feature(community): Enable using chat and llms without providing project/space/development id (lightweight engine IBM) #7781
Conversation
* Rename auth method in docs * Rename auth method in docs * Add deployment chat to chat class * Upadate Watsonx sdk * Rework interfaces in llms as well * Bump watsonx-ai sdk version * Remove unused code * Add fake auth
dosubot
bot
added
the
size:L
dosubot
bot
added
auto:documentation
dosubot
bot
added
size:M
FilipZmijewski
changed the title
| if (!("projectId" in fields || "spaceId" in fields || "idOrName" in fields)) | ||
| throw new Error( | ||
| "No id specified! At least id of 1 type has to be specified" | ||
| ); |
There was a problem hiding this comment.
Can we keep this check and only bypass it when the lightweight engine is in use?
There was a problem hiding this comment.
So that would enforce us to use a flag ex. lightweight: true. Unless user uses a lightweight engine the API is going to throw an error "No spaceId nor projectId is present" so this is taken care of there.
Also, we want to match with our Python Langchain SDK implementation. Assuming whenever id is not passed user chooses lightweight engine. If id of any type is mistakenly not passed the API will throw an error letting user know.
There was a problem hiding this comment.
@FilipZmijewski just to clarify, are you saying that the API does check this and therefore the check in our code is unnecessary? The benefit of having the check in our code is that it causes us to throw a nice/readable error. Is there no other way that we can infer that the user is using the lightweight engine?
By the way, feel free to reply in your native language if you're more comfortable that way. LLMs are very good at translating these days 😄
There was a problem hiding this comment.
@benjamincburns We would like to match the solution with what we have in SDK (both python and node.js) and also langchain IBM python meaning "no id provided -> lightweight engine in use". In case of lightweight engine unavailable the API will throw an error, quite clear to understand, Missing either space_id or project_id or wml_instance_crn. One thing that could be added is disabling error check depending on serviceUrl provided but this is not ideal since not all IBM watsonx.ai software instalations will allow user to use lightweight engine.
Haha, thanks. I enjoy speaking/writing in English so that works!
| if (!("projectId" in fields || "spaceId" in fields || "idOrName" in fields)) | ||
| throw new Error( | ||
| "No id specified! At least id of 1 type has to be specified" | ||
| ); |
| // @ts-expect-error Intentionally passing not enough parameters | ||
| const testPropsServiceUrl: WatsonxInputLLM = { | ||
| serviceUrl: process.env.WATSONX_AI_SERVICE_URL as string, | ||
| }; | ||
| expect( | ||
| () => | ||
| new WatsonxLLM({ | ||
| ...testPropsServiceUrl, | ||
| ...fakeAuthProp, | ||
| }) | ||
| ).toThrowError(); |
There was a problem hiding this comment.
Assuming it's possible to check for the lightweight engine, you could keep this logic here, but pass whatever is necessary to detect that use
FilipZmijewski
changed the title
dosubot
bot
added
lgtm
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
1 Skipped Deployment
|
FilipZmijewski
force-pushed
the
main
branch
2 times, most recently
from
61f672a to
ccbe52d
Compare
dosubot
bot
added
size:XL
|
@benjamincburns previous tests failed for some reason, I merged with main and hopefully this is fine now. |
FilipZmijewski
changed the title
|
Thank you! |
Providing small changes in order to allow user to use inference methods in both LLM and chat class