Fix delta key in last streaming response in case there is empty content

[dkalinowski]

No description provided.

[Copilot]

Pull request overview

This PR fixes OpenAI Chat Completions streaming output to always include a delta object in each streamed choices[] entry, including the final “finish_reason-only” chunk when the parser returns an empty document (e.g., when generation ends on a swallowed token). This aligns OVMS streaming output with the expected OpenAI-compatible schema and adds a regression test to prevent future regressions.

Changes:

• Ensure serializeStreamingChunk() emits "delta": {} when the parsed streaming delta document has no "delta" member.
• Add a unit test covering the empty-document final chunk case, asserting both presence of "delta":{} and correct finish_reason serialization.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/llm/apis/openai_completions.cpp	Adds fallback emission of an empty delta object when the parser provides no delta for a streaming chunk.
src/test/http_openai_handler_test.cpp	Adds a regression test asserting that the streaming chunk always contains a delta field even when the parsed delta is empty.