Updates to 04-deploy-ai

labs/03-orchestration/04-ACS/acs-sk-csharp.ipynb

@@ -1070,9 +1070,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
-   "language": "python",
-   "name": "python3"
+   "display_name": ".NET (C#)",
+   "language": "C#",
+   "name": ".net-csharp"
   },
   "language_info": {
    "codemirror_mode": {
@@ -1086,7 +1086,18 @@
    "pygments_lexer": "ipython3",
    "version": "3.11.6"
   },
-  "orig_nbformat": 4
+  "orig_nbformat": 4,
+  "polyglot_notebook": {
+   "kernelInfo": {
+    "defaultKernelName": "csharp",
+    "items": [
+     {
+      "aliases": [],
+      "name": "csharp"
+     }
+    ]
+   }
+  }
  },
  "nbformat": 4,
  "nbformat_minor": 2

labs/04-deploy-ai/01-backend-api/aais-lc-python-api/README.md

@@ -0,0 +1,253 @@
+# 04 - Deploy ACS Langchain Python API
+
+In this folder you will find a sample AI App that is built using Python, Langchain and Azure AI Search.
+
+The entire solution is in this folder, but we also have all the step by step instructions so you can see how it was built.
+
+## Complete Solution
+
+To test the version of the app in this folder, you should just be able to run the command below. It should read the environment variable values from the `.env` file located in the root of this repository, so if you've already configured that there shouldn't be anything else to do. Otherwise, you'll need to fill in the `.env` file with the necessary values.
+
+You can run the app using the command below.
+
+```bash
+uvicorn main:app --reload --host=0.0.0.0 --port=5291
+```
+
+## Step by Step Instructions
+
+### Create Python Project and Solution
+
+```bash
+mkdir aais-lc-python
+cd aais-lc-python
+```
+
+### Add Dependencies
+
+```bash
+echo "azure-core==1.30.1
+azure-identity==1.16.0
+azure-search-documents==11.4.0
+fastapi==0.110.1
+uvicorn==0.25.0
+openai==1.27.0
+langchain==0.1.19
+langchain-openai==0.1.3
+tiktoken==0.6.0
+python-dotenv==1.0.1
+chainlit==1.0.506" > requirements.txt
+```
+
+```bash
+pip install -r requirements.txt
+```
+
+### Create main.py
+
+```bash
+echo "" > main.py
+```
+
+The first thing we need to do is to add some import and from statements to the `main.py` file.
+
+```python
+import os
+import logging
+from langchain_openai import AzureOpenAIEmbeddings, AzureChatOpenAI
+from dotenv import load_dotenv
+from fastapi import FastAPI
+from pydantic import BaseModel
+from fastapi.responses import JSONResponse, HTMLResponse
+from langchain.prompts import PromptTemplate
+from azure.search.documents import SearchClient
+from azure.search.documents.models import VectorizedQuery
+from azure.core.credentials import AzureKeyCredential
+from langchain.chains import LLMChain
+from langchain_core.output_parsers import StrOutputParser
+```
+
+We configure some logging so we can see what's happening behind the scenes
+
+```python
+logging.basicConfig(format='%(levelname)-10s%(message)s', level=logging.INFO)
+```
+
+Next we read in the environment variables.
+
+```python
+if load_dotenv():
+    logging.info("Azure OpenAI Endpoint: " + os.getenv("AZURE_OPENAI_ENDPOINT"))
+    logging.info("Azure AI Search: " + os.getenv("AZURE_AI_SEARCH_SERVICE_NAME"))
+else: 
+    print("No file .env found")
+```
+
+Now we add the Embeddings and ChatCompletion instances of Azure OpenAI that we will be using.
