tjardoo / openai-client Goto Github PK

I get the following error:
thread 'tokio-runtime-worker' panicked at src/ai_api/api.rs:56:59:
called Result::unwrap() on an Err value: ParseError("invalid value: map, expected map with a single key")

The relevant function:

pub async fn create_image_with_prompt(prompt: &str) -> String {

    let env_variables = read_file(".env").unwrap();
    let api_key: String = (&env_variables["OPENAI_API_KEY"]).to_owned();

    let client = Client::new(api_key);
    let parameters = CreateImageParameters {
        prompt: "A cute baby dog".to_string(),
        model: Some("dall-e-3".to_string()),
        n: Some(1),
        quality: None,
        response_format: Some(ResponseFormat::Url),
        size: Some(ImageSize::Size1024X1024),
        style: None,
        user: None,
    };

    let result = client.images().create(parameters).await.unwrap();
    let paths = result.save("./images").await.unwrap();
    let models_result = client.models().list().await.unwrap();

    println!("{:#?}", models_result);
    println!("{:?}", paths);

    println!("{:#?}", result);
    return "request complete".to_string();
}

Greetings
When ChatCompletionParameters is serialized and stop is Some, it's serialized in "stop":{"String":"stop_text"} which causes an issue with the OpenAI APIs: EndpointError("{\n \"error\": {\n \"message\": \"{'String': '{thoughts'} is not valid under any of the given schemas - 'stop'\",\n \"type\": \"invalid_request_error\",\n \"param\": null,\n \"code\": null\n }\n}\n")

I think the issue is with how the StopToken enum is serialized.

https://openai.com/index/introducing-structured-outputs-in-the-api

https://platform.openai.com/docs/guides/structured-outputs

Whenever OpenAI returns a 502 or doesn't respond over the network, an unwrap within the dive client fails and crashes our entire web server. We are currently on openai-dive 0.4.6

The fix is probably changing the unwrap at

We may get to pring this soon, but wanted to create the issue for visibility.

https://platform.openai.com/docs/guides/safety-best-practices/end-user-ids

How do you get the tool call id, or the id of the tool?
Tool output has this field

tool_call_id: Option<String>

The ID of the tool call in the ‘required_action’ object within the run object the output is being submitted for.

But, the run object doesn't have a required_action object
https://platform.openai.com/docs/api-reference/runs/object

I'm currently using openai_dive in a project to simplify my usage of the OpenAI API.. One thing that might be useful to the wider user base would be access to the Rate Limiting Headers as described in the OpenAI Docs specifically they mention x-ratelimit-limit-requests, x-ratelimit-limit-tokens,x-ratelimit-remaining-requests,x-ratelimit-remaining-tokens,x-ratelimit-reset-requests, and x-ratelimit-reset-tokens. Those all sound very useful for backing off the API appropriately.
Would you consider exposing the return headers in the various response objects?

https://platform.openai.com/docs/api-reference/fine-tunes/create

• n_epochs
• batch_size
• learning_rate_multiplier
• prompt_loss_weight
• compute_classification_metrics
• classification_n_classes
• classification_positive_class
• classification_betas

cargo fmt -- --check
cargo fmt

https://platform.openai.com/docs/api-reference/batch

Hi, thanks for the work on this library.
How do I submit a run tool output after my RunStatus becomes RequiresAction?

Hi there,

I'm thinking of implementing function calling (https://openai.com/blog/function-calling-and-other-api-updates), but I don't want to step on toes if this is already being implemented. Are there any current plans to support function calling? If not, I'll add support.

https://platform.openai.com/docs/api-reference/fine-tuning

Hey @tjardoo

Are there any current plans to support OpenAI Vision API?

https://platform.openai.com/docs/guides/vision

Please let me know. Was thinking to raise a PR if so. Thank you.

At present, EndpointError is constructed using just the text of the response:

Ideally, this would also return the status code so that the calling code can decide what to do with the failure (i.e. if it's a 5xx error, the request could be retried; if it's a rate limit, the code could wait a bit.)

I'm also not sure validate_response (below that code) is called anywhere, and given that check_status_code will return an EndpointError for any successful responses, it's not clear that there could be an opportunity for it to be called. It seems like these two functions should be merged so that specific errors from the OpenAI API can be surfaced, while non-specific errors can be dealt with based on the status code.

thread '<unnamed>' panicked at /Users/sjud/.cargo/registry/src/index.crates.io-6f17d22bba15001f/openai_dive-0.3.5/src/v1/resources/shared.rs:109:18:

for reference

 x_ratelimit_limit_requests: value
                .get("x-ratelimit-limit-requests")
                .unwrap()

I can make a pull request for error handling here if helpful?