final changes to this branch

api.py

@@ -118,7 +118,7 @@ def _title_from_hf(query: str, client: InferenceClient, model_id: str) -> str |
             {"role": "user", "content": user_prompt},
         ],
         max_tokens=24,
-        temperature=0.2,
+        temperature=0.3,
     )
     if not response or not response.choices:
         return None

retriever/generator.py

@@ -5,16 +5,9 @@ class RAGGenerator:
     def generate_prompt(self, query, retrieved_contexts):
         """Prepares the academic prompt template."""
         context_text = "\n\n".join([f"--- Source {i+1} ---\n{c}" for i, c in enumerate(retrieved_contexts)])
-        
         return f"""You are an expert academic assistant. Use the following pieces of retrieved context to answer the question. 
 If the answer isn't in the context, say you don't know based on the provided documents.
 
-Formatting requirements:
-- Write the response in clean Markdown.
-- Use short headings and bullet points when helpful.
-- Use fenced code blocks for code.
-- Use simple markdown tables only when they improve clarity.
-
 Context:
 {context_text}