+
+```python
+# Create an Embeddings Instance of Azure OpenAI
+azure_openai_embeddings = AzureOpenAIEmbeddings(
+    azure_deployment = os.getenv("AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME"),
+    openai_api_version = os.getenv("OPENAI_EMBEDDING_API_VERSION"),
+    model= os.getenv("AZURE_OPENAI_EMBEDDING_MODEL")
+)
+
+azure_openai = AzureChatOpenAI(
+    azure_deployment = os.getenv("AZURE_OPENAI_COMPLETION_DEPLOYMENT_NAME"),
+    temperature=0.1,
+    max_tokens=500
+)
+logging.info('Completed creation of embedding and completion instances.')
+```
+
+Next we start the app.
+
+```python
+# Start the App
+app = FastAPI()
+```
+
+Next we define the class and routing methods that will be used.
+
+```python
+class CompletionRequest(BaseModel):
+    Question: str
+
+class CompletionResponse(BaseModel):
+    completion: str
+```
+
+This route is for the root of the app, it will return a link to the Swagger UI.
+
+```python
+@app.get("/", response_class=HTMLResponse)
+def read_root():
+    return "<a href=docs>Swagger Endpoint: docs</a>"
+```
+
+This route is for the completion endpoint, it will take a question and return a completion. The question is first passed to Azure AI Search to get the top 5 results, then the question and results are passed to Azure OpenAI to get a completion.
+
+```python
+@app.post("/completion/", response_class=JSONResponse)
+def execute_completion(request: CompletionRequest):
+    # Ask the question
+    # The question is being passed in via the message body.
+    # request: CompletionRequest
+
+    # Create a prompt template with variables, note the curly braces
+    prompt = PromptTemplate(
+        input_variables=["original_question","search_results"],
+        template="""
+        Question: {original_question}
+
+        Do not use any other data.
+        Only use the movie data below when responding.
+        {search_results}
+        """,
+    )
+
+    # Search Vector Store
+    search_client = SearchClient(
+        os.getenv("AZURE_AI_SEARCH_ENDPOINT"),
+        os.getenv("AZURE_AI_SEARCH_INDEX_NAME"),
+        AzureKeyCredential(os.getenv("AZURE_AI_SEARCH_API_KEY"))
+    )
+
+    vector = VectorizedQuery(vector=azure_openai_embeddings.embed_query(request.Question), k_nearest_neighbors=5, fields="vector")
+
+    results = list(search_client.search(
+        search_text=request.Question,
+        query_type="semantic",
+        semantic_configuration_name="movies-semantic-config",
+        include_total_count=True,
+        vector_queries=[vector],
+        select=["title","genre","overview","tagline","release_date","popularity","vote_average","vote_count","runtime","revenue","original_language"],
+        top=5
+    ))
+
+    output_parser = StrOutputParser()
+    chain = prompt | azure_openai | output_parser
+    response = chain.invoke({"original_question": request.Question, "search_results": results})
+    logging.info("Response from LLM: " + response)
+    return CompletionResponse(completion = response)
+```
+
+### Test the App
+
+Now that we have all the code in place let's run it.
+
+Remember to create a `.env` file in the same folder as the `main.py` file and add the environment variables and values we used from the `.env` file in the Jupyter notebooks.
+
+```bash
+uvicorn main:app --reload --host=0.0.0.0 --port=5291
+```
+
+Once the app is started, open a browser and navigate to http://127.0.0.1:5291/docs
+>**Note:** the port number may be different to `5291`, so double check the output from the `uvicorn main:app` command.
+
+Click on the "POST /completion" endpoint, click on "Try it out", enter a Prompt, "List the movies about ships on the water.", then click on "Execute".
+
+### Build and Test Docker Image
+
+Let's now package the solution into a Docker Image so it can be deployed to a container service like Azure Kubernetes Serivce (AKS) or Azure Container Apps (ACA).
