nlpcloud / nlpcloud-js Goto Github PK

Hey, just wanted to provide some feedback.

It'd be really convenient if you were to put in an enum of sorts that contained every model that's available by request type.

Would save a trip to the website and help cleanup user code.

Running the example for question answering:

const NLPCloudClient = require('nlpcloud');

const client = new NLPCloudClient({ model: 'finetuned-llama-2-70b', token: '<redacted>', gpu: true })
client.question({
    question: `How many threats are blocked per day?`,
    context: `We have built security into every component of the network. Cloudflare’s 155 Tbps network blocks an average of 70 billion threats per day, including some of the largest DDoS attacks ever recorded. Each and every login, request, and response that goes through our network strengthens the machine learning that we apply to detect and block threats at the edge, before they ever reach your organization.`
}).then(function (response) {
    console.log(response.data);
}).catch(function (err) {
    console.error(err.response.status);
    console.error(err.response.data.detail);
});

throws the following error:

ReferenceError: params is not defined
    at Client.question (C:\Users\user\Documents\AI Answering\node_modules\nlpcloud\index.js:244:20)
    at Object.<anonymous> (C:\Users\user\Documents\AI Answering\index.js:4:8)
    at Module._compile (node:internal/modules/cjs/loader:1105:14)
    at Object.Module._extensions..js (node:internal/modules/cjs/loader:1159:10)
    at Module.load (node:internal/modules/cjs/loader:981:32)
    at Function.Module._load (node:internal/modules/cjs/loader:822:12)
    at Function.executeUserEntryPoint [as runMain] (node:internal/modules/run_main:77:12)
    at node:internal/main/run_main_module:17:47

Tried out the API like this and it fails with 404 Request failed with status code 404.

It works if I change the model to roberta-base-squad2.

const NLPCloudClient = require('nlpcloud');
const apiToken = 'token';
const client = new NLPCloudClient('fast-gpt-j',apiToken);

async function test11() {
  const response = await client.question(`When can plans be stopped?`,
    `All NLP Cloud plans can be stopped anytime. You only pay for the time you used the service. In case of a downgrade, you will get a discount on your next invoice.`)
  console.log('response', response);
}

test11().catch(console.error);

It would be awesome if you could include typings for TypeScript either as part of this package (preferably) or alternatively as a DT.

Thanks for providing an awesome NLP service!

the translation function does not allow me to specify source and target languages of the text, it only takes the text as a parameter despite the examples in the endpoints documentation showing 'text' , 'source' , 'target' as the three parameters of this function.
Am I missing something here?

https://docs.nlpcloud.com/?javascript#translation

Note: I got this probelm while using the nlpcloud package in my NestJs application

Hello,
I really like NL cloud and wanted to give you a quick feedback in regards to the typing issues I'm facing. I'm using the summarization function, and the return type is hard to use and a bit confusing:

 Promise<{
        status: number;
        statusText: string;
        data: {
            summary_text: string;
        };
    } | {
        status: number;
        statusText: string;
        data: {
            url: string;
        }
    }

TS is confused when I get the response back as it does not know if data contains an url or a summary_text. I'm not sure which case we get an url or a string, but would be nice to know in advance, otherwise a quick fix could look like this:

 Promise<{
        status: number;
        statusText: string;
        data: DataTypeText;
    } | {
        status: number;
        statusText: string;
        data: DataTypeUrl;
    }

and simply export the data type so we can type it when we get back the response ourself, right now I'm simply using // @ts-ignore which I would like to avoid.

Even without out the account limits endpoint and built-in rate limiting requested in #3 , the option to bring our own Axios (or fetch since it will be future node core?) where we have implemented the rate limit at the client would be have some nice ergonomics.

Context argument is still missing even after update to 1.0.24.

Thank you

the generation endpoint is expecting

• temperature : float,
• length_penalty : float,
• repetition_penalty : float,
• top_p : float,

but in the typescript package, the generation constructor has these values defined as type 'number' which is causing a 422 error from the server

the constructor also does not have top_p as a parameter at all

Hi! I got this sting in response for client.asr(fileUrl), which is hard to parse.
{ url: 'https://api.nlpcloud.io/v1/get-async-result/02762942-5796-4230-9558-b00f6d6c1b6d'}

Docs says that I should get a regular JSON string with quotes around "url" like {"url":"https://api.nlpcloud.io/v1/get-async-result/86cfbf00-f442-40b1-bb89-275d7d32fd48"}

Is this an error on your side?

Your latest release 1.0.49 introduced a functional change in the way the parameters are passed into the NLPCloudClient interface, and was released as a patch version update.

This broke our integration with NLPCloud, and loss of trust in the service, trying to figure out what is happening.

I would suggest you do the following:

• Whenever you add breaking changes in the public interface of this library, add a major library update so that people can upgrade your library with confidence.
• Use minor when adding new functionality and patch when fixing bugs.
• Maintain a proper changelog in this repo, or document significant changes like this in your Github releases.

Dear nlpcloud,

Is there the possibility to train a custom model, providing the necessary computation power required?

We get this data back from the embeddings() method:

response.data = {"embedding": [....]}

instead of the TS definition:

response.data = {"score": [.....]}

I think it would be slick if the the client could discover the account limits, such as rate limits the token is approved for and to configure Axios to use that limit.

This just a nice to have in being "good" consumers without implementing 429 handling.