fix assistant stream truncation handling

public/js/clinicalAssistant.js

@@ -21,6 +21,7 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
   var exportCacheKey = '';
   var exportCacheItems = null;
   var markdownRenderer = null;
+  var assistantBusy = false;
 
   document.addEventListener('tabChanged', function (e) {
     if (e.detail && e.detail.tab === 'assistant') initIfNeeded();
@@ -100,6 +101,10 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
     var input = document.getElementById('assistant-input');
     var includeContext = document.getElementById('assistant-include-context');
     var text = input ? input.value.trim() : '';
+    if (assistantBusy) {
+      if (typeof showToast === 'function') showToast('Assistant is still finishing the current answer', 'error');
+      return;
+    }
     if (!text) {
       if (typeof showToast === 'function') showToast('Enter a clinical question', 'error');
       return;
@@ -202,6 +207,10 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
       var tail = parseSseEvent(buffer);
       if (tail) handleEvent(tail.type, tail.data);
     }
+    if (!doneData) {
+      updateLoadingMessage(loading, 'Stream ended early. Retrying without streaming...');
+      doneData = await fetchAssistantFallback(payload);
+    }
 
     setBusy(false, 'Ready');
     var answer = (doneData && (doneData.answer || doneData.markdown)) || partial;
@@ -215,6 +224,18 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
     }
   }
 
+  async function fetchAssistantFallback(payload) {
+    var response = await fetch('/api/clinical-assistant/chat', {
+      method: 'POST',
+      headers: getAuthHeaders(),
+      credentials: 'same-origin',
+      body: JSON.stringify(payload)
+    });
+    var data = await response.json().catch(function () { return {}; });
+    if (!response.ok || !data.success) throw new Error(data.error || ('Request failed (' + response.status + ')'));
+    return data;
+  }
+
   function parseSseEvent(block) {
     var type = 'message';
     var data = '';
@@ -899,9 +920,11 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
   }
 
   function setBusy(isBusy, text, isError) {
+    assistantBusy = !!isBusy;
     var status = document.getElementById('assistant-status');
     var label = document.getElementById('assistant-status-text');
     var send = document.getElementById('btn-assistant-send');
+    var input = document.getElementById('assistant-input');
     if (status) {
       status.classList.toggle('busy', !!isBusy);
       status.classList.toggle('error', !!isError);
@@ -911,6 +934,7 @@ import { escapeAttr, escapeHtml, renderAssistantMarkdown, stripSourcesSection }
       send.disabled = !!isBusy;
       send.innerHTML = isBusy ? '<i class="fas fa-spinner fa-spin"></i> Searching' : '<i class="fas fa-paper-plane"></i> Ask';
     }
+    if (input) input.disabled = !!isBusy;
   }
 
   function sanitize(html) { return window.DOMPurify ? window.DOMPurify.sanitize(html, { ADD_ATTR: ['target'] }) : html; }

src/routes/clinicalAssistant.js

@@ -241,6 +241,18 @@ router.post('/clinical-assistant/chat', async function(req, res) {
       maxTokens: 2600
     });
     var answer = stripModelSourcesSection(String(ai.content || '').trim());
+    if (shouldRegenerateTruncatedAnswer(answer, ai.finishReason)) {
+      console.warn('[clinical-assistant] non-stream answer looked truncated; regenerating', { finishReason: ai.finishReason, chars: answer.length });
+      var completed = await callAI(prepared.messages, {
+        model: prepared.chatModel || undefined,
+        temperature: 0.15,
+        maxTokens: 5000
+      });
+      answer = stripModelSourcesSection(String(completed.content || '').trim()) || answer;
+      ai.model = completed.model || ai.model;
+      ai.provider = completed.provider || ai.provider;
+      ai.finishReason = completed.finishReason || ai.finishReason;
+    }
 
     logger.audit(req.user.id, 'clinical_assistant_query', 'Clinical assistant query', req, {
       category: 'clinical', model: ai.model || prepared.chatModel, duration: Date.now() - started
@@ -294,11 +306,12 @@ router.post('/clinical-assistant/chat/stream', async function(req, res) {
     });
     var answer = stripModelSourcesSection(String(ai.content || '').trim());
     if (shouldRegenerateTruncatedAnswer(answer, ai.finishReason)) {
+      console.warn('[clinical-assistant] stream answer looked truncated; regenerating', { finishReason: ai.finishReason, chars: answer.length });
       sendEvent('status', { message: 'Completing answer...' });
       var completed = await callAI(prepared.messages, {
         model: prepared.chatModel || undefined,
         temperature: 0.15,
-        maxTokens: 3200
+        maxTokens: 5000
       });
       answer = stripModelSourcesSection(String(completed.content || '').trim()) || answer;
       ai.model = completed.model || ai.model;

src/utils/ai.js

@@ -167,7 +167,8 @@ async function callOpenRouter(messages, model, temperature, maxTokens) {
     content: completion.choices[0].message.content,
     model: model,
     provider: 'openrouter',
-    usage: completion.usage || null
+    usage: completion.usage || null,
+    finishReason: completion.choices[0].finish_reason || null
   };
 }
 
@@ -189,7 +190,8 @@ async function callAzure(messages, model, temperature, maxTokens) {
     content: completion.choices[0].message.content,
     model: process.env.AZURE_DEPLOYMENT_NAME || model,
     provider: 'azure',
-    usage: completion.usage || null
+    usage: completion.usage || null,
+    finishReason: completion.choices[0].finish_reason || null
   };
 }
 
@@ -391,7 +393,8 @@ async function callLiteLLM(messages, model, temperature, maxTokens) {
     content: completion.choices[0].message.content,
     model: model,
     provider: 'litellm',
-    usage: completion.usage || null
+    usage: completion.usage || null,
+    finishReason: completion.choices[0].finish_reason || null
   };
 }