messages
-    let hasAudio = false;
     for (let msg of messages) {
         if (Array.isArray(msg.content)) {
             for (let part of msg.content) {
                 if (part.type === 'audio_base64') {
-                    hasAudio = true;
-                    // Try to transcribe
                     const text = await transcribeAudioWithHF(part.data);
                     if (text) {
                         part.type = 'text';
-                        part.text = `[Voice Input Transcribed]: ${text}`;
-                        delete part.data; // Remove heavy data
+                        part.text = `[语音转文字]: ${text}`;
+                        delete part.data;
                     } else {
-                        throw new Error("Failed to transcribe audio for non-Gemini model (Check HF Token)");
+                        throw new Error("语音转文字失败 (请检查 HF Token)");
                     }
                 }
             }
@@ -214,125 +176,68 @@ async function streamOpenRouter(baseParams, res) {
         for (const modelName of models) {
             const modelConfig = config?.openRouterModels?.find(m => m.id === modelName);
             const baseURL = modelConfig?.apiUrl ? modelConfig.apiUrl : "https://openrouter.ai/api/v1";
-            const providerLabel = modelConfig?.apiUrl ? 'Custom API' : 'OpenRouter';
-            
-            const client = new OpenAI({ baseURL, apiKey, defaultHeaders: { "HTTP-Referer": "https://smart.com", "X-Title": "Smart School" } });
-            
+            const client = new OpenAI({ baseURL, apiKey });
             try {
-                console.log(`[AI] 🚀 Attempting ${providerLabel} Model: ${modelName}`);
-                
                 const stream = await client.chat.completions.create({ model: modelName, messages, stream: true });
-                
-                console.log(`[AI] ✅ Connected to ${providerLabel}: ${modelName}`);
                 recordUsage(modelName, PROVIDERS.OPENROUTER);
-                
                 let fullText = '';
                 for await (const chunk of stream) {
                     const text = chunk.choices[0]?.delta?.content || '';
                     if (text) {
                         fullText += text;
                         res.write(`data: ${JSON.stringify({ text: text })}\n\n`);
-                        if (res.flush) res.flush();
                     }
                 }
                 return fullText;
-            } catch (e) { 
-                console.warn(`[AI] ⚠️ ${providerLabel} ${modelName} Error: ${e.message}`);
-                if (isQuotaError(e)) {
-                    break; 
-                }
-            }
+            } catch (e) { if (isQuotaError(e)) break; }
         }
     }
-    throw new Error("OpenRouter/Custom stream failed (All models exhausted)");
+    throw new Error("OpenRouter exhausted");
 }
 
 async function streamGemma(baseParams, res) {
     const { GoogleGenAI } = await import("@google/genai");
     const models = ['gemma-3-27b-it', 'gemma-3-12b-it'];
     const keys = await getKeyPool('gemini'); 
-    if (keys.length === 0) throw new Error("No keys for Gemma");
-    
     for (const apiKey of keys) {
         const client = new GoogleGenAI({ apiKey });
         for (const modelName of models) {
             try {
-                console.log(`[AI] 🚀 Attempting Gemma Model: ${modelName}`);
                 const result = await client.models.generateContentStream({ ...baseParams, model: modelName });
-                
                 let hasStarted = false;
                 let fullText = "";
                 for await (const chunk of result) {
-                    if (!hasStarted) {
-                        console.log(`[AI] ✅ Connected to Gemma: ${modelName}`);
-                        recordUsage(modelName, PROVIDERS.GEMMA);
-                        hasStarted = true;
-                    }
+                    if (!hasStarted) { recordUsage(modelName, PROVIDERS.GEMMA); hasStarted = true; }
                     if (chunk.text) {
                         fullText += chunk.text;
                         res.write(`data: ${JSON.stringify({ text: chunk.text })}\n\n`);
-                        if (res.flush) res.flush();
                     }
                 }
                 return fullText;
-            } catch (e) { 
-                console.warn(`[AI] ⚠️ Gemma ${modelName} Error: ${e.message}`);
-                if (isQuotaError(e)) continue; 
-            }
+            } catch (e) { if (isQuotaError(e)) continue; }
         }
     }
-    throw new Error("Gemma stream failed");
+    throw new Error("Gemma failed");
 }
 
 async function streamContentWithSmartFallback(baseParams, res) {
-    let hasAudio = false;
-    const contentsArray = Array.isArray(baseParams.contents) ? baseParams.contents : [baseParams.contents];
-    
-    contentsArray.forEach(c => {
-        if (c && c.parts) {
-            c.parts.forEach(p => { if (p.inlineData && p.inlineData.mimeType.startsWith('audio/')) hasAudio = true; });
-        }
-    });
-    
-    // FETCH CONFIG AND SET PROVIDER ORDER
     const config = await ConfigModel.findOne({ key: 'main' });
-    const configuredOrder = config?.aiProviderOrder && config.aiProviderOrder.length > 0 
-        ? config.aiProviderOrder 
-        : [PROVIDERS.GEMINI, PROVIDERS.OPENROUTER, PROVIDERS.GEMMA];
-
-    const runtimeSet = new Set(runtimeProviderOrder);
-    if (runtimeProviderOrder.length === 0 || runtimeProviderOrder.length !== configuredOrder.length || !configuredOrder.every(p => runtimeSet.has(p))) {
-        runtimeProviderOrder = [...configuredOrder];
-    }
+    const configuredOrder = config?.aiProviderOrder && config.aiProviderOrder.length > 0 ? config.aiProviderOrder : [PROVIDERS.GEMINI, PROVIDERS.OPENROUTER, PROVIDERS.GEMMA];
+    if (runtimeProviderOrder.length === 0) runtimeProviderOrder = [...configuredOrder];
 
     let finalError = null;
-    
     for (const provider of runtimeProviderOrder) {
         try {
-            console.log(`[AI] 👉 Trying Provider: ${provider}...`);
-            
-            // Note: Unlike before, we allow Audio to pass to OpenRouter/Gemma, 
-            // because those functions now have internal Fallback STT logic.
-            
             if (provider === PROVIDERS.GEMINI) return await streamGemini(baseParams, res);
             else if (provider === PROVIDERS.OPENROUTER) return await streamOpenRouter(baseParams, res);
             else if (provider === PROVIDERS.GEMMA) return await streamGemma(baseParams, res);
-            
         } catch (e) {
-            console.error(`[AI] ❌ Provider ${provider} Failed: ${e.message}`);
             finalError = e;
-            
-            if (isQuotaError(e)) { 
-                console.log(`[AI] 📉 Quota/Rate Limit detected. Switching provider...`);
-                deprioritizeProvider(provider); 
-                continue; 
-            }
+            if (isQuotaError(e)) { deprioritizeProvider(provider); continue; }
             continue; 
         }
     }
-    
-    console.error(`[AI] 💀 All providers failed.`);
-    throw finalError || new Error('All streaming models unavailable.');
+    throw finalError || new Error('All failed');
 }
 
 const checkAIAccess = async (req, res, next) => {
@@ -347,7 +252,6 @@ const checkAIAccess = async (req, res, next) => {
     next();
 };
 
