# Preparing Azure Functions

What is azure functions?

Azure Functions is a serverless computing service used for a variety of scenarios including microservices, data processing, and task automation.

Event-driven
Azure Functions provides an event-driven programming model where functions are executed in response to triggers
For example, HTTP Triggers can be requests, new messages to the message queue, database changes, etc.

Supported triggers and bindings
Various triggers and bindings are supported, including HTTP triggers, timer events, Blob storage changes, and receiving queue messages.
This allows for easy integration with external resources

Support for multiple programming languages
Azure Functions supports different programming languages ​​(C#, JavaScript, Python, PowerShell, etc.), allowing developers to choose the language of their choice. You can create a function by

## How to create functions
Since functions cannot be created using Python from the Portal, create functions using VSCode (text editor).

### About HTTP triggers
HTTP Trigger is a type of trigger type used in Azure Functions, which allows you to create functions that respond to external HTTP requests.
When using HTTP Trigger, processing defined in Azure Functions is executed in response to HTTP requests from external systems or clients.

* What are HTTP requests POST and GET? [click here](https://medium-company.com/http-get-post-%E9%81%95%E3%81%84/)
* If you want to know what you can receive with HTTP Trigger [Click here](https://learn.microsoft.com/ja-jp/python/api/azure-functions/azure.functions.httprequest?view=azure-python)

### Purpose
* Be able to implement and debug Azure Functions template functions
* By implementing Cognitive Search and AOAI processing in Azure Functions functions and skipping the question text with Postman, "Receive a question with an HTTP trigger and start a function" → "Search internal documents based on a question with Cognitive Search" → You will be able to create a series of steps in which you can send internal documents and questions to AOAI and have them respond.

## [Configure your environment](https://learn.microsoft.com/en-us/azure/azure-functions/create-first-function-vs-code-python?pivots=python-mode-decorators)
Before you begin, make sure that you have the following requirements in place:

An Azure account with an active subscription. Create an account for free.

Python versions that are supported by Azure Functions. For more information, see How to install Python.

Visual Studio Code on one of the supported platforms.

The Python extension for Visual Studio Code.

The Azure Functions extension for Visual Studio Code, version 1.8.1 or later.

The Azurite V3 extension local storage emulator. While you can also use an actual Azure storage account, this article assumes you're using the Azurite emulator.

### Execution steps

* [Create Azure Function Resource From Portal](https://hackmd.io/q094aW8ER7qhF_gj4ugTqg) for hosting, select consumption (serverless)
OR
* [Create Azure Function Resources from VScode](https://balconia.backlog.jp/alias/wiki/1331431)
* [Official Demo JP](https://learn.microsoft.com/ja-jp/azure/azure-functions/create-first-function-vs-code-python?pivots=python-mode-decorators)
* [Official Demo EN](https://learn.microsoft.com/en-us/azure/azure-functions/create-first-function-vs-code-python?pivots=python-mode-decorators)

2. Create an HTTP trigger function and incorporate the Cognitive Search and AOAI processing you have learned so far into the function

You have to replace two files:
0. local.settings.json should as below
There is a automatic way to fix ... I need to revisit

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "DefaultEndpointsProtocol=https;AccountName=functionappstorageaoi;AccountKey=erB+uVvn1LsWNOZ7UqwIVIOtsoh9Xo0PQ0fkqtWXwykhgjxb0p4Lum8WBWu3ctqLdPULSk7PdcHM+AStAQ/yTA==;EndpointSuffix=core.windows.net",
    "FUNCTIONS_WORKER_RUNTIME": "python",
    "AzureWebJobsFeatureFlags": "EnableWorkerIndexing",
    "FUNCTIONS_EXTENSION_VERSION": "~4",
    "WEBSITE_CONTENTAZUREFILECONNECTIONSTRING": "DefaultEndpointsProtocol=https;AccountName=functionappstorageaoi;AccountKey=erB+uVvn1LsWNOZ7UqwIVIOtsoh9Xo0PQ0fkqtWXwykhgjxb0p4Lum8WBWu3ctqLdPULSk7PdcHM+AStAQ/yTA==;EndpointSuffix=core.windows.net",
    "WEBSITE_CONTENTSHARE": "aoi-search-app92e6"
  }
}

2. requirements.txt with the following contents

azure-functions
azure-search-documents==11.4.0b6
openai[datalib]==0.27.8

# additional libraries
Flask==2.0.2
gunicorn==21.2.0
Werkzeug==2.2.2
# openai==0.28.1
# azure-search-documents==11.4.0b6
# azure-common==1.1.28
# azure-core==1.29.4
numpy

