update

app.py

@@ -44,6 +44,12 @@ def download_arxiv_source(paper_id):
         for member in tar.getmembers()
         if member.name.endswith(".tex")
     }
+    # Load all .tex files into memory, including their subdirectories
+    tex_files = {
+        member.name: tar.extractfile(member).read().decode("utf-8")
+        for member in tar.getmembers()
+        if member.isfile() and member.name.endswith(".tex")
+    }
 
     # Pattern to match \input{filename} and \include{filename}
     pattern = re.compile(r"\\(input|include){(.*?)}")
@@ -55,8 +61,8 @@ def download_arxiv_source(paper_id):
             match = re.search(pattern, line)
             if match:
                 command, filename = match.groups()
-                # LaTeX automatically adds .tex extension for \include command
-                if command == "include":
+                # LaTeX automatically adds .tex extension for \input and \include commands
+                if not filename.endswith(".tex"):
                     filename += ".tex"
                 if filename in tex_files:
                     output.append(replace_includes(tex_files[filename]))
@@ -88,9 +94,10 @@ class ContextualQA:
         self.context = text
 
     def ask_question(self, question):
-        leading_prompt = "Consider the following paper:"
-        trailing_prompt = "Now, answer the following question using Markdown syntax:"
-        prompt = f"{anthropic.HUMAN_PROMPT} {leading_prompt}\n\n{self.context}\n\n{trailing_prompt}\n\n{anthropic.HUMAN_PROMPT} {question} {anthropic.AI_PROMPT}"
+        leading_prompt = "Here is the content of a paper:"
+        trailing_prompt = "Now, answer the following question:"
+        markdown_notice = "Note:  You can use markdown syntax to format your question."
+        prompt = f"{anthropic.HUMAN_PROMPT} {leading_prompt}\n\n{self.context}\n\n{trailing_prompt}\n\n{anthropic.HUMAN_PROMPT} {question} \n({markdown_notice})\n\n{anthropic.AI_PROMPT}"
         response = self.client.completion_stream(
             prompt=prompt,
             stop_sequences=[anthropic.HUMAN_PROMPT],