+
+First you'll need to create a Dockerfile
+
+```bash
+echo "" > Dockerfile
+```
+
+Next, paste the following into the Dockerfile
+
+```Dockerfile
+FROM python:3.11-slim AS builder
+WORKDIR /app
+ENV PYTHONDONTWRITEBYTECODE 1
+ENV PYTHONUNBUFFERED 1
+RUN apt-get update && apt-get install -y \
+  build-essential \
+  curl \
+  software-properties-common \
+  git \
+  && rm -rf /var/lib/apt/lists/*
+# Create a virtualenv to keep dependencies together
+RUN python -m venv /opt/venv
+ENV PATH="/opt/venv/bin:$PATH"
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Stage 2 - Copy only necessary files to the runner stage
+FROM python:3.11-slim
+WORKDIR /app
+COPY --from=builder /opt/venv /opt/venv
+ENV PATH="/opt/venv/bin:$PATH"
+COPY main.py .
+EXPOSE 5291
+CMD ["uvicorn", "main:app", "--host=0.0.0.0", "--port=5291"]
+```
+
+Next, we build the Docker Image.
+
+```bash
+docker build -t aais-lc-python:v1 .
+```
+
+Finally, we can test the image. We pass in the environment variable values as we don't want to have sensitive information embedded directly into the image.
+
+```bash
+docker run -it --rm \
+    --name aaislcpython \
+    -p 5291:5291 \
+    -e AZURE_OPENAI_API_KEY="<YOUR AZURE OPENAI API KEY>" \
+    -e AZURE_OPENAI_ENDPOINT="<YOUR AZURE OPENAI ENDPOINT>" \
+    -e OPENAI_API_VERSION="2024-03-01-preview" \
+    -e AZURE_OPENAI_COMPLETION_DEPLOYMENT_NAME="<YOUR AZURE OPENAI COMPLETIONS DEPLOYMENT NAME - e.g. gpt-35-turbo>" \
+    -e AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME="<YOUR AZURE OPENAI EMBEDDINGS DEPLOYMENT NAME - e.g. text-embedding-ada-002>" \
+    -e AZURE_OPENAI_EMBEDDING_MODEL="<YOUR AZURE OPENAI EMBEDDINGS MODEL NAME - e.g. text-embedding-ada-002>" \
+    -e AZURE_AI_SEARCH_SERVICE_NAME="<YOUR AZURE AI SEARCH SERVICE NAME - e.g. cognitive-search-service>" \
+    -e AZURE_AI_SEARCH_ENDPOINT="<YOUR AZURE AI SEARCH ENDPOINT NAME - e.g. https://cognitive-search-service.search.windows.net" \
+    -e AZURE_AI_SEARCH_INDEX_NAME="<YOUR AZURE AI SEARCH INDEX NAME - e.g. cognitive-search-index>" \
+    -e AZURE_AI_SEARCH_API_KEY="<YOUR AZURE AI SEARCH ADMIN API KEY - e.g. cognitive-search-admin-api-key>" \
+    aais-lc-python:v1
+```

labs/04-deploy-ai/01-backend-api/acs-lc-python-api/acs-lc-python/README.md → labs/04-deploy-ai/01-backend-api/aais-lc-python-api/aais-lc-python/README.md

@@ -16,5 +16,5 @@ pip install -r requirements.txt
 uvicorn main:app --reload --host=0.0.0.0 --port=5291
 ```
 