3. another is function_app.py by following contents

import azure.functions as func
from azure.search.documents import SearchClient
from azure.core.credentials import AzureKeyCredential
import logging
import json
from azure.search.documents.models import QueryType
import openai

app = func.FunctionApp(http_auth_level=func.AuthLevel.ANONYMOUS)

base_resource_name = "oai-jp-gpt35"
search_service_name = "cognitivesearch-jpeast-aoi-standard" # Cognitive Searchのデプロイ名

search_index_name = "search_index_name" # index名
search_api_key = "api_key" # cognitive search/search service  APIキー
search_endpoint = f"https://{search_service_name}.search.windows.net" # Endポイント
vector_field_name = "vector" # index内でベクトルを格納しているフィールド名 特に変更していなければ「vector」
context_field_name = "chunk" # index内でchunkを格納しているフィールド名 特に変更していなければ「chunk」
semantic_vector_name = "vector-1706155385792-semantic-configuration"

openai.api_type = "azure" # AOAIを使用するためazure
openai.api_base = f"https://{base_resource_name}.openai.azure.com/" #endポイント
openai.api_version = "2023-07-01-preview" #Azure AI Studioのチャットプレイグランドで送信例のコード表示から
openai.api_key = "api_key" # APIキー
embedding_deploy_name = "text-embedding-AOI" # AOAIのembeddingモデルのデプロイ名
chat_deploy_name = "gpt35" # AOAIのChatモデルのデプロイ名

search_client = SearchClient(endpoint=search_endpoint,
                            index_name=search_index_name,
                            credential=AzureKeyCredential(search_api_key))


# endpoint = "https://aid-cosmosdb.documents.azure.com:443/"
# key = "<azure cosmos db apikey>"


def search(query_text):
    query_vector = openai.Embedding.create(engine=embedding_deploy_name, input=query_text)["data"][0]["embedding"] #queryベクトルに変換
    if query_text:
        
        results = list(
                search_client.search(
                    query_text,
                    filter=None,
                    query_type=QueryType.SEMANTIC,
                    query_language="ja-jp",  # 日本語の場合は ja-jp
                    query_speller="none",
                    semantic_configuration_name=semantic_vector_name,
                    top=3,
                    query_caption="extractive|highlight-false" if True else None,
                    vector=query_vector,
                    top_k=50 if query_vector else None,
                    vector_fields=vector_field_name if query_vector else None
                )
            )
        return results



def chat(query_text, results):
    if results:

        context = results[0][context_field_name]
        response = openai.ChatCompletion.create(
            engine=chat_deploy_name,
            messages=[{
                    "role": "user",
                    "content": f"{query_text}/n以下のcontextを使って質問に回答してください。/n### context ###/n{context}"
                    # "content": "shree"
                }],
            temperature=0,
            max_tokens=2000,
            top_p=0.95,
            frequency_penalty=0,
            presence_penalty=0,
            stop=None
        )
        r = response.choices[0].message['content']
        return r

@app.route(route="http_trigger")
def http_trigger(req: func.HttpRequest) -> func.HttpResponse:
    logging.info('Python HTTP trigger function processed a request.')


    body = req.get_json()
    message = body['message']

    
    search_result = search(message)
    
    output_text = chat(message, search_result)

    if body:
        # UnicodeDecodeError with the following code
        #  'output_text': output_text
        logging.info(output_text)
        # Convert bytes to string before JSON serialization
        output_text_str = output_text

        return func.HttpResponse(json.dumps({
        'message': message,
        'output_text': output_text 
        }),
        status_code=200,
        mimetype="applicatuin/json"
        )
    else:
        return func.HttpResponse(
             "This HTTP triggered function executed successfully. Pass a name in the query string or in the request body for a personalized response.",
             status_code=200
        )

note: a bit confusing here
Sign in to your visual studio code and follow the instruction provided to create azure function with http trigger

## Run the function app to check
Click F1 > Azurite start or click debug in WORKSPACE >Local Project > Functions > Error

Function_app.py and requirements.txt created in execution step 2 Replace the following files with

Note: function may gives error when you click the url (value error)
Continue to postman debug
3. Postman Debug
[Postman](https://www.postman.com/downloads/)

Note: Web version needs registration can use gmail
Note: Application version doesnt requires registration

* Download and create workspace as below and change GET to POST


2. Switch to Body and raw with POST


3. Set up POST URL generated from azure function
eg. http://localhost:7071/api/http_trigger
create a message in json format and send

4. Error on the function_app
Fix implemented

   if body:
        # UnicodeDecodeError with the following code
        #  'output_text': output_text
        logging.info(output_text)
        # Convert bytes to string before JSON serialization
        output_text_str = output_text

        return func.HttpResponse(json.dumps({
        'message': message,
        'output_text': output_text 
        }),
        status_code=200,
        mimetype="applicatuin/json"
        )
    else:
        return func.HttpResponse(
             "This HTTP triggered function executed successfully. Pass a name in the query string or in the request body for a personalized response.",
             status_code=200
        )

5. Preview of result

Succesfully message received:

Now it is ready to be deployed