insure correct pdf retrieval

modules/text/module_text_llm/module_text_llm/retrieval_augmented_generation/agents.py

@@ -41,7 +41,7 @@ def __init__(self, session_id="test-session"):
         self.model = ChatOpenAI(model="gpt-4o-mini") #gpt-4o-2024-08-06 , gpt-4o-mini
         self.memory = InMemoryChatMessageHistory(session_id=session_id)
         all_docs = []
-        file_paths = glob.glob("pdfs/*.pdf")
+        file_paths = glob.glob("module_text_llm/retrieval_augmented_generation/pdfs/*.pdf")
 
         for file_path in file_paths:
             loader = PyPDFLoader(file_path)
@@ -56,18 +56,43 @@ def __init__(self, session_id="test-session"):
 
         retriever = vectorstore.as_retriever()
         retriever_tool = create_retriever_tool(retriever, name="retrieve_document", description="Retrieves the pdf documents from the relevant lecture")
+        system_message = """\
+        You are an AI tutor for text assessment at a prestigious university.
 
+        # Task
+        Create graded feedback suggestions for a student\'s text submission that a human tutor would accept. \
+        Meaning, the feedback you provide should be applicable to the submission with little to no modification.
+        You have access to the provided document lecture slides to help you provide feedback. If you do use them, please reference the title and the page on your feedback.
+
+        # Style
+        1. Constructive, 2. Specific, 3. Balanced, 4. Clear and Concise, 5. Actionable, 6. Educational, 7. Contextual
+
+        Make use of the lecture slides provided. State clearly on your feedback which lecture you are using.
+
+        # Problem statement
+        {problem_statement}
+
+        # Example solution
+        {example_solution}
+
+        # Grading instructions
+        {grading_instructions}
+        Max points: {max_points}, bonus points: {bonus_points}\
+            
+        Respond only in json with the provided Assessment Feedback schema.
+        """
         # Define the prompt template with a system message placeholder
         self.prompt = ChatPromptTemplate.from_messages(
             [
-                ("system", """ You are an AI tutor for text assessment at a prestigious university.
-
-# Task
-Create graded feedback suggestions for a student\'s text submission that a human tutor would accept. \
-Meaning, the feedback you provide should be applicable to the submission with little to no modification.
-You have access to the provided document lecture slides to help you provide feedback. If you do use them, please reference the title and the page on your feedback.
-ALWAYS RESPONG IN A PYDNATIC OBJECT WITH THE PROVIDED ASSESSMENT MODEL SCHEMA.
-problem satement: {problem_statement}, max points: {max_points}"""),
+                ("system", system_message),
+#                 ("system", """ You are an AI tutor for text assessment at a prestigious university.
+
+# # Task
+# Create graded feedback suggestions for a student\'s text submission that a human tutor would accept. \
+# Meaning, the feedback you provide should be applicable to the submission with little to no modification.
+# You have access to the provided document lecture slides to help you provide feedback. If you do use them, please reference the title and the page on your feedback.
+# ALWAYS RESPONG IN A PYDNATIC OBJECT WITH THE PROVIDED ASSESSMENT MODEL SCHEMA.
+# problem satement: {problem_statement}, max points: {max_points}"""),
                 # ("placeholder", "{chat_history}"),  # History of interactions
                 ("human", "{submission}"),
                 ("placeholder", "{agent_scratchpad}"),  # Internal for steps created through function calling

modules/text/module_text_llm/module_text_llm/retrieval_augmented_generation/prompt_generate_suggestions.py

@@ -13,6 +13,12 @@
 # Style
 1. Constructive, 2. Specific, 3. Balanced, 4. Clear and Concise, 5. Actionable, 6. Educational, 7. Contextual
 
+Make use of the lecture slides provided. State clearly on your feedback which lecture you are using. If you
+believe that the student could benefit from the slide refer it on your feedback.
+
+The grading instructions are there to guide you on which criteria to give points. 
+You can comment with 0 points about grammar and spelling errors, but you should not give or remove points for them.
+
 # Problem statement
 {problem_statement}
 

modules/text/module_text_llm/pyproject.toml

@@ -15,6 +15,7 @@ gitpython = "3.1.41"
 nltk = "3.8.1"
 python-dotenv = "1.0.0"
 tiktoken = "0.7.0"
+pypdf = "5.1.0"
 
 [tool.poetry.dev-dependencies]
 pydantic = "1.10.17"