Update app.py

app.py

@@ -40,15 +40,11 @@ def encode_image(image_files):
 
 st.set_page_config(page_title="Document/Image AI Analyst", layout="centered", initial_sidebar_state="collapsed")
 # Streamlit page setup
-st.title("Document/Image AI Analyst: `GPT-4 with Vision` 👀")
+st.title("Document/Image AI Analyst: `GPT-4 with Vision v2` 👀")
 st.write("")
 st.write("")
 st.write("")
 
-# Retrieve the OpenAI API Key from secrets
-#load_dotenv()
-# api_key = os.getenv("OPENAI_API_KEY")t
-
 # Guide
 st.subheader('What can it do?')
 with st.expander('Read more details', expanded=False):
@@ -81,98 +77,39 @@ st.subheader('2. Upload Images: ')
 uploaded_files = st.file_uploader("", help='Up to five images only.', type=["jpg", "png", "jpeg"], accept_multiple_files=True)
 
 
-#############
-
-# loader = TextLoader("./kb_document_requirements_v1.txt", encoding="utf-8")
-# documents = loader.load()
-
-# text_splitter = CharacterTextSplitter(chunk_size=300, chunk_overlap=0) #you reduced chunk_size from 1000
-# texts = text_splitter.split_documents(documents)
-# for i, text in enumerate(texts):
-#     text.metadata["source"] = f"{i+1}-page"
-# embeddings = OpenAIEmbeddings(api_key=api_key)
-# docsearch = Chroma.from_documents(texts, embeddings) 
-
-# llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo-0613", api_key=api_key)
-# qa_chain = create_qa_with_sources_chain(llm)
-
-# doc_prompt = PromptTemplate(
-#     template="Content: {page_content}\nSource: {source}",
-#     input_variables=["page_content", "source"],
-# )
-
-# final_qa_chain = StuffDocumentsChain(
-#     llm_chain=qa_chain,
-#     document_variable_name="context",
-#     document_prompt=doc_prompt,
-# )
-
-# retrieval_qa = RetrievalQA(
-#     retriever=docsearch.as_retriever(), combine_documents_chain=final_qa_chain
-# )
-#############
-
-# Retrieve the OpenAI API Key from secrets
-# load_dotenv()
-# api_key = os.getenv("OPENAI_API_KEY")
-
-# docs = [
-#     Document(
-#             page_content="Rules and requirements of IELTS",
-#             metadata={"type": "english proficiency certification", "name": "ielts", "filename": "./docs/ielts_kb_doc.txt"},
-#         ),
-#     Document(
-#             page_content="Rules and requirements of IELTS UKVI Academic",
-#             metadata={"type": "english proficiency certification", "name": "ielts ukvi", "filename": "./docs/pearson_pte_kb_doc.txt"},
-#         ),
-#     Document(
-#             page_content="Rules and requirements of TOEFL iBT (Test of English as a Foreign Language internet-based test)",
-#             metadata={"type": "english proficiency certification", "name": "toefl", "filename": "./docs/toefl_kb_doc.txt"},
-#         ),
-#     Document(
-#             page_content="Rules and requirements of Pearson PTE Academic",
-#             metadata={"type": "english proficiency certification", "name": "pearson pte academic", "filename": "./docs/ielts_kb_doc.txt"},
-#         ),
-#     Document(
-#             page_content="Requirements of international passport",
-#             metadata={"type": "passport", "name": "international passport", "filename": "./docs/passport_kb_doc.txt"},
-#         ),
-# ]
-
-# docsearch = Chroma.from_documents(docs, OpenAIEmbeddings(), persist_directory="./knowledge_base")
-
-# load from disk
-docsearch = Chroma(persist_directory="./knowledge_base", embedding_function=OpenAIEmbeddings())
-
-metadata_field_info = [
-    AttributeInfo(
-        name="type",
-        description="The type of document",
-        type="string",
-    ),
-    AttributeInfo(
-        name="name",
-        description="The name of the document",
-        type="string",
-    ),
-    AttributeInfo(
-        name="filename",
-        description="The file name of the document",
-        type="string",
-    ),    
-]
-
-document_content_description = "Type of a document"
-llm = ChatOpenAI(temperature=0)
-
-retriever = SelfQueryRetriever.from_llm(
-    llm,
-    docsearch,
-    document_content_description,
-    metadata_field_info,
-    enable_limit=True,
-    search_kwargs={"k": 1}
-)
+if api_key:
+    # load from disk
+    docsearch = Chroma(persist_directory="./knowledge_base", embedding_function=OpenAIEmbeddings())
+
+    metadata_field_info = [
+        AttributeInfo(
+            name="type",
+            description="The type of document",
+            type="string",
+        ),
+        AttributeInfo(
+            name="name",
+            description="The name of the document",
+            type="string",
+        ),
+        AttributeInfo(
+            name="filename",
+            description="The file name of the document",
+            type="string",
+        ),    
+    ]
+
+    document_content_description = "Type of a document"
+    llm = ChatOpenAI(temperature=0)
+
+    retriever = SelfQueryRetriever.from_llm(
+        llm,
+        docsearch,
+        document_content_description,
+        metadata_field_info,
+        enable_limit=True,
+        search_kwargs={"k": 1}
+    )
 
 # This sample for querying
 # results = retriever.invoke("requirements in passport")
@@ -400,11 +337,10 @@ if uploaded_files is not None and api_key and analyze_button:
             message_placeholder.markdown("**Identified Documents:**\n\n" + markdown_string)
 
             ### TO DO : The ABOVE code can be in a function or object.
-            # Initialize an empty array to store the results
+            # Initialize an empty array to store the full prompt as log 
             retrieved_knowledge = []
             
             # Iterate over each item in the full_response dictionary
-            #for filename, details in full_response_json.items():
             st.subheader("5. Analysis Results")
             for i, (filename, details) in enumerate(full_response_json.items()):
 
@@ -472,10 +408,12 @@ if uploaded_files is not None and api_key and analyze_button:
                     # Now you can work with file_content
                     #print(file_content)
                     # add data to array
-                    #retrieved_knowledge.append(validation_response) 
+                    retrieved_knowledge.extend([prompt_text_validation, messages]) 
 
                     validation_placeholder.markdown(validation_response)
-                
+
+            with st.expander("Process Logs: "):
+                st.json(retrieved_knowledge) 
 
             # Display the response in the app
             # st.write(response.choices[0].message.content)