newsgpt

The project aims to aggregate news from common sources and give user summaries based on their setting.

Website (on azure cloud): https://news-gpt.azurewebsites.net/

Architecture overview:

flowchart LR;
 
    A[Web UI] -->|Load static page| B[Azure Web App];
    A -->|Load summary content| C[Azure HTTP Functions];
    E[Azure Timer Trigger] -->|Summary generation| G[Langchain and LLM] 
    G -->|Persist to DB| D;
    C -->|Read latest summary| D[CosmoDB];
    H[News Websites] --> E
  subgraph AZ functions
   C
   E
   G
  end

Loading

Resources:

• Medium post on azure openai langchai: https://medium.com/microsoftazure/azure-openai-and-langchain-eba69f18f050
• Azure OpenAI service: https://learn.microsoft.com/en-gb/azure/cognitive-services/openai/quickstart?tabs=command-line&pivots=programming-language-python
• Langchain summarized docs: https://python.langchain.com/en/latest/modules/chains/index_examples/summarize.html
• Azure function: https://learn.microsoft.com/en-us/azure/azure-functions/create-first-function-cli-python

Run server locally

Run azure function

install requirements

pip install -r requirements.txt

# terminal 1
azurite

# terminal 2
func start

Try APIs: http://localhost:7071/api/get_last_update

Run webapp

Since azure function needs cold start, it normally has a high response time on first call. For API call, it is okay, but for front page, it is not ideal. Thus, we will host our front page in a free-tier Azure web app. The web app is a simple express web app which serve html template and do api call to our azure functions.

To start locally

cd expressapp

npm start

Open UI: http://localhost:3000

Deploy to cloud

Deploy the azure function

After testing the functions locally, start the deployment to the cloud by following the steps in this link

In our setup:

• <RESOURCE_GROUP_NAME>: newsgpt
• <APP_NAME>: newsgpt-ai

Lastly deploy the function to the cloud with the following command:

# Update app settings
az functionapp config appsettings set --name newsgpt-ai --resource-group newsgpt --settings AzureWebJobsFeatureFlags=EnableWorkerIndexing

# then deploy
func azure functionapp publish newsgpt-ai

Deploy the webapp

• Create a new App Services using Azure App Service in VSCode. For our page, we created new App Services named news-gpt
• Choose expressapp/ as the resource and deploy it to Azure

The website is available at https://news-gpt.azurewebsites.net/

Price Forecast

• Azure function: https://azure.microsoft.com/en-us/pricing/details/functions/
• Azure Cosmos DB: https://azure.microsoft.com/en-us/pricing/details/cosmos-db/
• Azure static web app: https://azure.microsoft.com/en-us/pricing/details/app-service/static/
• Azure OpenAI: https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/

Service	Price	Unit	Description
Azure Function	0.0	Execution	1 million executions per month
Azure Cosmos DB	0.0	Execution	1,000 request units per-second provisioned throughput with 25 GB storage
Azure web app	0.00	Execution	60 mins CPU time daily
Azure OpenAI	0.02	1000	tokens, $0.3 every 2 hours

Notes:

• Selenium is not supported in Azure functions, due to the use of webdriver chrome.
• The prompt is with bullet point, however it is not shown as output. Will need more debugging.