Use the OpenAI Python SDK Client with HTTP or HTTP/SSL

The OpenAI Python client works with the container when using HTTP or HTTP/SSL. You just need to specify the correct HTTP endpoint, along with a valid API KEY when using HTTP/SSL.

The OpenAI Python client requires Python 3.8 or greater.

1. Install Python 3.12 using the uv or equivalent Python version manager.

   cd ~
   
   curl -LsSf https://astral.sh/uv/install.sh | sh
   
   uv
   
   cd ~
   
   uv venv --python 3.12
   
   source .venv/bin/activate
   
   python -V
   
   mkdir /home/opc/python
   cd /home/opc/python

2. Install the OpenAI Python client.

   uv pip install openai

3. You can now create a Python program to list the available embedding models using HTTP. The API_KEY value is not checked with HTTP, but the API_KEY parameter is still needed.

   from openai import OpenAI
     
   my_url = "http://localhost:8080/v1"
   my_key = "Any string will do"
   
   client = OpenAI(base_url=my_url, api_key=my_key)
   
   models = client.models.list()
   for model in models:
      print(f"- {model.id}, {model.modelSize}, {model.modelCapabilities}")

   You can also create a Python program to create a vector using HTTP.

   from openai import OpenAI
     
   my_endpoint = "http://localhost:8080/v1"
   my_api_key  = "Any string will do"
   my_sentence = "Just some sample text."
   my_model    = "clip-vit-base-patch32-txt"
   
   client = OpenAI(base_url=my_endpoint, api_key=my_api_key)
   
   embeddings = client.embeddings.create(model=my_model,input=my_sentence)
   print(embeddings.data[0].embedding)

   Note:

   The /health and /metrics endpoints are helper endpoints and are not part of the OpenAI API.
4. You can now create a Python program to list the embedding models using HTTP/SSL.

   Note:

   Keep in mind, the following conditions must be true:

   • The protocol must be HTTPS.
   • The HTTPS Port must be correct, for example 8443.
   • The API_KEY must match the value of $SECRETS_DIR/api-key on the container machine.
   • The certificate must match the certificate in the $SECRETS_DIR/cert.pem file.
   • The OpenAI function must use the valid API_KEY and certificate values.

   from openai import OpenAI
   import httpx
   
   my_url = "https://your_FQDN:8443/v1"
   my_key = "Your_API_KEY_value"
   my_cert = "/home/opc/secrets/cert.pem"
   my_string = "Just some sample text."
   
   client = OpenAI(base_url=my_url,
                   api_key=my_key,
                   http_client=httpx.Client(verify=my_cert))
   
   models = client.models.list()
   for model in models:
      print(f"- {model.id}, {model.modelSize}, {model.modelCapabilities}")

   You can also create a Python program to create a vector embedding using HTTP/SSL.

   from openai import OpenAI
   import httpx
   
   my_url = "https://your_FQDN:9443/v1"
   my_key = "Your_API_KEY_value"
   my_cert = "/home/opc/secrets/cert.pem"
   my_model = "tinybert"
   my_string = "Just some sample text."
   
   client = OpenAI(base_url=my_url,
                   api_key=my_key,
                   http_client=httpx.Client(verify=my_cert))
   
   embeddings = client.embeddings.create(model=my_model,input=my_string)
   print(embeddings.data[0].embedding)