Clarify "deployments" in azure.md & s/azure/openai. #2537
Conversation
1) Clarified use of deployment name for `engine` in Azure. 2) As of Oct 15 stable release, it seems Continue no longer recognizes "azure" as a `provider` but only as an `apiType`. I'm sure of this for the gpt-4o chat model. I'm assuming it follows for the rest.
✅ Deploy Preview for continuedev ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
| @@ -10,13 +10,13 @@ We recommend configuring **GPT-4o** as your chat model. | |||
| ```json title="config.json" | |||
| "models": [{ | |||
| "title": "GPT-4o", | |||
| "provider": "azure", | |||
| "provider": "openai", // As of Oct 15, "azure" isn't recognized | |||
There was a problem hiding this comment.
Can you share what you mean by it not being recognized? As far as I know it is still listed in the config schema
There was a problem hiding this comment.
When we use "azure" for provider, Continue generates the wrong URL, by using model instead of engine.
When we use "openai", it gets it right.
Before this week in stable, "azure" also worked. (Before last week in pre-release.)
| "apiVersion": "<YOUR_API_VERSION>", | ||
| "apiType": "azure", | ||
| "apiKey": "<MY_API_KEY>" | ||
| "engine": "<YOUR_ENGINE>", // Your deployment, eg "gpt4o-beta" |
There was a problem hiding this comment.
We recently updated the provider so that it is able to automatically set the apiType, apiVersion, and engine, so these parameters actually aren't needed at all any more (https://github.com/continuedev/continue/blob/dev/core/llm/llms/Azure.ts#L7-L16)
Can we just remove engine, apiType, and apiVersion entirely from this example?
There was a problem hiding this comment.
If provider=azure works properly, then apiType could be eliminated, and apiVersion can use your default. Not so with engine.
As of this week, if I set provider to "azure", it generates the wrong URL: Prior to this week, it used the engine to name the deployment and get my endpoint. Now it uses model, which is incorrect.
The deployment name is part of my URL, presumably to manage multiple instances of the same model. So I think the azure provider must have a way to capture my deployment name. I'm not sure what engine was supposed to be, but it used to serve the role of putting my deployment name in the proper place in the URL.
Lacking engine (or deployment), then model is a good default, but we need a way to specify it.
Description
Clarified use of deployment name for
enginein Azure. Perhaps it's called "engine" somewhere, but our Azure OpenAI console calls it "deployment".As of Oct 15 stable release, it seems Continue no longer recognizes "azure" as a
provider-- not listed in the config model. It's still needed forapiTypeof course. I'm pretty sure of this for the gpt-4o chat model, as making that change un-broke it. I'm assuming it follows for the rest.Checklist
dev, rather thanmainI based from
mainas this is fixing a doc issue in today's stable release.Screenshots
[ For visual changes, include screenshots. ]
Testing
Ideally this should be checked by someone familiar with the recent changes to be sure my doc changes are correct instead of merely "works for me".