-// NEW: Endpoint to provide a temporary key for Client-Side Live API
 router.get('/live-access', checkAIAccess, async (req, res) => {
     try {
         const keys = await getKeyPool('gemini');
@@ -374,11 +278,7 @@ router.get('/stats', checkAIAccess, async (req, res) => {
     } catch (e) { res.status(500).json({ error: e.message }); }
 });
 
-router.post('/reset-pool', checkAIAccess, (req, res) => { 
-    runtimeProviderOrder = []; 
-    console.log('[AI] 🔄 Provider priority pool reset.');
-    res.json({ success: true }); 
-});
+router.post('/reset-pool', checkAIAccess, (req, res) => { runtimeProviderOrder = []; res.json({ success: true }); });
 
 router.post('/chat', checkAIAccess, async (req, res) => {
     const { text, audio, history } = req.body;
@@ -399,7 +299,7 @@ router.post('/chat', checkAIAccess, async (req, res) => {
 
         const answerText = await streamContentWithSmartFallback({
             contents: fullContents, 
-            config: { systemInstruction: "你是一位友善、耐心且知识渊博的中小学AI助教。请用简洁、鼓励性的语言回答学生的问题。回复支持 Markdown 格式。" }
+            config: { systemInstruction: "你是一位友善且知识渊博的中小学AI助教。请用简洁、鼓励性的语言回答。支持 Markdown。" }
         }, res);
 
         if (answerText) {
@@ -425,12 +325,10 @@ router.post('/chat', checkAIAccess, async (req, res) => {
         }
         res.write('data: [DONE]\n\n'); res.end();
     } catch (e) {
-        console.error("[AI Chat Route Error]", e);
         res.write(`data: ${JSON.stringify({ error: true, message: e.message })}\n\n`); res.end();
     }
 });
 
-// STREAMING ASSESSMENT ENDPOINT
 router.post('/evaluate', checkAIAccess, async (req, res) => {
     const { question, audio, image, images } = req.body;
     res.setHeader('Content-Type', 'text/event-stream');
@@ -440,42 +338,22 @@ router.post('/evaluate', checkAIAccess, async (req, res) => {
 
     try {
         res.write(`data: ${JSON.stringify({ status: 'analyzing' })}\n\n`);
+        const evalParts = [{ text: `请对学生的回答评分。题目：${question}。` }];
+        if (audio) evalParts.push({ inlineData: { mimeType: 'audio/webm', data: audio } });
+        if (images && Array.isArray(images)) images.forEach(img => { if(img) evalParts.push({ inlineData: { mimeType: 'image/jpeg', data: img } }); });
+        else if (image) evalParts.push({ inlineData: { mimeType: 'image/jpeg', data: image } });
 
-        const evalParts = [{ text: `请作为一名严谨的老师，对学生的回答进行评分。题目是：${question}。` }];
-        if (audio) {
-            evalParts.push({ text: "学生的回答在音频中。" });
-            evalParts.push({ inlineData: { mimeType: 'audio/webm', data: audio } });
-        }
-        
-        if (images && Array.isArray(images) && images.length > 0) {
-            evalParts.push({ text: "学生的回答写在以下图片中，请识别所有图片中的文字内容并进行批改：" });
-            images.forEach(img => {
-                if(img) evalParts.push({ inlineData: { mimeType: 'image/jpeg', data: img } });
-            });
-        } else if (image) {
-            evalParts.push({ text: "学生的回答写在图片中，请识别图片中的文字内容并进行批改。" });
-            evalParts.push({ inlineData: { mimeType: 'image/jpeg', data: image } });
-        }
-
-        evalParts.push({ text: `请分析：1. 内容准确性 2. 表达/书写规范。
-        必须严格按照以下格式输出(不要使用Markdown代码块包裹):
-        
+        evalParts.push({ text: `格式:
         ## Transcription
-        (在此处输出识别到的学生回答内容，如果是图片则为识别的文字)
-        
+        (内容)
         ## Feedback
-        (在此处输出简短的鼓励性评语和建议)
-        
+        (评语)
         ## Score
-        (在此处仅输出一个0-100的数字)` });
-
-        const fullText = await streamContentWithSmartFallback({
-            contents: [{ role: 'user', parts: evalParts }],
-        }, res);
+        (0-100数字)` });
 
+        const fullText = await streamContentWithSmartFallback({ contents: [{ role: 'user', parts: evalParts }] }, res);
         const feedbackMatch = fullText.match(/## Feedback\s+([\s\S]*?)(?=## Score|$)/i);
         const feedbackText = feedbackMatch ? feedbackMatch[1].trim() : "";
-
         if (feedbackText) {
             res.write(`data: ${JSON.stringify({ status: 'tts' })}\n\n`);
             try {
@@ -498,14 +376,9 @@ router.post('/evaluate', checkAIAccess, async (req, res) => {
                 else res.write(`data: ${JSON.stringify({ ttsSkipped: true })}\n\n`);
             } catch (ttsErr) { res.write(`data: ${JSON.stringify({ ttsSkipped: true })}\n\n`); }
         }
-
-        res.write('data: [DONE]\n\n');
-        res.end();
-
+        res.write('data: [DONE]\n\n'); res.end();
     } catch (e) {
-        console.error("AI Eval Error:", e);
-        res.write(`data: ${JSON.stringify({ error: true, message: e.message || "Evaluation failed" })}\n\n`);
-        res.end();
+        res.write(`data: ${JSON.stringify({ error: true, message: e.message })}\n\n`); res.end();
     }
 });
 

components/ai/AssessmentPanel.tsx

@@ -16,17 +16,8 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
     const [isAssessmentRecording, setIsAssessmentRecording] = useState(false);
     const [isWebSpeechListening, setIsWebSpeechListening] = useState(false);
     const [assessmentStatus, setAssessmentStatus] = useState<'IDLE' | 'UPLOADING' | 'ANALYZING' | 'TTS'>('IDLE');
-    
-    // For Web Speech Text Accumulation
     const [recognizedText, setRecognizedText] = useState('');
-
-    const [streamedAssessment, setStreamedAssessment] = useState<{
-        transcription: string;
-        feedback: string;
-        score: number | null;
-        audio?: string;
-    }>({ transcription: '', feedback: '', score: null });
-
+    const [streamedAssessment, setStreamedAssessment] = useState<{ transcription: string; feedback: string; score: number | null; audio?: string; }>({ transcription: '', feedback: '', score: null });
     const [toast, setToast] = useState<ToastState>({ show: false, message: '', type: 'success' });
 
     const mediaRecorderRef = useRef<MediaRecorder | null>(null);
@@ -35,14 +26,12 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
     const currentSourceRef = useRef<AudioBufferSourceNode | null>(null);
     const recognitionRef = useRef<any>(null);
 
