fix: enable useLegacyCompletionsEndpoint to use openai package streaming

[uinstinct]

Description

Add supports for useLegacyCompletionsEndpoint to use openai adapter streaming.

closes CON-5285

AI Code Review

• Team members only: AI review runs automatically when PR is opened or marked ready for review
• Team members can also trigger a review by commenting @continue-review

Checklist

• [] I've read the contributing guide
• [] The relevant docs, if any, have been updated or created
• [] The relevant tests, if any, have been updated or created

Screen recording or screenshot

[ When applicable, please include a short screen recording or screenshot - this makes it much easier for us as contributors to review and understand your changes. See this PR as a good example. ]

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

---

Continue Tasks: ▶️ 1 queued · ✅ 1 no changes — View all

---

Summary by cubic

Disables streaming for OpenAI o-series/GPT-5 when the legacy completions endpoint is used, aligning behavior with model constraints. Addresses CON-5285.

^{Written for commit be39317. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 4 files

[RomneyDa]

@uinstinct since stream is already in completion options, how does this impact functionality? I think we should avoid adding a duplicate configuration option.

[uinstinct]

@uinstinct since stream is already in completion options, how does this impact functionality? I think we should avoid adding a duplicate configuration option.

My bad - I thought since completionOptions.stream was removed from config.yaml reference doc, support was dropped.

Added that back to the config.yaml reference in docs

[RomneyDa]

@uinstinct could you clarify how useLegacyCompletionsEndpoint and stream are related? As far as I understand useLegacyCompletionsEndpoint is for using a different URL endpoint /completions which was used prior to chat/completions coming out. stream is the issue here (you can use chat/completions with stream = false)

[uinstinct]

openai models (which require org verifications) cannot use the responses endpoint - so when useLegacyCompletionsEndpoint is enabled, it should only use the completions endpoint.

[RomneyDa]

@uinstinct the legacy completions endpoint is the ancient /completions, users should rarely use that for newer models. The fallback from responses should be /chat/completions.

But I don't think we need to fallback to that either, I think they can use responses, they just can't stream it. Kilo code fixed this by adding a stream toggle, I don't think they routed to a different endpoint. I think the issue is that stream isn't supported properly, or that we don't have error handling to detect the verification issue and switch streaming off.||

I think a stream toggle (which only shows when error detected) is a decent solution, as long as the stream parameter actually works

[uinstinct]

@uinstinct the legacy completions endpoint is the ancient /completions, users should rarely use that for newer models. The fallback from responses should be /chat/completions.

But I don't think we need to fallback to that either, I think they can use responses, they just can't stream it. Kilo code fixed this by adding a stream toggle, I don't think they routed to a different endpoint. I think the issue is that stream isn't supported properly, or that we don't have error handling to detect the verification issue and switch streaming off.||

I think a stream toggle (which only shows when error detected) is a decent solution, as long as the stream parameter actually works

It seems we will have to use the /chat/completions endpoint because /responses does not support "reasoning" for these models. Also when using /chat/completions with reasoning enabled, the model does thinking but the thinking content is not returned in the api.

I investigated kilocode now and found that they are using /chat/completions for original OpenAI provider and /responses for OpenAI native provider. The same error is produced in kilocode's openai native provider.

Currently we support using /completions endpoint via config.yaml. Should we also add support for /chat/completions via config.yaml?