fix: template in generate_answer

src/tools/llm.py

@@ -48,6 +48,7 @@ class LlmAgent:
                     f"```{context}```"
                     f"\\n 2 You are consistent and avoid redundancies with the rest of the initial"
                     f" conversation delimited by triple backticks: ```{histo}```"
+                    f" If the langage of the answer is the same as the langage of the query, you shall be concise but keep the meaning of the answer"
                     )
 
         p = self.llm(template)