Update app.py

app.py

@@ -9,7 +9,7 @@ from langchain.document_loaders import PDFPlumberLoader
 from langchain_experimental.text_splitter import SemanticChunker
 from langchain_huggingface import HuggingFaceEmbeddings
 from langchain_chroma import Chroma
-from prompts import rag_prompt, relevancy_prompt, relevant_context_picker_prompt, response_synth
+from langchain.chains import SequentialChain, LLMChain
 
 # Set API Keys
 os.environ["GROQ_API_KEY"] = st.secrets.get("GROQ_API_KEY", "")
@@ -41,7 +41,7 @@ if "vector_created" not in st.session_state:
 st.title("Blah-2")
 
 # Step 1: Choose PDF Source
-pdf_source = st.radio("Upload or provide a link to a PDF:", ["Upload a PDF file", "Enter a PDF URL"], index=0, horizontal=True)
+pdf_source = st.radio("Upload or provide a link to a PDF:", ["Enter a PDF URL", "Upload a PDF file"], index=0, horizontal=True)
 
 if pdf_source == "Upload a PDF file":
     uploaded_file = st.file_uploader("Upload your PDF file", type="pdf")
@@ -54,7 +54,7 @@ if pdf_source == "Upload a PDF file":
         st.session_state.vector_created = False
 