-    // Initialize AudioContext
     useEffect(() => {
         // @ts-ignore
         const AudioCtor = window.AudioContext || window.webkitAudioContext;
         audioContextRef.current = new AudioCtor();
         return () => {
             stopPlayback();
-            window.speechSynthesis.cancel();
             if (recognitionRef.current) recognitionRef.current.abort();
         };
     }, []);
@@ -58,13 +47,8 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
     const speakWithBrowser = (text: string) => {
         if (!text) return;
         stopPlayback();
-        const cleanText = cleanTextForTTS(text);
-        const utterance = new SpeechSynthesisUtterance(cleanText);
+        const utterance = new SpeechSynthesisUtterance(cleanTextForTTS(text));
         utterance.lang = 'zh-CN';
-        utterance.rate = 1.0; 
-        const voices = window.speechSynthesis.getVoices();
-        const zhVoice = voices.find(v => v.lang === 'zh-CN' && !v.name.includes('Hong Kong') && !v.name.includes('Taiwan'));
-        if (zhVoice) utterance.voice = zhVoice;
         window.speechSynthesis.speak(utterance);
     };
 
@@ -76,9 +60,6 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
                 const AudioCtor = window.AudioContext || window.webkitAudioContext;
                 audioContextRef.current = new AudioCtor();
             }
-            if (audioContextRef.current?.state === 'suspended') {
-                await audioContextRef.current.resume();
-            }
             const bytes = base64ToUint8Array(base64Audio);
             const audioBuffer = decodePCM(bytes, audioContextRef.current!);
             const source = audioContextRef.current!.createBufferSource();
@@ -86,13 +67,11 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
             source.connect(audioContextRef.current!.destination);
             source.start(0);
             currentSourceRef.current = source;
-        } catch (e) {
-            console.error("Audio playback error", e);
-            setToast({ show: true, message: '语音播放失败', type: 'error' });
-        }
+        } catch (e) { console.error("Audio playback error", e); }
     };
 
     const startRecording = async () => {
+        console.log("[Assessment] Starting Recording...");
         // @ts-ignore
         const SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
         if (SpeechRecognition) {
@@ -101,6 +80,7 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
                 const recognition = new SpeechRecognition();
                 recognition.lang = 'zh-CN';
                 recognition.interimResults = true;
+                recognition.continuous = true;
                 
                 recognition.onstart = () => {
                     setIsWebSpeechListening(true);
@@ -113,36 +93,21 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
                     for (let i = event.resultIndex; i < event.results.length; ++i) {
                         if (event.results[i].isFinal) final += event.results[i][0].transcript;
                     }
-                    if (final) {
-                        setRecognizedText(prev => prev + final);
-                        // For assessment, we usually want full context, so keep appending
-                    }
-                };
-
-                recognition.onend = () => {
-                    // Logic to handle stop: if it stopped but user is still holding button?
-                    // Typically Web Speech stops on silence. We rely on stopRecording() to finalize.
-                    // If it stopped automatically, we might just restart or consider it done.
-                    // For now, let's update state but not trigger submit yet unless triggered by user.
-                    setIsWebSpeechListening(false);
+                    if (final) setRecognizedText(prev => prev + final);
                 };
 
                 recognition.onerror = (e: any) => {
-                    console.warn("Web Speech API Error:", e.error);
-                    recognition.stop();
-                    if (e.error !== 'aborted') {
-                        startAudioRecordingFallback();
-                    } else {
-                        setIsWebSpeechListening(false);
-                        setIsAssessmentRecording(false);
-                    }
+                    console.warn("[Assessment] Web Speech Error:", e.error);
+                    stopRecording();
+                    if (e.error !== 'aborted') startAudioRecordingFallback();
                 };
 
+                recognition.onend = () => setIsWebSpeechListening(false);
                 recognitionRef.current = recognition;
                 recognition.start();
                 return;
             } catch (e) {
-                console.warn("Web Speech Init Failed", e);
+                console.error("[Assessment] Web Speech Failed", e);
                 startAudioRecordingFallback();
             }
         } else {
@@ -156,13 +121,15 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
             const mediaRecorder = new MediaRecorder(stream);
             mediaRecorderRef.current = mediaRecorder;
             audioChunksRef.current = [];
-
             mediaRecorder.ondataavailable = (event) => {
-                if (event.data.size > 0) {
-                    audioChunksRef.current.push(event.data);
-                }
+                if (event.data.size > 0) audioChunksRef.current.push(event.data);
+            };
+            mediaRecorder.onstop = async () => {
+                const audioBlob = new Blob(audioChunksRef.current, { type: 'audio/webm' });
+                const base64 = await blobToBase64(audioBlob);
+                handleAssessmentStreamingSubmit({ audio: base64 });
+                stream.getTracks().forEach(track => track.stop());
             };
-
             mediaRecorder.start();
             setIsAssessmentRecording(true);
             setIsWebSpeechListening(false);
@@ -173,35 +140,11 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
 
     const stopRecording = () => {
         setIsAssessmentRecording(false);
-        
-        if (isWebSpeechListening) {
-            if (recognitionRef.current) recognitionRef.current.stop();
-            // Wait slightly for final results then submit
-            setTimeout(() => {
-                if (recognizedText) {
-                    // Send text as prompt supplement since we don't have audio
-                    // Construct a virtual "audio" payload which is just text instructions for the backend to handle?
-                    // No, backend expects audio or image. We should modify backend to accept text for assessment too.
-                    // Actually, simpler: backend takes 'audio' or 'image'.
-                    // Let's modify frontend to send `audio` as undefined but include text in prompt.
-                    
-                    // Hack: Prepend "学生回答内容：..." to the question if we have text but no audio.
-                    // Or, add a `studentAnswerText` field to the API. 
-                    // Let's modify the handleAssessmentStreamingSubmit to take text.
-                    handleAssessmentStreamingSubmit({ text: recognizedText });
-                } else {
-                    setToast({ show: true, message: '未检测到语音输入', type: 'error' });
-                }
-            }, 500);
+        if (isWebSpeechListening && recognitionRef.current) {
+            recognitionRef.current.stop();
+            setTimeout(() => { if (recognizedText) handleAssessmentStreamingSubmit({ text: recognizedText }); }, 500);
         } else if (mediaRecorderRef.current) {
             mediaRecorderRef.current.stop();
-            
-            mediaRecorderRef.current.onstop = async () => {
-                const audioBlob = new Blob(audioChunksRef.current, { type: 'audio/webm' });
-                const base64 = await blobToBase64(audioBlob);
-                handleAssessmentStreamingSubmit({ audio: base64 });
-                mediaRecorderRef.current?.stream.getTracks().forEach(track => track.stop());
-            };
         }
     };
 
@@ -211,11 +154,8 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
         stopPlayback();
 
         try {
-            // If we have text from Web Speech, augment the question
             let finalQuestion = assessmentTopic;
-            if (text) {
-                finalQuestion += `\n\n学生口述回答内容：${text}\n（请基于此文本进行评分，忽略音频缺失）`;
-            }
+            if (text) finalQuestion += `\n\n学生���述回答内容：${text}\n（请基于此文本进行评分）`;
 
             const response = await fetch('/api/ai/evaluate', {
                 method: 'POST',
@@ -229,11 +169,9 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
             });
 
             if (!response.ok) throw new Error(response.statusText);
-            if (!response.body) throw new Error('No response body');
-
             setAssessmentStatus('ANALYZING');
 
