Merge branch 'MSUSAzureAccelerators:main' into main

Author: TBunkley

06-First-RAG.ipynb

@@ -125,9 +125,10 @@
     "In `apps/frontend/` folder you will find the files necesary to build a simple Streamlit application that will have:\n",
     "\n",
     "1) <u>A Search Interface</u>: Using `utils.py` and `prompts.py` and streamlit functions\n",
-    "2) <u>A WebChat Interface</u>: Using the Bot Service Web Chat javascript library we can render the WebChat Channel inside Streamlit as an html component\n",
+    "2) <u>A BotService Chat Interface</u>: Using the Bot Service Web Chat javascript library we can render the WebChat Channel inside Streamlit as an html component\n",
+    "3) <u>A FastAPI Chat Interface</u>: Using LangServe/FastAPI as backend, we use streamlit components to provide a streaming chat interface (MORE ON THIS LATER)\n",
     "\n",
-    "Notice that in (1) the logic code is running in the Frontend Web App, however in (2) the logic code is running in the Backend Bot API and the Frontend is just using the WebChat channel from the Bot Service.\n",
+    "Notice that in (1) the logic code is running in the Frontend Web App, however in (2) and (3) the logic code is running in the Backend Bot API and the Frontend is just using the WebChat channel from the Bot Service.\n",
     "\n",
     "GO AHEAD NOW AND FOLLOW THE INSTRUCTIONS in `apps/frontend/README.md`"
    ]

09-BingChatClone.ipynb

@@ -8,20 +8,46 @@ This bot has been created using [LangServe](https://python.langchain.com/docs/la
 
 Below are the steps to run the LangServe Bot API as an Azure Wep App:
 
-1. We don't need to deploy again the Azure infrastructure, we did that already for the Bot Service API (Notebook 12). We are going to use the same App Service and just change the code.
+1. We don't need to deploy again the Azure infrastructure, we did that already for the Bot Service API (Notebook 12). We are going to use the same App Service, but we just need to add another SLOT to the service and have both APIs running at the same time. Note: the slot can have any name, we are using "staging".<br> In the terminal run:
+
+```bash
+az login -i
+az webapp deployment slot create --name "<name-of-backend-app-service>" --resource-group "<resource-group-name>" --slot staging --configuration-source "<name-of-backend-app-service>"
+```
+
+2. Zip the code of the bot by executing the following command in the terminal (**you have to be inside the apps/backend/langserve/ folder**):
 
-3. Zip the code of the bot by executing the following command in the terminal (**you have to be inside the apps/backend/langserve/ folder**):
 ```bash
 (cd ../../../ && zip -r apps/backend/langserve/backend.zip common) && zip -j backend.zip ./* && zip -j backend.zip ../../../common/requirements.txt && zip -j backend.zip app/*
 ```
-4. Using the Azure CLI deploy the bot code to the Azure App Service created on Step 2
+
+3. Using the Azure CLI deploy the bot code to the Azure App Service new SLOT created on Step 1:
+
 ```bash
-az login -i
-az webapp deployment source config-zip --resource-group "<resource-group-name>" --name "<name-of-backend-app-service>" --src "backend.zip"
+az webapp deployment source config-zip --resource-group "<resource-group-name>" --name "<name-of-backend-app-service>" --src "backend.zip" --slot staging
 ```
 
-5. **Wait around 5 minutes** and test your bot by running the next Notebook.
+4. Wait around 5 minutes and test your bot by running Notebook 14. Your Swagger (OpenAPI) definition should show here:
 
+```html
+https://<name-of-backend-app-service>-staging.azurewebsites.net/
+```
+
+5. Once you confirm that the API is working on step 4, you need to add the endpoint to the frontend page code. Go to `apps/frontend/pages`   and edit   `3_FastAPI_Chat.py`:
+
+```python
+    # ENTER HERE YOUR LANGSERVE FASTAPI ENDPOINT
+    # for example: "https://webapp-backend-botid-zf4fwhz3gdn64-staging.azurewebsites.net"
+
+    url = "https://<name-of-backend-app-service>-staging.azurewebsites.net" + "/agent/stream_events"
+```
+
+6. Re-deploy FrontEnd code: Zip the code of the bot by executing the following command in the terminal (you have to be inside the folder: **apps/frontend/** ):
+
+```bash
+zip frontend.zip ./* && zip frontend.zip ./pages/* && zip -j frontend.zip ../../common/*
+az webapp deployment source config-zip --resource-group "<resource-group-name>" --name "<name-of-frontend-app-service>" --src "frontend.zip"
+```
 
