async-openai
async-openai copied to clipboard
64bit
Rust library for OpenAI
First off, this is a really well done project. Thank you for all your work. There is one feature missing that is preventing me from fully scratching my current OAI...
It seems that with the launch of GPT4.5 they made a minor change to some return types: ``` Error: Failed to retrieve assistant Caused by: failed to deserialize api response:...
I am unable to use file api with this piece of code ``` pub async fn upload_separate_batch_files( app_data: web::Data, prompts: Vec, ) -> Result { let mut file_ids = Vec::new();...
When playing around with the tool-call-stream example in combination with [mistral's API](https://console.mistral.ai/) I encountered a strange bug where the arguments where duplicated (e.g. instead of `{ "city": "Berlin" }` I...
If `stream = True`, we are noticing that a 400 returns (in the server logs): ``` StreamError("Invalid status code: 400 Bad Request") ``` On the client side (Python consumption): ```...
I'm trying to pass a function tool to a chat completion, but I want to add the "strict": true additional param, but it looks like additional params isn't supported on...
Hello, I wanted to bring up the idea of unpinning major dependencies. Upon comparing the latest releases, I noticed that version `0.27` is approximately 5 MB smaller than `0.26`. This...
The `OpenAIError` type is implemented using the `thiserror` crate. It implements `From` via the `#[from]` present in the Reqwest variant. https://github.com/64bit/async-openai/blob/c48e62e72b1c8bd9658af7e682c996ba5c93e083/async-openai/src/error.rs#L4-L28 I think that the `#[from]` attribute should also be...
If I understand the code correctly, the current mechanism used for dealing with rate limits relies on exponential backoff to try again until it succeeds or it runs out of...