-3. Open a browser and navigate to http://localhost:8000/docs
+3. Open a browser and navigate to http://localhost:5291/docs
 4. Click on the "POST /completion" endpoint, click on "Try it out", enter a Prompt, then click on "Execute".

labs/04-deploy-ai/01-backend-api/aais-lc-python-api/aais-lc-python/main.py

@@ -0,0 +1,101 @@
+import os
+import logging
+from langchain_openai import AzureOpenAIEmbeddings, AzureChatOpenAI
+from dotenv import load_dotenv
+from fastapi import FastAPI
+from pydantic import BaseModel
+from fastapi.responses import JSONResponse, HTMLResponse
+from langchain.prompts import PromptTemplate
+from azure.search.documents import SearchClient
+from azure.search.documents.models import VectorizedQuery
+from azure.core.credentials import AzureKeyCredential
+from langchain.chains import LLMChain
+from langchain_core.output_parsers import StrOutputParser
+
+logging.basicConfig(format='%(levelname)-10s%(message)s', level=logging.INFO)
+
+# Load environment variables
+if load_dotenv():
+    logging.info("Azure OpenAI Endpoint: " + os.getenv("AZURE_OPENAI_ENDPOINT"))
+    logging.info("Azure AI Search: " + os.getenv("AZURE_AI_SEARCH_SERVICE_NAME"))
+else: 
+    print("No file .env found")
+
+azure_openai_api_key = os.getenv("AZURE_OPENAI_API_KEY")
+azure_openai_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
+openai_api_version = os.getenv("OPENAI_API_VERSION")
+azure_openai_completion_deployment_name = os.getenv("AZURE_OPENAI_COMPLETION_DEPLOYMENT_NAME")
+azure_openai_embedding_deployment_name = os.getenv("AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME")
+azure_ai_search_name = os.getenv("AZURE_AI_SEARCH_SERVICE_NAME")
+azure_ai_search_endpoint = os.getenv("AZURE_AI_SEARCH_ENDPOINT")
+azure_ai_search_index_name = os.getenv("AZURE_AI_SEARCH_INDEX_NAME")
+azure_ai_search_api_key = os.getenv("AZURE_AI_SEARCH_API_KEY")
+
+azure_openai_embeddings = AzureOpenAIEmbeddings(
+    azure_deployment = os.getenv("AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME"),
+    openai_api_version = os.getenv("OPENAI_EMBEDDING_API_VERSION"),
+    model= os.getenv("AZURE_OPENAI_EMBEDDING_MODEL")
+)
+
+azure_openai = AzureChatOpenAI(
+    azure_deployment = os.getenv("AZURE_OPENAI_COMPLETION_DEPLOYMENT_NAME"),
+    temperature=0.1,
+    max_tokens=500
+)
+logging.info('Completed creation of embedding and completion instances.')
+
+# Start the App
+app = FastAPI()
+
+class CompletionRequest(BaseModel):
+    Question: str
+
+class CompletionResponse(BaseModel):
+    completion: str
+
+@app.get("/", response_class=HTMLResponse)
+def read_root():
+    return "<a href=docs>Swagger Endpoint: docs</a>"
+
+@app.post("/completion/", response_class=JSONResponse)
+def execute_completion(request: CompletionRequest):
+    # Ask the question
+    # The question is being passed in via the message body.
+    # request: CompletionRequest
+
+    # Create a prompt template with variables, note the curly braces
+    prompt = PromptTemplate(
+        input_variables=["original_question","search_results"],
+        template="""
+        Question: {original_question}
+
+        Do not use any other data.
+        Only use the movie data below when responding.
+        {search_results}
+        """,
+    )
+
+    # Search Vector Store
+    search_client = SearchClient(
+        azure_ai_search_endpoint,
+        azure_ai_search_index_name,
+        AzureKeyCredential(azure_ai_search_api_key)
+    )
+
+    vector = VectorizedQuery(vector=azure_openai_embeddings.embed_query(request.Question), k_nearest_neighbors=5, fields="vector")
+
+    results = list(search_client.search(
+        search_text=request.Question,
+        query_type="semantic",
+        semantic_configuration_name="movies-semantic-config",
+        include_total_count=True,
+        vector_queries=[vector],
+        select=["title","genre","overview","tagline","release_date","popularity","vote_average","vote_count","runtime","revenue","original_language"],
+        top=5
+    ))
+
+    output_parser = StrOutputParser()
+    chain = prompt | azure_openai | output_parser
+    response = chain.invoke({"original_question": request.Question, "search_results": results})
+    logging.info("Response from LLM: " + response)
+    return CompletionResponse(completion = response)