truncated answer reponse

service/llm_service.py

@@ -17,7 +17,7 @@ class LLMService:
 
         output = self.model.generate(
             **inputs,
-            max_new_tokens=120,
+            max_new_tokens=256,
             do_sample=True,
             temperature=0.7,
             top_p=0.9,

service/rag_service.py

@@ -24,23 +24,26 @@ def generate_answer(question: str, session_id: str) -> str:
     memory = get_memory(session_id)
 
     prompt = f"""
-<|system|>
-You are a helpful assistant.
-Answer ONLY using the provided context.
-If the answer is not in the context, say "I don't know".
+    <|system|>
+    You are a helpful assistant.
+    Answer ONLY using the provided context.
+    Give a COMPLETE, well-formed answer.
+    Do not stop mid-sentence.
+    If the answer is not in the context, say "I don't know".
 
-Conversation memory:
-{memory}
+    Conversation memory:
+    {memory}
 
-<|user|>
-Context:
-{chr(10).join(context)}
+    <|user|>
+    Context:
+    {chr(10).join(context)}
 
-Question:
-{question}
+    Question:
+    {question}
+
+    <|assistant|>
+    """
 
-<|assistant|>
-"""
 
     answer = llm.generate(prompt)