V0

app.py

@@ -103,7 +103,6 @@ submit_button = st.button('Submit')
 Answer_tab, Content_tab, Info_tab = st.tabs(["Answer", "Content used to create answer", "Infrmation about this app"])
 
 
-
 # Placeholder for displaying the answer
 with Answer_tab:
     answer_placeholder = st.empty()
@@ -119,13 +118,26 @@ as it takes more work to retrieve the text from them. It does include most order
 This is a simple RAG (retrieval augmented generation) system and does not consider order of events when 
 retrieving onformation and generating responses. It can also easily missinterpret information, but information used to generate the 
 response is presented in the content tab with link to the full document so that you can read the details in its proper context. 
-
                 
 
-""" 
-                )
+""" )
             
 
+
+    with open('results.json', 'r') as file:
+        content = file.read()
+
+
+    data_to_download = content.encode()
+
+    # Create a download button
+    st.download_button(label="Download Prior responses",
+                   data=data_to_download,
+                   file_name="results.json",
+                   mime="json")
+
+
+
 # Logic to display an answer when the submit button is pressed
 if submit_button:
     if question:  # Check if there is a question typed
@@ -134,8 +146,8 @@ if submit_button:
             if started: 
                 #Awnser = rag_chain.invoke(question)
                 #contexts = retriever.get_relevant_documents(question)
-                answer, selected_items,selected_sources,selected_chunks,highest_simularities = ask(question)
-                answer_placeholder.markdown(answer)  # Display the answer
+                answer, selected_items,selected_sources,titles,selected_chunks,highest_simularities = ask(question)
+                answer_placeholder.markdown(escape_markdown(answer))  # Display the answer
 
                     # Prepare the data to be saved
                 
@@ -178,10 +190,10 @@ if submit_button:
 
                 string = ""
                 for k in range(len(selected_items)):
-                    temp = " [" +  selected_sources[k] + "](" + url  +  selected_sources[k] +   ")" +  "  text block: " + selected_chunks[k]   + "   Relevance: " +f"{highest_simularities[k]:.2f}"
-
+                    temp = " [" +  titles[k] + "](" + url  +  selected_sources[k] +   ")" +  "  text block: " + selected_chunks[k]   + "   Relevance: " +f"{highest_simularities[k]:.2f}"
+                    
 
-                    string = string + "### Paragraph used. \n" + selected_items[k]  + "\n\n  source:"  + temp  + "\n"
+                    string = string + "### Paragraph used. \n" + escape_markdown(selected_items[k])  + "\n\n  source:"  + temp  + "\n"
 
 
                 content_placeholder.markdown(string) 

embeddings.npy

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2e59b70d53bc2779e24f76bcc2377fd60b9d3cdabf20b26cd8cfc176ec316292
-size 66072704
+oid sha256:578f2b44347eaff3137c23cbeb70ece93360e78c3d447ca9e4d825296a8bd0b3
+size 43131008

start2.py

@@ -8,6 +8,39 @@ from langchain.schema.runnable import RunnablePassthrough
 from langchain.schema.output_parser import StrOutputParser
 from scipy.spatial.distance import cosine
 
+
+def find_first_with_docket(items):
+    # Loop through each item in the list
+    k = 0
+    for item in items:
+        # Check if "docket" is in the item (case-insensitive search)
+        if "docket" in item.lower():
+            return item
+        k = k + 1
+    # Return None if no item contains "docket"
+    return 0
+
+
+
+def escape_markdownold(text):
+    # List of markdown special characters to escape
+    special_chars = r"\*|_|#|\{|\}|\[|\]|\(|\)|\#|\+|\-|\.|\!|\\"
+    
+    # Use regex sub function to escape special characters by adding a backslash before them
+    escaped_text = re.sub(f"([{special_chars}])", r"\\\1", text)
+    
+    return escaped_text
+
+def escape_markdown(text):
+    # List of special characters in markdown that need escaping
+    markdown_chars = ["\\", "`", "*", "_", "{", "}", "[", "]", "(", ")", "#", "+", "-", ".", "!", "|", ">","$"]
+    # Escape each character with a backslash
+    for char in markdown_chars:
+        text = text.replace(char, "\\" + char)
+    return text
+
+
+
 if not started:
     print("------------starting------------")
 
@@ -18,7 +51,7 @@ if not started:
 
 
     # Path to the pickle file where you want to save your data
-    pickle_file_path = 'vectorstore2.pkl'
+    pickle_file_path = 'vectorstore.pkl'
 
 
     with open(pickle_file_path, 'rb') as file:
@@ -106,13 +139,15 @@ def askq(query):
 
 
     # Combine content and question into a single prompt
-    prompt = f"""Anwser question provided given content. If an awnser can't be found in the provided ontent respond that you could not find 
+    prompt = f"""Anwser the question or request provided given content. If an awnser can't be found in the provided ontent respond that you could not find 
     the awnser to the question, applogice and say that you will instead tell a lawyer joke follow that with the joke.   
     Content: {content}\n\nQuestion: {question}\nAnswer:"""
 
 
     # Define LLM
-    llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0.1)
+    llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0.2)
+    #llm = ChatOpenAI(model_name="gpt-4", temperature=0.2)
+    
     Awnser = llm.invoke(prompt).content
 
     return Awnser, selected_items,selected_sources,selected_chunks,highest_simularities
@@ -152,13 +187,14 @@ def ask(query):
     selected_items = [docs[i].page_content for i in highest_indices]
     selected_sources = [docs[i].metadata['source'] for i in highest_indices]
     selected_chunks = [docs[i].metadata['chunk'] for i in highest_indices]
+    titles = [docs[i].metadata['title'] for i in highest_indices]
 
     # Combine selected items into a single content string
     content = ' '.join(selected_items)
 
     # Prepare the prompt
-    prompt = f"""Answer the question provided given the content. If an answer can't be found in the provided content, 
-    respond that you could not find the answer to the question, apologize and instead provide a suggestion for where to search for more information related to teh question.
+    prompt = f"""Answer the question or request provided given the content. If an answer can't be found in the provided content, 
+    respond that you could not find the answer to the question, apologize and instead provide a suggestion for where to search for more information related to the question.
 
     -------------------   
     Content: {content}\n\nQuestion: {question}\nAnswer:
@@ -170,4 +206,4 @@ def ask(query):
     llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0.1)
     answer = llm.invoke(prompt).content
 
-    return answer, selected_items, selected_sources, selected_chunks, highest_simularities
+    return answer, selected_items, selected_sources, titles, selected_chunks, highest_simularities

vectorstore.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:507fc427f4ebce75592e4035cb1b4f8601f46d82c3170f5529e21c5844fc8440
-size 5553356
+oid sha256:54f335b6791bcf72f95e705e8418e6ba343585df5b6bb1bedf7e738f9a2b698f
+size 5449252

vectorstore2.pkl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:0eeb601bfdd128945a52712a20a89f9bfd89c85ea1d25215d552f68ca094b012
-size 5582531