Please add support for <a href="https://learn.microsoft.com/en-us/azure/cognitive-serv

API reference for Azure OpenAI service: <a href="https://learn

Thanks for the suggestion <a class="user-mention notranslate" data-hovercard-type="use

Completions Request Endpoint <p dir="aut

Currently blocked on async-openai supporting Azure: <a class="issue-link js-issue-link

Support Azure OpenAI service about gptcommit HOT 8 OPEN

LayZeeDK commented on August 19, 2024

Support Azure OpenAI service

from gptcommit.

Comments (8)

Laisky commented on August 19, 2024 2

You can try this project that converts Azure to OpenAI API.

https://github.com/haibbo/cf-openai-azure-proxy/blob/main/README_en.md

then edit the gptchat's config file(~/.config/gptcommit/config.toml)

Note: api_key should be azure openai service's key, not openai's api key.

from gptcommit.

LayZeeDK commented on August 19, 2024 1

API reference for Azure OpenAI service:

Completions and embeddings (Swagger)

from gptcommit.

zurawiki commented on August 19, 2024 1

Thanks for the context.

Currently, I'm blocked because I don't have access to the Azure OpenAI services. I applied but haven't heard back from Microsoft

from gptcommit.

zurawiki commented on August 19, 2024

Thanks for the suggestion @LayZeeDK. I'm interested by your request. What's motivating it?

I certainly agree that this project should be generalized to support different LLM APIs in the future. If we want to implement Azure, I'd be curious to learn:

Is there some feature Azure offers over OpenAI?
Will the setup for the Azure API be any different, in terms of authentication or data format?

from gptcommit.

LayZeeDK commented on August 19, 2024

Azure OpenAI service supports enterprise-grade security, policies, compliance, and regional availability. Would you be interested in API access?

from gptcommit.

zurawiki commented on August 19, 2024

Yup this is definitely something I'd be interested in making happen! I tagged the issue as help wanted

from gptcommit.

LayZeeDK commented on August 19, 2024

Completions

Request

Endpoint

gptcommit/src/openai.rs

Lines 66 to 71 in 6ee5bc5

 let request = self 

 .client 

 .post("https://api.openai.com/v1/completions") 

 .header("Content-Type", "application/json") 

 .header("Authorization", format!("Bearer {}", self.api_key)) 

 .json(&json_data);

Instead of https://api.openai.com/v1/completions, the endpoint URL would be in the format https://{resource_name}.openai.azure.com/openai/deployments/{deployment_id}/completions?api-version=<api_version> with 2022-12-01 being the latest API version so we must make the endpoint URL for completions configurable.

Headers

Instead of an Authorization header in the format Authorization: Bearer <api_key>, we would have the header api-key: <api_key>

Parameters

In the Azure OpenAI Service API, a completions request does not support a model parameter. Instead, this is represented by the {deployment_id} URL parameter.

Response

The OpenAI API response object has a usage property but the Azure OpenAI Service API does not.

The model property in the example response displays "ada". The models available for the Azure OpenAI Service are listed in Model Summary table and region availability. These are deployed then targeted in the {deployment_id} URL parameter for the completions endpoint. I'm not sure all the models supported by the Azure OpenAI Service are listed in

gptcommit/src/openai.rs

Lines 91 to 101 in 6ee5bc5

 pub(crate) fn get_prompt_token_limit_for_model(&self) -> usize { 

 match self.model.as_str() { 

 "text-davinci-003" => 4097, 

 "text-curie-001" => 2048, 

 "text-babbage-001" => 2048, 

 "text-ada-001" => 2048, 

 "code-davinci-002" => 8000, 

 "code-cushman-001" => 2048, 

 _ => 4097, 

 } 

 }

At least not if "ada" is a valid model response value.

from gptcommit.

zurawiki commented on August 19, 2024

Currently blocked on async-openai supporting Azure: 64bit/async-openai#32

from gptcommit.

Support Azure OpenAI service about gptcommit HOT 8 OPEN

Comments (8)

Completions

Request

Endpoint

Headers

Parameters

Response

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	let request = self
	.client
	.post("https://api.openai.com/v1/completions")
	.header("Content-Type", "application/json")
	.header("Authorization", format!("Bearer {}", self.api_key))
	.json(&json_data);

	pub(crate) fn get_prompt_token_limit_for_model(&self) -> usize {
	match self.model.as_str() {
	"text-davinci-003" => 4097,
	"text-curie-001" => 2048,
	"text-babbage-001" => 2048,
	"text-ada-001" => 2048,
	"code-davinci-002" => 8000,
	"code-cushman-001" => 2048,
	_ => 4097,
	}
	}