fix(api): improve stream token chunks robustness and sync year filter

frontend-next/app/page.tsx

@@ -296,11 +296,15 @@ export default function App() {
                 body: JSON.stringify({ 
                     question: originalQuery, 
                     top_k: topK, 
-                    filter_category: category === "All" ? undefined : category 
+                    filter_category: category === "All" ? undefined : category,
+                    filter_year_gte: yearFilter ? yearFrom : undefined
                 })
             });
 
-            if (!res.ok || !res.body) throw new Error("Stream failed");
+            if (!res.ok || !res.body) {
+                const errText = await res.text();
+                throw new Error(`API error: ${res.status} ${errText}`);
+            }
 
             const reader = res.body.getReader();
             const decoder = new TextDecoder();

src/rag/llm_client.py

@@ -72,7 +72,10 @@ class MultiModelClient:
                                 break
                             try:
                                 data = json.loads(data_str)
-                                token = data["choices"][0]["delta"].get("content", "")
+                                choices = data.get("choices", [])
+                                if not choices:
+                                    continue
+                                token = choices[0].get("delta", {}).get("content", "")
                                 if token:
                                     yield token
                             except:
@@ -95,7 +98,10 @@ class MultiModelClient:
         if stream:
             def generator():
                 for chunk in response:
-                    token = chunk.choices[0].delta.content
+                    choices = chunk.choices
+                    if not choices:
+                        continue
+                    token = choices[0].delta.content
                     if token:
                         yield token
             return generator()

test_stream.py

@@ -0,0 +1,16 @@
+import os
+import sys
+from dotenv import load_dotenv
+
+load_dotenv()
+sys.path.append(os.getcwd())
+
+from src.rag.pipeline import RAGPipeline
+
+try:
+    pipeline = RAGPipeline()
+    gen = pipeline.stream_query("What is LoRA?", top_k=2)
+    for x in gen:
+        print(x)
+except Exception as e:
+    print(f"Error: {e}")