Update app.py

app.py

@@ -50,8 +50,8 @@ def call_openai_api(persona, user_prompt):
             response = openai.ChatCompletion.create(
                 model="gpt-3.5-turbo",
                 messages=[
-                    {"role": "system", "content": f"You are a knowledgeable assistant with a persona based on this: \n{persona}"},
-                    {"role": "user", "content": user_prompt},
+                    {"role": "system", "content": f"You are a knowledgeable assistant that provides short factual answers"},
+                    {"role": "user", "content": f"{persona}{user_prompt}",
                 ]
             )
             return response['choices'][0]['message']['content']
@@ -72,8 +72,15 @@ def pdf_to_text(file, user_prompt):
             pdf = PdfReader(pdf_file_io)
             for page in pdf.pages:
                 aggregated_text += page.extract_text()
+    # Tokenize aggregated_text
+    tokens = nltk.word_tokenize(aggregated_text)
+    # Split into chunks if tokens are more than 4096
+    if len(tokens) > 4096:
+        # Here you may choose the strategy that fits best.
+        # For instance, the first 4096 tokens could be used.
+        tokens = tokens[:4096]
     # Create a single persona from all text
-    persona = create_persona(aggregated_text)
+    persona = create_persona(' '.join(tokens))
     # Using OpenAI API
     response = call_openai_api(persona, user_prompt)
     return response