completions

Endpoint Documentation: /v1/completions

This endpoint allows you to make a POST request to complete a task or answer a question with a JSON output.

Request

• HTTP Method: POST
• URL: https://api.lamini.ai/v1/completions
• Headers:
• Authorization: Bearer <LAMINI_API_KEY>
• Content-Type: application/json
• Example Body (JSON):

{
    "model_name": "<YOUR_MODEL_NAME>",
    "prompt": "What is the hottest day of the year?",
    "out_type": {"Answer": "string"},
}

Parameters:

• model_name: str, the name of your base or finetuned model
• prompt: str or List[str] for a batch, ex.

      [
          "How old is Carl, the llama with a hat?",
          "How old is Paul, the llama with a hat?",
      ]
  

• out_type: Dict[str, str]. Type Schema of the output. Ex.

      {
          "age": "int",
          "units": "str",
      }
  

  The valid types are str, int, float, and bool.

Response

If the web request is successful, you will see a response with an answer to the provided questions like below:

• Success Status Code: 200
• Body (JSON):
• Output will be formatted as specified by the out_type argument passed in above.

    [
        {
            "age":10,
            "units":"years"
        },{
            "age":10,
            "units":"years"
        }
    ]
  

Otherwise, the request will return an error code, and the response json will contain specific error details like invalid token or incompatible data.

Example

Request

curl --location 'https://api.lamini.ai/v1/completions' \
--header 'Authorization: Bearer <LAMINI_API_KEY>' \
--header 'Content-Type: application/json' \
--data '{
    "model_name": "meta-llama/Llama-2-7b-chat-hf",
    "prompt": [
            "How old is Carl, the llama with a hat?",
            "How old is Paul, the llama with a hat?"
        ],
    "out_type": {
        "age": "int",
        "units": "str"
    }
}'

Response

Note the result is a hash, so the order of keys may be different from below.

[{"age":10,"units":"years"},{"age":10,"units":"years"}]