-            const reader = response.body.getReader();
+            const reader = response.body!.getReader();
             const decoder = new TextDecoder();
             let accumulatedRaw = '';
             let buffer = '';
@@ -241,7 +179,6 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
             while (true) {
                 const { done, value } = await reader.read();
                 if (done) break;
-                
                 buffer += decoder.decode(value, { stream: true });
                 const parts = buffer.split('\n\n');
                 buffer = parts.pop() || ''; 
@@ -250,72 +187,38 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
                     if (line.startsWith('data: ')) {
                         const jsonStr = line.replace('data: ', '').trim();
                         if (jsonStr === '[DONE]') break;
-                        
                         try {
                             const data = JSON.parse(jsonStr);
-                            
-                            if (data.status) {
-                                if (data.status === 'analyzing') setAssessmentStatus('ANALYZING');
-                                if (data.status === 'tts') setAssessmentStatus('TTS');
-                            }
-
+                            if (data.status === 'tts') setAssessmentStatus('TTS');
                             if (data.text) {
                                 accumulatedRaw += data.text;
-                                const transcriptionMatch = accumulatedRaw.match(/## Transcription\s+([\s\S]*?)(?=## Feedback|$)/i);
-                                const feedbackMatch = accumulatedRaw.match(/## Feedback\s+([\s\S]*?)(?=## Score|$)/i);
-                                const scoreMatch = accumulatedRaw.match(/## Score\s+(\d+)/i);
-
+                                const trMatch = accumulatedRaw.match(/## Transcription\s+([\s\S]*?)(?=## Feedback|$)/i);
+                                const fbMatch = accumulatedRaw.match(/## Feedback\s+([\s\S]*?)(?=## Score|$)/i);
+                                const scMatch = accumulatedRaw.match(/## Score\s+(\d+)/i);
                                 setStreamedAssessment(prev => ({
                                     ...prev,
-                                    transcription: transcriptionMatch ? transcriptionMatch[1].trim() : (text || prev.transcription), // Use Web Speech text if avail
-                                    feedback: feedbackMatch ? feedbackMatch[1].trim() : prev.feedback,
-                                    score: scoreMatch ? parseInt(scoreMatch[1]) : prev.score
+                                    transcription: trMatch ? trMatch[1].trim() : (text || prev.transcription),
+                                    feedback: fbMatch ? fbMatch[1].trim() : prev.feedback,
+                                    score: scMatch ? parseInt(scMatch[1]) : prev.score
                                 }));
                             }
-
                             if (data.audio) {
                                 setStreamedAssessment(prev => ({ ...prev, audio: data.audio }));
                                 playPCMAudio(data.audio);
                             }
-                            
                             if (data.ttsSkipped) {
                                 const fb = streamedAssessment.feedback || accumulatedRaw.match(/## Feedback\s+([\s\S]*?)(?=## Score|$)/i)?.[1] || '';
                                 if (fb) speakWithBrowser(fb);
                             }
-
-                            if (data.error) {
-                                setToast({ show: true, message: data.message || '评分出错', type: 'error' });
-                            }
-
+                            if (data.error) setToast({ show: true, message: data.message || '评分出错', type: 'error' });
                         } catch (e) {}
                     }
                 }
             }
             setAssessmentStatus('IDLE');
-
         } catch (error: any) {
-            console.error("Assessment error", error);
-            setToast({ show: true, message: '评分失败: ' + error.message, type: 'error' });
-            setAssessmentStatus('IDLE');
-        }
-    };
-
-    const handleImageUpload = (e: React.ChangeEvent<HTMLInputElement>) => {
-        if (e.target.files && e.target.files.length > 0) {
-            setSelectedImages(prev => [...prev, ...Array.from(e.target.files!)]);
-        }
-    };
-
-    const confirmImageSubmission = async () => {
-        if (selectedImages.length === 0) return;
-        setAssessmentStatus('UPLOADING');
-        try {
-            const base64Promises = selectedImages.map(file => compressImage(file));
-            const base64Images = await Promise.all(base64Promises);
-            handleAssessmentStreamingSubmit({ images: base64Images });
-        } catch(e) {
+            setToast({ show: true, message: '评分失败', type: 'error' });
             setAssessmentStatus('IDLE');
-            setToast({ show: true, message: '图片压缩上传失败', type: 'error' });
         }
     };
 
@@ -325,7 +228,6 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
             <button onClick={stopPlayback} className="absolute top-4 right-4 z-50 bg-white/80 backdrop-blur p-2 rounded-full shadow-md text-red-500 hover:bg-white border border-gray-200" title="停止播放"><StopIcon size={20}/></button>
 
             <div className="max-w-3xl mx-auto space-y-6">
-                {/* Topic Card */}
                 <div className="bg-white p-6 rounded-2xl border border-purple-100 shadow-sm">
                     <h3 className="text-lg font-bold text-gray-800 mb-2 flex items-center justify-between">
                         <span className="flex items-center"><Brain className="mr-2 text-purple-600"/> 今日测评题目</span>
@@ -335,111 +237,48 @@ export const AssessmentPanel: React.FC<AssessmentPanelProps> = ({ currentUser })
                         </div>
                     </h3>
                     <textarea className="w-full bg-purple-50/50 border border-purple-100 rounded-xl p-4 text-gray-700 font-medium text-lg resize-none focus:ring-2 focus:ring-purple-200 outline-none" value={assessmentTopic} onChange={e => setAssessmentTopic(e.target.value)} rows={3}/>
-                    
                     <div className="mt-6 flex justify-center">
                         {assessmentMode === 'audio' ? (
-                            <button 
-                                onMouseDown={startRecording} onMouseUp={stopRecording} onTouchStart={startRecording} onTouchEnd={stopRecording}
-                                disabled={assessmentStatus !== 'IDLE'}
-                                className={`px-8 py-4 rounded-full font-bold text-white flex items-center gap-3 shadow-lg transition-all ${isAssessmentRecording ? 'bg-red-500 scale-105' : 'bg-gradient-to-r from-purple-600 to-indigo-600 hover:shadow-purple-200 hover:scale-105 disabled:opacity-50'}`}
-                            >
+                            <button onMouseDown={startRecording} onMouseUp={stopRecording} onTouchStart={startRecording} onTouchEnd={stopRecording} disabled={assessmentStatus !== 'IDLE'} className={`px-8 py-4 rounded-full font-bold text-white flex items-center gap-3 shadow-lg transition-all ${isAssessmentRecording ? 'bg-red-500 scale-105' : 'bg-gradient-to-r from-purple-600 to-indigo-600 hover:shadow-purple-200 hover:scale-105 disabled:opacity-50'}`}>
                                 {assessmentStatus !== 'IDLE' ? <Loader2 className="animate-spin"/> : (isAssessmentRecording ? <StopCircle/> : <Mic/>)}
-                                {assessmentStatus === 'UPLOADING' ? '上传中...' : assessmentStatus === 'ANALYZING' ? 'AI 正在分析...' : assessmentStatus === 'TTS' ? '生成语音...' : isAssessmentRecording ? (isWebSpeechListening ? '正在识别...' : '松开结束录音') : '按住开始回答'}
+                                {assessmentStatus === 'UPLOADING' ? '上传中...' : assessmentStatus === 'ANALYZING' ? 'AI 分析中...' : assessmentStatus === 'TTS' ? '生成语音...' : isAssessmentRecording ? '正在识别...' : '长按开始回答'}
                             </button>
                         ) : (
                             <div className="w-full">
                                 <div className="relative border-2 border-dashed border-purple-200 rounded-xl p-4 text-center hover:bg-purple-50 transition-colors cursor-pointer min-h-[160px] flex flex-col items-center justify-center">
-                                    <input 
-                                        type="file" 
-                                        accept="image/*" 
-                                        multiple
-                                        className="absolute inset-0 opacity-0 cursor-pointer w-full h-full z-10" 
-                                        onChange={handleImageUpload}
-                                        onClick={(e) => (e.currentTarget.value = '')}
-                                    />
-                                    {selectedImages.length === 0 ? (
-                                        <>
-                                            <ImageIcon className="mx-auto text-purple-300 mb-2" size={40}/>
-                                            <p className="text-purple-600 font-bold">点击上传作业图片</p>
-                                            <p className="text-xs text-gray-400">支持批量上传 • 自动压缩处理</p>
-                                        </>
-                                    ) : (
-                                        <div className="z-0 w-full pointer-events-none opacity-50 flex items-center justify-center">
-                                            <Plus className="text-purple-300" size={40}/>
-                                            <span className="text-purple-400 font-bold ml-2">继续添加图片</span>
-                                        </div>
-                                    )}
+                                    <input type="file" accept="image/*" multiple className="absolute inset-0 opacity-0 cursor-pointer w-full h-full z-10" onChange={(e) => { if (e.target.files) setSelectedImages(prev => [...prev, ...Array.from(e.target.files!)]); }} onClick={(e) => (e.currentTarget.value = '')} />
+                                    {selectedImages.length === 0 ? (<><ImageIcon className="mx-auto text-purple-300 mb-2" size={40}/><p className="text-purple-600 font-bold">点击上传作业图片</p></>) : (<div className="z-0 w-full pointer-events-none opacity-50 flex items-center justify-center"><Plus className="text-purple-300" size={40}/><span className="text-purple-400 font-bold ml-2">继续添加</span></div>)}
                                 </div>
-
                                 {selectedImages.length > 0 && (
                                     <div className="mt-4 grid grid-cols-3 sm:grid-cols-4 gap-3 animate-in fade-in">
-                                        {selectedImages.map((file, idx) => (
-                                            <div key={idx} className="relative group aspect-square">
-                                                <img src={URL.createObjectURL(file)} className="w-full h-full object-cover rounded-lg shadow-sm border border-gray-200" />
-                                                <button 
-                                                    onClick={() => setSelectedImages(prev => prev.filter((_, i) => i !== idx))} 
-                                                    className="absolute -top-2 -right-2 bg-red-500 text-white rounded-full p-1 shadow-md hover:bg-red-600 transition-colors z-20 scale-90 hover:scale-100"
-                                                >
-                                                    <X size={14}/>
-                                                </button>
-                                            </div>
-                                        ))}
+                                        {selectedImages.map((file, idx) => (<div key={idx} className="relative group aspect-square"><img src={URL.createObjectURL(file)} className="w-full h-full object-cover rounded-lg shadow-sm border border-gray-200" /><button onClick={() => setSelectedImages(prev => prev.filter((_, i) => i !== idx))} className="absolute -top-2 -right-2 bg-red-500 text-white rounded-full p-1 shadow-md hover:bg-red-600 transition-colors z-20 scale-90 hover:scale-100"><X size={14}/></button></div>))}
+                                        <button onClick={async () => { setAssessmentStatus('UPLOADING'); try { const images = await Promise.all(selectedImages.map(f => compressImage(f))); handleAssessmentStreamingSubmit({ images }); } catch(e) { setAssessmentStatus('IDLE'); } }} disabled={assessmentStatus !== 'IDLE'} className="mt-6 w-full px-8 py-3 bg-purple-600 text-white rounded-lg font-bold hover:bg-purple-700 flex items-center justify-center gap-2 shadow-md transition-all">开始批改</button>
                                     </div>
                                 )}
-
-                                {selectedImages.length > 0 && (
-                                    <button 
-                                        onClick={confirmImageSubmission} 
-                                        disabled={assessmentStatus !== 'IDLE'} 
-                                        className="mt-6 w-full px-8 py-3 bg-purple-600 text-white rounded-lg font-bold hover:bg-purple-700 flex items-center justify-center gap-2 shadow-md transition-all"
-                                    >
-                                        {assessmentStatus !== 'IDLE' ? <Loader2 className="animate-spin" size={18}/> : <CheckCircle size={18}/>}
-                                        {assessmentStatus === 'UPLOADING' ? '压缩上传中...' : assessmentStatus === 'ANALYZING' ? 'AI 正在分析...' : assessmentStatus === 'TTS' ? '生成语音...' : `开始批改 (${selectedImages.length}张)`}
-                                    </button>
-                                )}
                             </div>
                         )}
                     </div>
                 </div>
-
-                {/* Streamed Result Card */}
                 {(streamedAssessment.transcription || streamedAssessment.feedback || streamedAssessment.score !== null) && (
                     <div className="bg-white p-6 rounded-2xl border border-gray-200 shadow-lg animate-in slide-in-from-bottom-4">
                         <div className="flex items-center justify-between border-b border-gray-100 pb-4 mb-4">
                             <div className="flex items-center gap-2">
                                 <h3 className="font-bold text-xl text-gray-800">测评报告</h3>
-                                {assessmentStatus !== 'IDLE' && (
-                                    <div className="flex items-center gap-1 text-xs px-2 py-1 bg-purple-50 text-purple-600 rounded-full animate-pulse">
-                                        <Zap size={12}/>
-                                        {assessmentStatus === 'ANALYZING' ? '正在智能分析内容...' : assessmentStatus === 'TTS' ? '正在生成语音点评...' : '处理中...'}
-                                    </div>
-                                )}
+                                {assessmentStatus !== 'IDLE' && <div className="flex items-center gap-1 text-xs px-2 py-1 bg-purple-50 text-purple-600 rounded-full animate-pulse"><Zap size={12}/>正在处理...</div>}
                             </div>
                             <div className="flex items-center gap-4">
-                                {streamedAssessment.audio && (
-                                    <button onClick={() => playPCMAudio(streamedAssessment.audio!)} className="flex items-center gap-1 text-sm bg-purple-100 text-purple-700 px-3 py-1 rounded-full hover:bg-purple-200 animate-in fade-in">
-                                        <Volume2 size={16}/> 听AI点评
-                                    </button>
-                                )}
-                                {streamedAssessment.score !== null ? (
-                                    <div className={`text-3xl font-black ${streamedAssessment.score >= 80 ? 'text-green-500' : streamedAssessment.score >= 60 ? 'text-yellow-500' : 'text-red-500'}`}>
-                                        {streamedAssessment.score}<span className="text-sm text-gray-400 ml-1">分</span>
-                                    </div>
-                                ) : (
-                                    <div className="text-sm text-gray-400 italic">评分中...</div>
-                                )}
+                                {streamedAssessment.audio && <button onClick={() => playPCMAudio(streamedAssessment.audio!)} className="flex items-center gap-1 text-sm bg-purple-100 text-purple-700 px-3 py-1 rounded-full hover:bg-purple-200"><Volume2 size={16}/> 听AI点评</button>}
+                                {streamedAssessment.score !== null && <div className={`text-3xl font-black ${streamedAssessment.score >= 80 ? 'text-green-500' : streamedAssessment.score >= 60 ? 'text-yellow-500' : 'text-red-500'}`}>{streamedAssessment.score}<span className="text-sm text-gray-400 ml-1">分</span></div>}
                             </div>
                         </div>
                         <div className="space-y-4">
                             <div className="bg-gray-50 p-4 rounded-xl">
                                 <p className="text-xs font-bold text-gray-500 uppercase mb-1">AI 识别内容</p>
-                                <p className="text-gray-700 leading-relaxed text-sm whitespace-pre-wrap">{streamedAssessment.transcription || <span className="text-gray-400">正在识别...</span>}</p>
+                                <p className="text-gray-700 leading-relaxed text-sm whitespace-pre-wrap">{streamedAssessment.transcription || '正在识别...'}</p>
                             </div>
                             <div>
                                 <p className="text-xs font-bold text-gray-500 uppercase mb-2">AI 点评建议</p>
-                                <div className="p-4 bg-purple-50 text-purple-900 rounded-xl border border-purple-100 text-sm leading-relaxed whitespace-pre-wrap">
-                                    {streamedAssessment.feedback || <span className="text-purple-300">AI 正在思考评语...</span>}
-                                </div>
+                                <div className="p-4 bg-purple-50 text-purple-900 rounded-xl border border-purple-100 text-sm leading-relaxed whitespace-pre-wrap">{streamedAssessment.feedback || 'AI 思考中...'}</div>
                             </div>
                         </div>
                     </div>