 elif pdf_source == "Enter a PDF URL":
-    pdf_url = st.text_input("Enter PDF URL:")
+    pdf_url = st.text_input("Enter PDF URL:", value = "https://arxiv.org/pdf/2406.06998")
     if pdf_url and not st.session_state.pdf_path:
         with st.spinner("Downloading PDF..."):
             try:
@@ -157,101 +157,136 @@ if query:
         # Debugging: Check extracted context
         st.write("Extracted Context (page_content):", context)
         st.write("Number of Extracted Contexts:", len(context))
-    # ----------------- Run Individual Chains Explicitly -----------------
-
-    context_relevancy_checker_prompt = PromptTemplate(input_variables=["retriever_query","context"],template=relevancy_prompt)
-
-    if "context_relevancy_evaluation_chain" in st.session_state:
-        del st.session_state["context_relevancy_evaluation_chain"]
-
-
-
-    context_relevancy_evaluation_chain = LLMChain( llm=llm_judge, prompt=PromptTemplate(input_variables=["retriever_query","context"],template=relevancy_prompt), output_key="relevancy_response" )
-
-    st.write("🚀 Debugging Expected Keys for `context_relevancy_evaluation_chain`:")
-    st.write(context_relevancy_evaluation_chain.input_keys)
-
-    response_crisis = context_relevancy_evaluation_chain.invoke({"context":context,"retriever_query":query})
-
-    relevant_prompt = relevant_prompt = PromptTemplate(input_variables=["relevancy_response"], template= relevant_context_picker_prompt)
-
-    pick_relevant_context_chain = LLMChain(llm=llm_judge, prompt=relevant_prompt, output_key="context_number")
-
-    relevant_response = pick_relevant_context_chain.invoke({"relevancy_response":response_crisis['relevancy_response']})
-
-    context_prompt = PromptTemplate(input_variables=["context_number"],template=response_synth)
-
-    relevant_contexts_chain = LLMChain(llm=llm_judge, prompt=context_prompt, output_key="relevant_contexts")
-
-    contexts = relevant_contexts_chain.invoke({"context_number":relevant_response['context_number'],"context":context})
-
-    #temp
-    st.subheader("Relevant Contexts")
-    st.json(contexts['relevant_contexts']) 
-
-    response_chain = LLMChain(llm=rag_llm,prompt=final_prompt,output_key="final_response")
+        
+        relevancy_prompt = """You are an expert judge tasked with evaluating whether the  EACH OF THE CONTEXT provided in the CONTEXT LIST is self sufficient to answer the QUERY asked.
+        Analyze the provided QUERY AND  CONTEXT to determine if each Ccontent in the CONTEXT LIST contains Relevant information to answer the QUERY.
+        
+        Guidelines:
+        1. The content must not introduce new information beyond what's provided in the QUERY.
+        2. Pay close attention to the subject of statements. Ensure that attributes, actions, or dates are correctly associated with the right entities (e.g., a person vs. a TV show they star in).
+        3. Be vigilant for subtle misattributions or conflations of information, even if the date or other details are correct.
+        4. Check that the content in the CONTEXT LIST doesn't oversimplify or generalize information in a way that changes the meaning of the QUERY.
+        
+        Analyze the text thoroughly and assign a relevancy score 0 or 1 where:
+        - 0: The content has all the necessary information to answer the QUERY
+        - 1: The content does not has the necessary information to answer the QUERY
+        
+        ```
+        EXAMPLE:
+        INPUT (for context only, not to be used for faithfulness evaluation):
+        What is the capital of France?
+        
+        CONTEXT:
+        ['France is a country in Western Europe. Its capital is Paris, which is known for landmarks like the Eiffel Tower.',
+        'Mr. Naveen patnaik has been the chief minister of Odisha for consequetive 5 terms']
+        
+        OUTPUT:
+        The Context has sufficient information to answer the query.
+        
+        RESPONSE:
+        {{"score":0}}
+        ```
+        
+        CONTENT LIST:
+        {context}
+        
+        QUERY:
+        {retriever_query}
+        Provide your verdict in JSON format  with a single key 'score' and no preamble or explanation:
+        [{{"content:1,"score": <your score either 0 or 1>,"Reasoning":<why you have chose the score as 0 or 1>}},
+        {{"content:2,"score": <your score either 0 or 1>,"Reasoning":<why you have chose the score as 0 or 1>}},
+        ...]
+        """
+        
+        context_relevancy_checker_prompt = PromptTemplate(input_variables=["retriever_query","context"],template=relevancy_prompt)
+
+        relevant_prompt = PromptTemplate(
+            input_variables=["relevancy_response"],
+            template="""
+            Your main task is to analyze the json structure as a part of the Relevancy Response.
+            Review the Relevancy Response and do the following:-
+            (1) Look at the Json Structure content
+            (2) Analyze the 'score' key in the Json Structure content.
+            (3) pick the value of 'content' key against those 'score' key value which has 0.
+            
+            Relevancy Response:
+            {relevancy_response}
+            
+            Provide your verdict in JSON format  with a single key 'content number' and no preamble or explanation:
+            [{{"content":<content number>}}]
+            """
+        )
+        
+        context_prompt = PromptTemplate(
+            input_variables=["context_number"],
+            template="""
+            You main task is to analyze the json structure as a part of the Context Number Response and the list of Contexts provided in the 'Content List' and perform the following steps:-
+            (1) Look at the output from the Relevant Context Picker Agent.
+            (2) Analyze the 'content' key in the Json Structure format({{"content":<<content_number>>}}).
+            (3) Retrieve the value of 'content' key and pick up the context corresponding to that element from the Content List provided.
+            (4) Pass the retrieved context for each corresponing element number referred in the 'Context Number Response'
+            
+            Context Number Response:
+            {context_number}
+            
+            Content List:
+            {context}
+            
+            Provide your verdict in JSON format  with a two key 'relevant_content' and 'context_number' no preamble or explanation:
+            [{{"context_number":<content1>,"relevant_content":<content corresponing to that element 1 in the Content List>}},
+            {{"context_number":<content4>,"relevant_content":<content corresponing to that element 4 in the Content List>}},
+            ...
+            ]
+            """
+        )
+
+        rag_prompt = """ You are ahelpful assistant very profiient in formulating clear and meaningful answers from the context provided.Based on the CONTEXT Provided ,Please formulate
+        a clear concise and meaningful answer for the QUERY asked.Please refrain from making up your own answer in case the COTEXT provided is not sufficient to answer the QUERY.In such a situation please respond as 'I do not know'.
+        
+        QUERY:
+        {query}
+        
+        CONTEXT
+        {context}
+        
+        ANSWER:
+        """
 
+        context_relevancy_evaluation_chain = LLMChain(llm=llm_judge, prompt=context_relevancy_checker_prompt, output_key="relevancy_response")
 
