feat(prompts): implement TOML-backed prompt loading and refactor message generation

python/ebook_search/answer.py

@@ -6,6 +6,7 @@ import logging
 from typing import TYPE_CHECKING
 
 from python.ebook_search.llm_interface import request_chat_completion
+from python.ebook_search.prompts import load_prompt
 
 if TYPE_CHECKING:
     import httpx
@@ -42,16 +43,7 @@ async def answer_query(
     content = await request_chat_completion(
         client,
         config,
-        [
-            {
-                "role": "system",
-                "content": (
-                    "Answer only from the provided context. Cite sources with bracketed numbers like [1]. "
-                    "If the context is insufficient, say so."
-                ),
-            },
-            {"role": "user", "content": f"Question:\n{query}\n\nContext:\n{context}"},
-        ],
+        load_prompt("answer").messages(query=query, context=context),
     )
 
     logger.info(f"ebook_answer_request_complete {config.chat_model=} answer_length={len(content)}")