components/ai/ChatPanel.tsx

@@ -1,7 +1,7 @@
 
 import React, { useState, useRef, useEffect } from 'react';
 import { AIChatMessage, User } from '../../types';
-import { Bot, Mic, Square, Volume2, Send, Sparkles, Loader2, StopCircle, Trash2 } from 'lucide-react';
+import { Bot, Mic, Volume2, Send, Sparkles, Loader2, StopCircle, Trash2 } from 'lucide-react';
 import ReactMarkdown from 'react-markdown';
 import remarkGfm from 'remark-gfm';
 import { blobToBase64, base64ToUint8Array, decodePCM, cleanTextForTTS } from '../../utils/mediaHelpers';
@@ -22,16 +22,11 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
                 timestamp: Date.now() 
             }];
         } catch (e) {
-            return [{ 
-                id: 'welcome', 
-                role: 'model', 
-                text: '你好！我是你的 AI 智能助教。有什么可以帮你的吗？', 
-                timestamp: Date.now() 
-            }];
+            return [{ id: 'welcome', role: 'model', text: '你好！我是你的 AI 智能助教。', timestamp: Date.now() }];
         }
     });
+    
     const [textInput, setTextInput] = useState('');
-    const [inputMode, setInputMode] = useState<'text' | 'audio'>('text');
     const [isChatProcessing, setIsChatProcessing] = useState(false);
     const [isChatRecording, setIsChatRecording] = useState(false);
     const [isWebSpeechListening, setIsWebSpeechListening] = useState(false);
@@ -43,32 +38,18 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
     const currentSourceRef = useRef<AudioBufferSourceNode | null>(null);
     const messagesEndRef = useRef<HTMLDivElement>(null);
     const recognitionRef = useRef<any>(null);
+    const inputRef = useRef<HTMLInputElement>(null);
 
-    // Initialize AudioContext
     useEffect(() => {
         // @ts-ignore
         const AudioCtor = window.AudioContext || window.webkitAudioContext;
         audioContextRef.current = new AudioCtor();
         return () => {
             stopPlayback();
-            window.speechSynthesis.cancel();
             if (recognitionRef.current) recognitionRef.current.abort();
         };
     }, []);
 
-    // Persist messages
-    useEffect(() => {
-        try {
-            const MAX_COUNT = 50;
-            const welcome = messages.find(m => m.id === 'welcome');
-            const others = messages.filter(m => m.id !== 'welcome');
-            const recent = others.slice(-MAX_COUNT);
-            const messagesToSave = (welcome ? [welcome] : []).concat(recent);
-            localStorage.setItem('ai_chat_history', JSON.stringify(messagesToSave));
-        } catch (e) {}
-    }, [messages]);
-
-    // Scroll to bottom
     useEffect(() => {
         messagesEndRef.current?.scrollIntoView({ behavior: isChatProcessing ? 'auto' : 'smooth', block: 'end' });
     }, [messages, isChatProcessing]);
@@ -84,13 +65,8 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
     const speakWithBrowser = (text: string) => {
         if (!text) return;
         stopPlayback();
-        const cleanText = cleanTextForTTS(text);
-        const utterance = new SpeechSynthesisUtterance(cleanText);
+        const utterance = new SpeechSynthesisUtterance(cleanTextForTTS(text));
         utterance.lang = 'zh-CN';
-        utterance.rate = 1.0; 
-        const voices = window.speechSynthesis.getVoices();
-        const zhVoice = voices.find(v => v.lang === 'zh-CN' && !v.name.includes('Hong Kong') && !v.name.includes('Taiwan'));
-        if (zhVoice) utterance.voice = zhVoice;
         window.speechSynthesis.speak(utterance);
     };
 
@@ -114,61 +90,63 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
             currentSourceRef.current = source;
         } catch (e) {
             console.error("Audio playback error", e);
-            setToast({ show: true, message: '语音播放失败', type: 'error' });
         }
     };
 
-    const startRecording = async () => {
-        // Priority 1: Web Speech API
+    const startRecording = async (e?: React.MouseEvent | React.TouchEvent) => {
+        if (e) {
+            e.preventDefault();
+            e.stopPropagation();
+        }
+        
+        console.log("[Voice] Starting...");
         // @ts-ignore
         const SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
+        
         if (SpeechRecognition) {
             try {
                 if (recognitionRef.current) recognitionRef.current.abort();
+
                 const recognition = new SpeechRecognition();
                 recognition.lang = 'zh-CN';
                 recognition.interimResults = true;
+                recognition.continuous = true; 
                 
                 recognition.onstart = () => {
+                    console.log("[Voice] Web Speech Active");
                     setIsWebSpeechListening(true);
-                    setIsChatRecording(true); // Reuse UI state
+                    setIsChatRecording(true);
                 };
                 
                 recognition.onresult = (event: any) => {
-                    let final = '';
-                    let interim = '';
+                    let transcript = '';
                     for (let i = event.resultIndex; i < event.results.length; ++i) {
-                        if (event.results[i].isFinal) final += event.results[i][0].transcript;
-                        else interim += event.results[i][0].transcript;
+                        transcript += event.results[i][0].transcript;
                     }
-                    if (final || interim) {
-                        setTextInput(final || interim);
-                        setInputMode('text'); // Switch UI to text to show result
+                    if (transcript) {
+                        setTextInput(transcript);
                     }
                 };
 