 ## (optional) Running in Docker
 

11-Smart_Agent.ipynb

@@ -169,7 +169,7 @@ def get_session_history(session_id: str, user_id: str) -> CosmosDBChatMessageHis
 
 chatgpt_search = ChatGPTTool(llm=llm,
                      name="chatgpt",
-                    description="use for general questions, profile, greeting-like questions and when the questions includes the term: chatgpt",
+                    description="useful when the questions includes the term: chatgpt",
                     verbose=False)
 
 tools = [doc_search, book_search, www_search, sql_search, chatgpt_search]

11.5-Smart_Agent-LangGraph.ipynb

@@ -12,12 +12,12 @@ Also includes a Search experience.
 [![Deploy To Azure](https://aka.ms/deploytoazurebutton)](https://portal.azure.com/#create/Microsoft.Template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Fpablomarin%2FGPT-Azure-Search-Engine%2Fmain%2Fapps%2Ffrontend%2Fazuredeploy-frontend.json)
 
 2. Zip the code of the bot by executing the following command in the terminal (you have to be inside the folder: apps/frontend/ ):
+
 ```bash
-zip frontend.zip ./*
-zip frontend.zip ./pages/*
-zip -j frontend.zip ../../common/*
+zip frontend.zip ./* && zip frontend.zip ./pages/* && zip -j frontend.zip ../../common/*
 ```
 3. Using the Azure CLI deploy the frontend code to the Azure App Service created on Step 2
+
 ```bash
 az login -i
 az webapp deployment source config-zip --resource-group "<resource-group-name>" --name "<name-of-frontend-app-service>" --src "frontend.zip"

12-Building-Apps.ipynb

@@ -3,7 +3,7 @@
 import streamlit.components.v1 as components
 
 # From here down is all the StreamLit UI.
-st.set_page_config(page_title="GPT Smart Agent", page_icon="📖", layout="wide")
+st.set_page_config(page_title="BotService Backend Bot", page_icon="🤖", layout="wide")
 # Add custom CSS styles to adjust padding
 st.markdown("""
         <style>
@@ -45,7 +45,7 @@
 - @bing, what movies are showing tonight in Seattle?
 - Please tell me a joke
     """)
-    
+
 st.markdown("""
         <style>
                .block-container {
@@ -155,4 +155,4 @@
     </script>
   </body>
 </html>
-""", height=800)
+""", height=800)

14-LangServe-API.ipynb

@@ -0,0 +1,162 @@
+import os
+import streamlit as st
+from langchain_core.messages import AIMessage, HumanMessage
+from langchain_openai import AzureChatOpenAI
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.prompts import ChatPromptTemplate
+from langserve import RemoteRunnable
+import uuid
+import requests
+import json
+import sys
+import time
+import random
+
+# Env variables needed by langchain
+os.environ["OPENAI_API_VERSION"] = os.environ.get("AZURE_OPENAI_API_VERSION")
+
+# app config
+st.set_page_config(page_title="FastAPI Backend Bot", page_icon="🤖", layout="wide")
+
+with st.sidebar:
+    st.markdown("""# Instructions""")
+    st.markdown("""
+
+This Chatbot is hosted in an independent Backend Azure Web App and was created using the Bot Framework SDK.
+The Bot Interface is just a window to a Bot Service app hosted in Azure.
+
+It has access to the following tools/pluggins:
+
+- Bing Search (***use @bing in your question***)
+- ChatGPT for common knowledge (***use @chatgpt in your question***)
+- Azure SQL for covid statistics data (***use @sqlsearch in your question***)
+- Azure Search for documents knowledge - Arxiv papers and Covid Articles (***use @docsearch in your question***)
+- Azure Search for books knowledge - 5 PDF books (***use @booksearch in your question***)
+- API Search for real-time covid statistics for US States, Countries and Continents (***use @apisearch in your question***)
+
+Note: If you don't use any of the tool names beginning with @, the bot will try to use it's own knowledge or tool available to answer the question.
+
+Example questions:
+
+- Hello, my name is Bob, what's yours?
+- @bing, What's the main economic news of today?
+- @chatgpt, How do I cook a chocolate cake?
+- @booksearch, what normally rich dad do that is different from poor dad?
+- @docsearch, Why Covid doesn't affect kids that much compared to adults?
+- @apisearch, What is the state with most covid deaths in USA?
+- @sqlsearch, How many people where hospitalized in Arkansas in June 2020?
+- @docsearch, List the authors that talk about Boosting Algorithms
+- @bing, what movies are showing tonight in Seattle?
+- Please tell me a joke
+    """)
+
+st.markdown("""
+        <style>
+               .block-container {
+                    padding-top: 1rem;
+                    padding-bottom: 0rem;
+                }
+        </style>
+        """, unsafe_allow_html=True)
+
+
+# ENTER HERE YOUR LANGSERVE FASTAPI ENDPOINT
+# for example: "https://webapp-backend-botid-zf4fwhz3gdn64-staging.azurewebsites.net"
+
+url = "https://<name-of-backend-app-service>-staging.azurewebsites.net" + "/agent/stream_events"