-    #temp
-    st.subheader("Response Chain")
-    st.json(response_chain)
+        response_crisis = context_relevancy_evaluation_chain.invoke({"context":context,"retriever_query":query})
 
-    #response = chain.invoke({"query":query,"context":contexts['relevant_contexts']})
+        pick_relevant_context_chain = LLMChain(llm=llm_judge, prompt=relevant_prompt, output_key="context_number")
 
-    #temp
-    #st.subheader("blah response")
-    #st.json(response.content)
+        relevant_response = pick_relevant_context_chain.invoke({"relevancy_response":response_crisis['relevancy_response']})
 
-    # Orchestrate using SequentialChain
-    context_management_chain = SequentialChain(
-        chains=[context_relevancy_evaluation_chain ,pick_relevant_context_chain, relevant_contexts_chain,response_chain],
-        input_variables=["context","retriever_query","query"],
-        output_variables=["relevancy_response", "context_number","relevant_contexts","final_response"]
-    )
-    final_output = context_management_chain({"context":context,"retriever_query":query,"query":query})
-    st.subheader("Final Output from Context Management chain")
-    st.json(final_output)
-    
-    st.subheader("Context of Final Output from Context Management chain")
-    st.json(final_output['context'])
-    
-    st.header("Relevancy Response")
-    st.json(final_output['relevancy_response'])
-    
-    st.subheader("Relevant Context")
-    st.json(final_output['relevant_contexts'])
-    
-    response = chain.invoke({"query":query,"context":final_output['relevant_contexts']})
-    
-    st.subheader("Final Response")
-    st.json(response.content)
+        relevant_contexts_chain = LLMChain(llm=llm_judge, prompt=context_prompt, output_key="relevant_contexts")
 
-    # ----------------- Display All Outputs -----------------
-    #st.subheader("response_crisis")
-    #st.json((response_crisis))
+        contexts = relevant_contexts_chain.invoke({"context_number":relevant_response['context_number'],"context":context})
 
-    #st.subheader("response_crisis['relevancy_response']")
-    #st.json((response_crisis['relevancy_response']))
+        final_prompt = PromptTemplate(input_variables=["query","context"],template=rag_prompt)
 
+        response_chain = LLMChain(llm=rag_llm,prompt=final_prompt,output_key="final_response")
 
-    
-    #st.markdown("### Context Relevancy Evaluation")
-    #st.json(response_crisis["relevancy_response"])
+        response = response_chain.invoke({"query":query,"context":contexts['relevant_contexts']})
 
-    #st.markdown("### Picked Relevant Contexts")
-    #st.json(relevant_response["context_number"])
+        # Orchestrate using SequentialChain
+        context_management_chain = SequentialChain(
+            chains=[context_relevancy_evaluation_chain ,pick_relevant_context_chain, relevant_contexts_chain,response_chain],
+            input_variables=["context","retriever_query","query"],
+            output_variables=["relevancy_response", "context_number","relevant_contexts","final_response"]
+        )
 
-    #st.markdown("### Extracted Relevant Contexts")
-    #st.json(contexts["relevant_contexts"])
+        final_output = context_management_chain({"context":context,"retriever_query":query,"query":query})
 
-    #st.subheader("context_relevancy_evaluation_chain Statement")
-    #st.json(final_response["relevancy_response"])
+        st.subheader('final_output["relevancy_response"]')
+        st.json(final_output["relevancy_response"] )
 
-    #st.subheader("pick_relevant_context_chain Statement")
-    #st.json(final_response["context_number"])
+        st.subheader('final_output["context_number"]')
+        st.json(final_output["context_number"])
 
-    #st.subheader("relevant_contexts_chain Statement")
-    #st.json(final_response["relevant_contexts"])
+        st.subheader('final_output["relevant_contexts"]')
+        st.json(final_output["relevant_contexts"])
 
-    #st.subheader("RAG Response Statement")
-    #st.json(final_response["final_response"])
+        st.subheader('final_output["final_response"]')
+        st.json(final_output["final_response"])