+                recognition.onerror = (e: any) => {
+                    console.warn("[Voice] Web Speech Error:", e.error);
+                    if (e.error === 'not-allowed') {
+                        setToast({ show: true, message: '请允许麦克风访问', type: 'error' });
+                    }
+                    stopRecording();
+                };
+
                 recognition.onend = () => {
+                    console.log("[Voice] Web Speech Stopped");
                     setIsWebSpeechListening(false);
                     setIsChatRecording(false);
                 };
 
-                recognition.onerror = (e: any) => {
-                    console.warn("Web Speech API Error:", e.error);
-                    recognition.stop();
-                    // Fallback to audio recording if it was a permission/network issue not aborted by user
-                    if (e.error !== 'aborted') {
-                        startAudioRecordingFallback();
-                    } else {
-                        setIsWebSpeechListening(false);
-                        setIsChatRecording(false);
-                    }
-                };
-
                 recognitionRef.current = recognition;
                 recognition.start();
                 return;
             } catch (e) {
-                console.warn("Web Speech Init Failed, using fallback", e);
+                console.error("[Voice] Web Speech Failed, falling back", e);
                 startAudioRecordingFallback();
             }
         } else {
@@ -177,6 +155,7 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
     };
 
     const startAudioRecordingFallback = async () => {
+        console.log("[Voice] Using MediaRecorder Fallback");
         try {
             const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
             const mediaRecorder = new MediaRecorder(stream);
@@ -184,55 +163,51 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
             audioChunksRef.current = [];
 
             mediaRecorder.ondataavailable = (event) => {
-                if (event.data.size > 0) {
-                    audioChunksRef.current.push(event.data);
-                }
+                if (event.data.size > 0) audioChunksRef.current.push(event.data);
+            };
+
+            mediaRecorder.onstop = async () => {
+                const audioBlob = new Blob(audioChunksRef.current, { type: 'audio/webm' });
+                const base64 = await blobToBase64(audioBlob);
+                handleChatSubmit(undefined, base64);
+                stream.getTracks().forEach(track => track.stop());
             };
 
             mediaRecorder.start();
             setIsChatRecording(true);
             setIsWebSpeechListening(false);
         } catch (e) {
-            setToast({ show: true, message: '无法访问麦克风', type: 'error' });
+            setToast({ show: true, message: '麦克风不可用', type: 'error' });
         }
     };
 
     const stopRecording = () => {
-        if (isWebSpeechListening) {
-            if (recognitionRef.current) recognitionRef.current.stop();
-            // Do not auto submit, user sees text and clicks send
+        console.log("[Voice] Stopping...");
+        if (isWebSpeechListening && recognitionRef.current) {
+            recognitionRef.current.stop();
         } else if (mediaRecorderRef.current && isChatRecording) {
             mediaRecorderRef.current.stop();
-            setIsChatRecording(false);
-            
-            mediaRecorderRef.current.onstop = async () => {
-                const audioBlob = new Blob(audioChunksRef.current, { type: 'audio/webm' });
-                const base64 = await blobToBase64(audioBlob);
-                handleChatSubmit(undefined, base64);
-                mediaRecorderRef.current?.stream.getTracks().forEach(track => track.stop());
-            };
         }
+        setIsChatRecording(false);
+        setIsWebSpeechListening(false);
     };
 
     const handleChatSubmit = async (text?: string, audioBase64?: string) => {
-        if (!text && !audioBase64) return;
+        const finalContent = text || textInput;
+        if (!finalContent && !audioBase64) return;
+        
         stopPlayback();
         const historyPayload = messages.filter(m => m.id !== 'welcome').map(m => ({ role: m.role, text: m.text }));
         
         const newUserMsg: AIChatMessage = { 
             id: Date.now().toString(), 
             role: 'user', 
-            text: text || '(语音消息)', 
+            text: finalContent || '(语音消息)', 
             isAudioMessage: !!audioBase64, 
             timestamp: Date.now() 
         };
         const newAiMsgId = (Date.now() + 1).toString();
-        const newAiMsg: AIChatMessage = { 
-            id: newAiMsgId, 
-            role: 'model', 
-            text: '', 
-            timestamp: Date.now() 
-        };
+        const newAiMsg: AIChatMessage = { id: newAiMsgId, role: 'model', text: '', timestamp: Date.now() };
 
         setMessages(prev => [...prev, newUserMsg, newAiMsg]);
         setTextInput('');
@@ -247,13 +222,11 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
                     'x-user-role': currentUser?.role || '',
                     'x-school-id': currentUser?.schoolId || ''
                 },
-                body: JSON.stringify({ text, audio: audioBase64, history: historyPayload })
+                body: JSON.stringify({ text: finalContent, audio: audioBase64, history: historyPayload })
             });
 
             if (!response.ok) throw new Error(response.statusText);
-            if (!response.body) throw new Error('No response body');
-
-            const reader = response.body.getReader();
+            const reader = response.body!.getReader();
             const decoder = new TextDecoder();
             let aiTextAccumulated = '';
             let buffer = '';
@@ -280,36 +253,23 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
                                 playPCMAudio(data.audio);
                             }
                             if (data.ttsSkipped) {
-                                setToast({ show: true, message: 'AI 语音额度已用尽，已切换至本地语音播报', type: 'error' });
                                 speakWithBrowser(aiTextAccumulated);
                             }
-                            if (data.error) {
-                                setMessages(prev => prev.map(m => m.id === newAiMsgId ? { ...m, text: `⚠️ 错误: ${data.message || '未知错误'}` } : m));
-                            }
                         } catch (e) {}
                     }
                 }
             }
         } catch (error: any) {
-            setMessages(prev => prev.map(m => m.id === newAiMsgId ? { ...m, text: '抱歉，连接断开或发生错误，请重试。' } : m));
+            setMessages(prev => prev.map(m => m.id === newAiMsgId ? { ...m, text: '抱歉，连接断开，请重试。' } : m));
         } finally { setIsChatProcessing(false); }
     };
 
-    const clearHistory = () => {
-        setMessages([{ 
-            id: 'welcome', 
-            role: 'model', 
-            text: '你好！我是你的 AI 智能助教。有什么可以帮你的吗？', 
-            timestamp: Date.now() 
-        }]);
-    };
-
     return (
         <div className="flex-1 flex flex-col max-w-4xl mx-auto w-full min-h-0 relative overflow-hidden h-full">
             {toast.show && <Toast message={toast.message} type={toast.type} onClose={()=>setToast({...toast, show: false})}/>}
             
