HTTP request failed with status: 404 when trying to connect to ooba

[Bibrrr]

Hello :)
I have an issue with this app when it comes to connection to ooba with openai api.
I changed port to 5000 in text-default.json config file to match port with that from ooba. Bot turns on just fine, it also appears to connect to ooba. However when messaging bot in discord, I get this

I'm trying to figure out if it has something to do with openklyde, or openai api extension itself. Extention loads without that /v1 at the end

[badgids]

Looks like your missing a trailing / after v1 for your API address.

…

On Fri, Apr 19, 2024, 9:16 AM Bibrrr ***@***.***> wrote: Hello :) I have an issue with this app when it comes to connection to ooba with openai api. I changed port to 5000 in text-default.json config file to match port with that from ooba. Bot turns on just fine, it also appears to connect to ooba. However when messaging bot in discord, I get this obraz.png (view on web) <https://github.com/badgids/OpenKlyde/assets/115658726/58dfd243-cbdd-4bad-a456-5d3ab095b735> I'm trying to figure out if it has something to do with openklyde, or openai api extension itself. Extention loads without that /v1 at the end obraz.png (view on web) <https://github.com/badgids/OpenKlyde/assets/115658726/2a67f650-c594-4b42-b014-f89b5977d564> — Reply to this email directly, view it on GitHub <#11>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABC7MFZRUE6UKREGJNBPOOTY6EKMBAVCNFSM6AAAAABGPFRHZKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGI2TEOJYGI2TEOA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Bibrrr]

That's a silly typo... but it's still not working

I noticed that when you run koboldcpp there's an actual api website when you go to 127.0.0.1:5001/api. I assume there should be something similar on 127.0.0.1:5000/api, but there's nothing

[badgids]

It's been a minute since I used Ooba. Did you launch Textgen-UI with the --listen --api and the --openai extension?

For Ooba the API endpoint is http://127.0.0.1:5000/v1/

It was correct the first go 'round, just missing the trailing / after v1

https://github.com/oobabooga/text-generation-webui/wiki/12-%E2%80%90-OpenAI-API

[Bibrrr]

I completely forgot about /docs... sorry about that.
I think I see where's the issue. Your bot is looking for localhost:5000/api/v1/model, while in ooba it's localhost:5000/v1/models

for text generation in kobold it's 127.0.0.1:5001/api/v1/generate while in ooba it's 127.0.0.1:5000/v1/chat/completions . That's why it's throwing 404 errors.

I will look into the code and check if changing those values helps

[Bibrrr]

It's getting somewhere !
I changed text-default config to the following

and it succesfully sends a request to ooba and generates text. There's a new issue... It looks like it can't retrieve data from openai api.
I tried multiple things. I mostly focued on line 380 in bot.py file. I changed data = llm_response['results'][0]['text'] to data = llm_response['choices'][0]['object'] as it's suggests in ooba docs. Now I get this error

I also tried to change like 380 to data = llm_response['choices'][0]['message']['content'] as per openai api docs. It errors out at message (also included in 2nd screenshot).

Further changes that I tried...
I tried to change "generation" in config to "generation": "chat/completions". It gives HTML error 422. With no text generation from ooba

[badgids]

Yeah it looks like Ooba has changed some things in the API from when I first wrote this program last year.

[badgids]

If you get this working again with Ooba before I have some more time to look into it, put in a PR. If we have to change so much it breaks the other API's, we'll just put the Ooba version in it's own branch.

It's getting somewhere !
I changed text-default config to the following

and it succesfully sends a request to ooba and generates text. There's a new issue... It looks like it can't retrieve data from openai api.
I tried multiple things. I mostly focued on line 380 in bot.py file. I changed data = llm_response['results'][0]['text'] to data = llm_response['choices'][0]['object'] as it's suggests in ooba docs. Now I get this error

I also tried to change like 380 to data = llm_response['choices'][0]['message']['content'] as per openai api docs. It errors out at message (also included in 2nd screenshot).

Further changes that I tried...
I tried to change "generation" in config to "generation": "chat/completions". It gives HTML error 422. With no text generation from ooba