+
+
+def get_or_create_ids():
+    """Generate or retrieve session and user IDs."""
+    if 'session_id' not in st.session_state:
+        st.session_state['session_id'] = str(uuid.uuid4())
+    if 'user_id' not in st.session_state:
+        st.session_state['user_id'] = str(uuid.uuid4())
+    return st.session_state['session_id'], st.session_state['user_id']
+
+    
+def consume_api(url, user_query, session_id, user_id):
+    """Uses requests POST to talk to the FastAPI backend, supports streaming."""
+    headers = {'Content-Type': 'application/json'}
+    config = {"configurable": {"session_id": session_id, "user_id": user_id}}
+    payload = {'input': {"question": user_query}, 'config': config}
+    
+    with requests.post(url, json=payload, headers=headers, stream=True) as response:
+        try:
+            response.raise_for_status()  # Raises an HTTPError if the response is not 200.
+            for line in response.iter_lines():
+                if line:  # Check if the line is not empty.
+                    decoded_line = line.decode('utf-8')
+                    if decoded_line.startswith('data: '):
+                        # Extract JSON data following 'data: '.
+                        json_data = decoded_line[len('data: '):]
+                        try:
+                            data = json.loads(json_data)
+                            if "event" in data:
+                                kind = data["event"]
+                                if kind == "on_chat_model_stream":
+                                    content = data["data"]["chunk"]["content"]
+                                    if content:  # Ensure content is not None or empty.
+                                        yield content  # Two newlines for a paragraph break in Markdown.
+                                elif kind == "on_tool_start":
+                                        tool_inputs = data['data'].get('input')
+                                        if isinstance(tool_inputs, dict):
+                                            # Joining the dictionary into a string format key: 'value'
+                                            inputs_str = ", ".join(f"'{v}'" for k, v in tool_inputs.items())
+                                        else:
+                                            # Fallback if it's not a dictionary or in an unexpected format
+                                            inputs_str = str(tool_inputs)
+                                        yield f"Searching Tool: {data['name']} with input: {inputs_str} ⏳\n\n"
+                                elif kind == "on_tool_end":
+                                        yield "Search completed.\n\n"
+                            elif "content" in data:
+                                # If there is immediate content to print, with added Markdown for line breaks.
+                                yield f"{data['content']}\n\n"
+                            elif "steps" in data:
+                                yield f"{data['steps']}\n\n"
+                            elif "output" in data:
+                                yield f"{data['output']}\n\n"
+                        except json.JSONDecodeError as e:
+                            yield f"JSON decoding error: {e}\n\n"
+                    elif decoded_line.startswith('event: '):
+                        pass
+                    elif ": ping" in decoded_line:
+                        pass
+                    else:
+                        yield f"{decoded_line}\n\n"  # Adding line breaks for plain text lines.
+        except requests.exceptions.HTTPError as err:
+            yield f"HTTP Error: {err}\n\n"
+        except Exception as e:
+            yield f"An error occurred: {e}\n\n"
+
+
+# session state
+if "chat_history" not in st.session_state:
+    st.session_state.chat_history = [AIMessage(content="Hello, I am a GPT-3.5 bot hosted in Azure using FastAPI Streaming. How can I help you?")]
+
+    
+# conversation
+for message in st.session_state.chat_history:
+    if isinstance(message, AIMessage):
+        with st.chat_message("AI"):
+            st.write(message.content)
+    elif isinstance(message, HumanMessage):
+        with st.chat_message("Human"):
+            st.write(message.content)
+
+# user input
+
+session_id, user_id = get_or_create_ids()
+
+user_query = st.chat_input("Type your message here...")
+
+if user_query is not None and user_query != "":
+    st.session_state.chat_history.append(HumanMessage(content=user_query))
+
+    with st.chat_message("Human"):
+        st.markdown(user_query)
+
+    with st.chat_message("AI"):
+        response = st.write_stream(consume_api(url, user_query, session_id, user_id))
+
+    st.session_state.chat_history.append(AIMessage(content=response))