call_llama

call_code_llama(prompt: str) → str

Call a Code Llama API to generate SQL queries. API is hosted on AWS EC2 instance using VLLM.

Parameters:

prompt (str) – Prompt to generate SQL query from.

Raises:

HTTPError if the API request fails.

Returns:

Generated SQL query or “Error” if the API call fails.

Return type:

str

codellama.json

Configuration file for the Code Llama API parameters. This file should be a JSON file containing the following keys:

• max_tokens: Maximum number of tokens to generate. (int)

• temperature: Sampling temperature. (float)

• top_p: Top-p sampling parameter. (float)

• n: Number of sequences to generate. (int)

• frequency_penalty: Frequency penalty. (float)

• presence_penalty: Presence penalty. (float)

• aws_url: URL of the Code Llama API endpoint. (str)

Example:

{
  "max_tokens": 2048,
  "temperature": 0,
  "top_p": 1,
  "n": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "aws_url": "http://<public IPv4 address>:8000/generate"
}