             <div className="absolute top-2 right-4 z-10">
-                <button onClick={clearHistory} className="text-xs text-gray-400 hover:text-red-500 flex items-center gap-1 bg-white/80 p-1.5 rounded-lg border border-transparent hover:border-red-100 transition-all shadow-sm backdrop-blur">
+                <button onClick={() => setMessages([{ id: 'welcome', role: 'model', text: '你好！我是你的 AI 智能助教。有什么可以帮你的吗？', timestamp: Date.now() }])} className="text-xs text-gray-400 hover:text-red-500 flex items-center gap-1 bg-white/80 p-1.5 rounded-lg border border-transparent hover:border-red-100 transition-all shadow-sm backdrop-blur">
                     <Trash2 size={14}/> 清除记录
                 </button>
             </div>
@@ -322,20 +282,68 @@ export const ChatPanel: React.FC<ChatPanelProps> = ({ currentUser }) => {
                         </div>
                         <div className={`max-w-[80%] p-3 rounded-2xl text-sm overflow-hidden ${msg.role === 'user' ? 'bg-blue-600 text-white rounded-tr-none' : 'bg-white border border-gray-200 text-gray-800 rounded-tl-none shadow-sm'}`}>
                             <div className="markdown-body"><ReactMarkdown remarkPlugins={[remarkGfm]}>{msg.text || ''}</ReactMarkdown></div>
-                            {msg.role === 'model' && !msg.text && isChatProcessing && <div className="flex items-center gap-2 text-gray-400 py-1"><Loader2 className="animate-spin" size={14}/><span className="text-xs">思考中...</span></div>}
-                            {msg.audio ? (<button onClick={() => playPCMAudio(msg.audio!)} className="mt-2 flex items-center gap-2 text-xs bg-blue-50 text-blue-600 px-3 py-1.5 rounded-full hover:bg-blue-100 border border-blue-100 transition-colors w-fit"><Volume2 size={14}/> 播放语音 (AI)</button>) : (msg.role === 'model' && msg.text && !isChatProcessing) && (<button onClick={() => speakWithBrowser(msg.text!)} className="mt-2 flex items-center gap-2 text-xs bg-gray-50 text-gray-600 px-3 py-1.5 rounded-full hover:bg-gray-100 border border-gray-200 transition-colors w-fit"><Volume2 size={14}/> 朗读 (本地)</button>)}
+                            {msg.audio && <button onClick={() => playPCMAudio(msg.audio!)} className="mt-2 flex items-center gap-2 text-xs bg-blue-50 text-blue-600 px-3 py-1.5 rounded-full hover:bg-blue-100 border border-blue-100 transition-colors w-fit"><Volume2 size={14}/> 播放语音 (AI)</button>}
                         </div>
                     </div>
                 ))}
+                {isChatProcessing && (
+                    <div className="flex gap-3">
+                         <div className="w-10 h-10 rounded-full bg-blue-100 text-blue-600 flex items-center justify-center shrink-0">
+                            <Loader2 className="animate-spin" size={20}/>
+                         </div>
+                         <div className="bg-white border border-gray-100 p-3 rounded-2xl rounded-tl-none shadow-sm flex items-center gap-2 text-gray-400 text-xs">
+                             思考中...
+                         </div>
+                    </div>
+                )}
                 <div ref={messagesEndRef} />
             </div>
             
+            {/* Unified Input Bar (No switching, Mic and Text coexist) */}
             <div className="p-4 bg-white border-t border-gray-200 shrink-0 z-20">
-                <div className="flex items-center gap-2 max-w-4xl mx-auto bg-gray-100 p-1.5 rounded-full border border-gray-200">
-                    <button onClick={() => setInputMode(inputMode === 'text' ? 'audio' : 'text')} className="p-2 rounded-full hover:bg-white text-gray-500 transition-colors">{inputMode === 'text' ? <Mic size={20}/> : <Square size={20}/>}</button>
-                    {inputMode === 'text' ? (<input className="flex-1 bg-transparent border-none outline-none px-2 text-sm" placeholder="输入问题..." value={textInput} onChange={e => setTextInput(e.target.value)} onKeyDown={e => e.key === 'Enter' && handleChatSubmit(textInput)}/>) : (<div className="flex-1 text-center text-sm font-medium text-blue-600 animate-pulse">{isChatRecording ? (isWebSpeechListening ? '正在识别...' : '正在录音...') : '按住麦克风说话'}</div>)}
-                    {inputMode === 'text' ? (<button onClick={() => handleChatSubmit(textInput)} className="p-2 bg-blue-600 rounded-full text-white hover:bg-blue-700 disabled:opacity-50" disabled={!textInput.trim() || isChatProcessing}><Send size={18}/></button>) : (<button onMouseDown={startRecording} onMouseUp={stopRecording} onTouchStart={startRecording} onTouchEnd={stopRecording} className={`p-3 rounded-full text-white transition-all ${isChatRecording ? 'bg-red-500 scale-110 shadow-lg ring-4 ring-red-200' : 'bg-blue-600 hover:bg-blue-700'}`}>{isChatRecording ? <StopCircle size={20}/> : <Mic size={20}/>}</button>)}
+                <div className="flex items-center gap-3 max-w-4xl mx-auto bg-gray-50 p-2 rounded-2xl border border-gray-200">
+                    {/* Recording Status Dot */}
+                    {isChatRecording && (
+                        <div className="flex items-center gap-1.5 px-3 py-1 bg-red-100 text-red-600 rounded-full animate-pulse text-[10px] font-bold shrink-0">
+                            <div className="w-2 h-2 bg-red-600 rounded-full"></div>
+                            {isWebSpeechListening ? '正在识别' : '正在录制'}
+                        </div>
+                    )}
+
+                    <input 
+                        ref={inputRef}
+                        className="flex-1 bg-transparent border-none outline-none px-3 text-sm py-2" 
+                        placeholder={isChatRecording ? "正在倾听..." : "输入问题..."}
+                        value={textInput} 
+                        onChange={e => setTextInput(e.target.value)} 
+                        onKeyDown={e => e.key === 'Enter' && !isChatProcessing && handleChatSubmit(textInput)}
+                        disabled={isChatProcessing}
+                    />
+
+                    <div className="flex items-center gap-2 shrink-0">
+                        {/* Mic Button - Coexists with input */}
+                        <button 
+                            onMouseDown={startRecording} 
+                            onMouseUp={stopRecording} 
+                            onTouchStart={startRecording} 
+                            onTouchEnd={stopRecording}
+                            className={`p-3 rounded-xl transition-all ${isChatRecording ? 'bg-red-500 scale-110 shadow-lg text-white ring-4 ring-red-100' : 'bg-gray-100 text-gray-500 hover:bg-gray-200'}`}
+                            title="按住说话"
+                        >
+                            {isChatRecording ? <StopCircle size={22}/> : <Mic size={22}/>}
+                        </button>
+
+                        {/* Send Button */}
+                        <button 
+                            onClick={() => handleChatSubmit(textInput)} 
+                            className={`p-3 rounded-xl transition-all ${!textInput.trim() || isChatProcessing ? 'bg-gray-100 text-gray-300' : 'bg-blue-600 text-white hover:bg-blue-700 shadow-md'}`} 
+                            disabled={!textInput.trim() || isChatProcessing}
+                        >
+                            <Send size={22}/>
+                        </button>
+                    </div>
                 </div>
+                <div className="text-[10px] text-gray-400 text-center mt-2">支持文字输入或按住麦克风图标进行语音提问</div>
             </div>
         </div>
     );