From 86ac00cc09531636591b141814f7dc167858917a Mon Sep 17 00:00:00 2001
From: Rob Kerr <mail@robkerr.com>
Date: Sat, 18 Nov 2023 22:34:08 -0500
Subject: [PATCH] Add Streamlit python source app.py

---
 .../app.py                                    | 84 +++++++++++++++++++
 1 file changed, 84 insertions(+)
 create mode 100644 azure-ai-search-integrated-vector-embeddings/app.py

diff --git a/azure-ai-search-integrated-vector-embeddings/app.py b/azure-ai-search-integrated-vector-embeddings/app.py
new file mode 100644
index 0000000..0ee6e1f
--- /dev/null
+++ b/azure-ai-search-integrated-vector-embeddings/app.py
@@ -0,0 +1,84 @@
+import streamlit as st
+import openai, os, requests, json
+
+openai.api_type = "azure"
+openai.api_version = "2023-08-01-preview"
+
+# Azure OpenAI setup
+openai.api_base = "https://<openai service name>.openai.azure.com/" 
+openai.api_key = os.getenv("AZURE_OPENAI_KEY") 
+deployment_id = "turbo-35-16k" 
+
+# Azure AI Search setup
+search_endpoint = "https://robkerrai-demo.search.windows.net"
+search_key = os.getenv("AZURE_SEARCH_KEY")
+search_index_name = "<index name>" 
+document_library = "https://<storage account name>.blob.core.windows.net/<storage account container>"
+
+# Add the text embedding RAG middleware to the OpenAI session used to retrieve LLM responses
+def setup_byod(deployment_id: str) -> None:
+    class BringYourOwnDataAdapter(requests.adapters.HTTPAdapter):
+
+        def send(self, request, **kwargs):
+            request.url = f"{openai.api_base}/openai/deployments/{deployment_id}/extensions/chat/completions?api-version={openai.api_version}"
+            return super().send(request, **kwargs)
+
+    session = requests.Session()
+
+    # Mount a custom adapter which will use the extensions endpoint for any call using the given `deployment_id`
+    session.mount(
+        prefix=f"{openai.api_base}/openai/deployments/{deployment_id}",
+        adapter=BringYourOwnDataAdapter()
+    )
+
+    openai.requestssession = session
+
+setup_byod(deployment_id)
+
+st.title('Generative AI using Azure AI Search Vector Embeddings')
+
+@st.cache_data
+def fetch_response(prompt_text):
+  completion = openai.ChatCompletion.create(
+    messages=prompt_text,
+    deployment_id=deployment_id,
+    dataSources=[ 
+        {
+            "type": "AzureCognitiveSearch",
+            "parameters": {
+                "endpoint": search_endpoint,
+                "key": search_key,
+                "indexName": search_index_name,
+            }
+        }
+    ])
+  
+  return completion
+
+# this is the text box where the user enters their question
+# Example questions:
+#    What models does directive 2019-25-55 affect? Format as a bulleted list.
+#    What is the aviation authority for Israel?
+#    what is cirrus design's contact information?
+prompt = st.text_area("I'm an expert in FAA Airworthiness Directives. What do you want to know?")
+
+# Present an button the user can click to send the prompt to the OpenAI LLM for a response
+# Azure AI Search will:
+#   1. create a vector embedding of the prompt
+#   2. Search the index for vectors close to the prompt
+#   3. Extract chunks from the vector index and add them to the prompt as context
+#   4. Forward the updated prompt to the OpenAI LLM for a response
+if st.button("Ask"):
+    data_load_state = st.text('Loading Response...')
+    message_text = [{"role": "user", "content": prompt}]
+    completion = fetch_response(message_text)
+    
+    # Write LLM Response to output
+    data_load_state.text("Done!")
+    st.write(completion.choices[0].message.content)
+    
+    # Present a button to read the original PDF document 
+    context_json_string = completion.choices[0].message.context.messages[0].content
+    context_json = json.loads(context_json_string)
+    doc_url = f"{document_library}/{context_json['citations'][0]['title']}"
+    st.link_button("Read Source Document", doc_url)