-
-
Notifications
You must be signed in to change notification settings - Fork 309
[Request] modify codebase to not require to
prefix for toapi
routes.
#316
Comments
In my opinion this is undesirable. The proxy contains the official API: https://platform.openai.com/docs/api-reference, /to/v1/chat/completions. The purpose is to allow users to distinguish between using the |
But it is not impossible to do this. The server distinguishes the user's Authenticate key type for forwarding. |
Would it be possible to potentially make it configurable via a command line argument? I personally would love this because it breaks a few applications where they just want a base URL instead of a completions url |
and I don't want to pay for the API when I already pay for pro :D |
The next plan is to develop an account pool management (but relatively speaking, a front-end WebUI is needed for account management. Unfortunately, I am not familiar with html and js). It is to map Access-Token to API-Key, and then use it with the official API. After merging, the path will be consistent with the official one and will be insensitive to users. |
Couldn't you do the api for this in the background, and if you dont have the time yourself, you could potentially have someone else make a frontend to call those endpoints exposed by ninja to map access tokens to api keys. If need be, I can create a very basic UI for managing these but as a disclaimer, I will be super basic but functional |
My plan is to only need a simple multi-account group management. Each group of accounts can generate several mapped API-Keys, and some pages for adding, deleting, checking and modifying will be enough. |
If you can provide me with an API which handles this in rust. I can then create a barebone UI for you. but I cant do anything unless I know what api calls to make |
It takes a while to design the data model |
That's fine, keep this issue open to track the progress, once your side of things are done you can let me know and I'll make something barebone |
OpenAI is facing the risk of employees running away. I am considering whether to continue the plan. |
with the recent changes and them getting the ceo back. do you think you will get back to this project? |
I'll be back, now |
For account Pooling via API, use OneAPI https://github.com/songquanpeng/one-api in-front of Ninja. OneAPI can load-balance across multiple channels. Each Channel can send a unique access token key to Ninja. Can even have multiple instances of Ninja, as you can set Ninja up as a proxy for each Channel. In my setup I have chat going straight to Ninja backend-api, but API I send to OneAPI first then back to Ninja. |
I don't like putting things related to different AI agents together |
I don't know what you want me to do with this issue? I am happy with the change you made so we could close it or do you want to keep it here? @gngpp |
stay here until done |
Can one-api automatically update access token? Or manually? |
From the reading I have done it should be possible to simply move the router setup for
toapi
above where you defineAxum does things in the order things are initialized / written. So if we do something like
the custom handler for completions will be prioritized under the normal v1 path. I am only making this suggestion as it appears to me it sometimes breaks third party applications where they require a base URL which is usually the one that first leads up to
/v1
.The text was updated successfully, but these errors were encountered: