openai-poc-az-func

poc using the Azure OpenAI cognitive service (chatgpg 3.5 turbo) and llama_index to build and index to use with the LLM

How-to

Pull this project and create your virtual env, install packages and you can simply run func start (see prerequisites)

python3 -m venv .venv
source .venv/bin/activate
pip install openai azure-identity azure-keyvault-secrets llama-index azure-storage-blob langchain

Then once you finish the setup for configs bellow you can simply func start and use your function service.

Deploy to Azure.

Using the Azure extension you can selection your function and click on Deploy ... make sure to select the appropriate subscription and then you will be prompted with a list of applications you can deploy to, select the one that is tied to your proper resource group. You can also, once deployed, pushs your local.settings.json in the cloud directly (as env settings, on the App screen you can view those in the Configuration menu).

Pre-requisites

VSCode with Azure Dev tools such as anything needed to run app dev locally.
.NET runtime ..
etc.

Set KeyVault setting to your local.settings.json we will be using DefaultAzureCredential to conect to Key Vault.

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "",
    "FUNCTIONS_WORKER_RUNTIME": "python",
    "KEY_VAULT_NAME": "ScDC-CIO-DTO-Infra-kv",
    "OPENAI_ENDPOINT_NAME":"scdc-cio-dto-openai-poc-oai",
    "OPENAI_DEPLOYMENT_NAME":"gpt-35-poc-dev"
  }
}

Documentation

python langchain documentation
llamahub to get custom loaders
llama_index documentation on how to build the index and load it

dto-btn / openai-poc-az-func Goto Github PK