Skip to content

OpenAI Chat

OpenAIChat

Bases: LocalStep

Interaction with OpenAI service.

Model is represented by the OpenAIExecutor.

This step is a wrapper, extracting the prompt from the item, interact with the API, and saves the response to the answer key in the item.

Source code in ragfoundry/processing/local_steps/api/openai.py 
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
class OpenAIChat(LocalStep):
    """
    Interaction with OpenAI service.

    Model is represented by the `OpenAIExecutor`.

    This step is a wrapper, extracting the prompt from the item, interact with the API, and
    saves the response to the `answer` key in the item.
    """

    def __init__(self, model, instruction, prompt_key, answer_key, **kwargs):
        """
        Args:
            model (dict): Configuration for the OpenAIExecutor.
            instruction (str): Path to the system instruction file.
            prompt_key (str): Key to the prompt in the item.
            answer_key (str): Key to store the response.
        """
        super().__init__(**kwargs)

        self.model = OpenAIExecutor(**model)
        self.prompt_key = prompt_key
        self.answer_key = answer_key
        self.instruction = open(instruction).read()

    def process_item(self, item, index, datasets, **kwargs):
        answer = self.model.chat(item[self.prompt_key], self.instruction)
        item[self.answer_key] = answer
        return item

__init__(model, instruction, prompt_key, answer_key, **kwargs)

Parameters:

  • model (dict) –

    Configuration for the OpenAIExecutor.

  • instruction (str) –

    Path to the system instruction file.

  • prompt_key (str) –

    Key to the prompt in the item.

  • answer_key (str) –

    Key to store the response.

Source code in ragfoundry/processing/local_steps/api/openai.py 
16
17
18
19
20
21
22
23
24
25
26
27
28
29
def __init__(self, model, instruction, prompt_key, answer_key, **kwargs):
    """
    Args:
        model (dict): Configuration for the OpenAIExecutor.
        instruction (str): Path to the system instruction file.
        prompt_key (str): Key to the prompt in the item.
        answer_key (str): Key to store the response.
    """
    super().__init__(**kwargs)

    self.model = OpenAIExecutor(**model)
    self.prompt_key = prompt_key
    self.answer_key = answer_key
    self.instruction = open(instruction).read()