#api-request #ai #llm #shorthand

gpt-batch-scribe

contains the GptBatchAPIRequest struct which helps create gpt4 batch requests

4 releases

0.1.3 Jul 31, 2024
0.1.2 Jun 28, 2024
0.1.1 Jun 28, 2024
0.1.0 Jun 28, 2024

#14 in #shorthand

Download history 293/week @ 2024-06-25 42/week @ 2024-07-02 16/week @ 2024-07-23 131/week @ 2024-07-30

147 downloads per month

MIT license

39KB
584 lines

gpt-batch-scribe

A Rust crate for handling batch API requests to GPT models. This crate provides structures and enumerations to construct, serialize, and deserialize JSON requests and handle errors appropriately.

Features

  • Define request structures with custom fields and methods.
  • Serialize and deserialize using Serde.
  • Handle various error types using a custom error tree.
  • Support for different GPT models and API endpoints.
  • Custom serialization for enums.

Example Usage

Basic Example

use gpt_batch_scribe::{GptBatchAPIRequest, HttpMethod, GptApiUrl, GptRequestBody, GptMessage, GptModelType};

let request = GptBatchAPIRequest::new_basic(
    1,
    "You are a helpful assistant.",
    "Hello world!"
);

println!("{}", request);

Structures and Enumerations

GptBatchAPIRequest

Represents the complete request structure.

  • custom_id: Identifier for the custom request.
  • method: HTTP method used for the request.
  • url: URL of the API endpoint.
  • body: Body of the request.

Methods

  • new_basic(idx: usize, system_message: &str, user_message: &str) -> Self: Creates a basic request.
  • new_with_image(idx: usize, system_message: &str, user_message: &str, image_b64: &str) -> Self: Creates a request with an image.

GptApiUrl

Enumeration of API URLs.

  • ChatCompletions: API endpoint for chat completions.

GptMessage

Represents individual message details in the request body.

  • role: Role of the participant (system/user).
  • content: Content of the message.

Methods

  • system_message(msg: &str) -> Self: Creates a system message.
  • user_message(msg: &str) -> Self: Creates a user message.
  • user_message_with_image(msg: &str, image_b64: &str) -> Self: Creates a user message with an image.

GptModelType

Supported model types.

  • Gpt4o: Model type GPT-4o.
  • Gpt4Turbo: Model type GPT-4 Turbo.

GptRequestBody

Details of the API request body.

  • model: Model used for the request.
  • messages: Array of messages exchanged in the request.
  • max_tokens: Maximum number of tokens to be used by the model.

Methods

  • default_max_tokens() -> u32: Default maximum tokens.
  • default_max_tokens_given_image(image_b64: &str) -> u32: Default maximum tokens when an image is included.
  • new_basic(system_message: &str, user_message: &str) -> Self: Creates a basic request body.
  • new_with_image(system_message: &str, user_message: &str, image_b64: &str) -> Self: Creates a request body with an image.

HttpMethod

Enumeration of possible HTTP methods.

  • Get: HTTP GET method.
  • Post: HTTP POST method.

Error Handling

ParseTokenDescriptionLineError

Enumeration for parsing token description line errors.

  • MissingToken
  • MissingDescription

TokenizerError

Enumeration for tokenizer errors.

  • TokenizerError(String)

GptBatchCreationError

Enumeration for batch creation errors.

  • OpenAIError(OpenAIError)
  • IOError(std::io::Error)
  • TokenizerError(TokenizerError)
  • ParseTokenDescriptionLineError(ParseTokenDescriptionLineError)
  • SerdeJsonError(serde_json::Error)

License

This project is licensed under the MIT License.

Dependencies

~15–26MB
~407K SLoC