feat: chat to responses convert filds

[zijiren233]

No description provided.

[Copilot]

Pull request overview

This PR extends the Chat Completions → Responses API conversion to cover additional request fields (tools/function-call compatibility, response_format/json_schema, logprobs, parallel tool calls, and multimodal image content), and updates the stream-fake plugin to correctly preserve multiple streamed choices when converting back to a non-streaming Chat Completions response.

Changes:

• Update stream-fake streaming aggregation to track per-choice state (by choice.index) and emit multiple final choices in the non-stream response.
• Enhance OpenAI chat→responses request conversion to support legacy functions/function_call, named tool choices, response_format → text.format, logprobs include flags, and image content parts.
• Extend relay model structs to represent the newly supported request/response fields (schema text format and input_image fields, plus n, modalities, parallel_tool_calls).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
core/relay/plugin/streamfake/fake.go	Track streaming deltas per choice.index and rebuild non-stream responses with multiple choices plus prompt-level filter fields.
core/relay/plugin/streamfake/fake_test.go	Add/adjust tests for multi-choice preservation and updated choice-state handling in streamfake.
core/relay/model/response.go	Expand Responses API structs (text format schema fields; input_image fields on input content).
core/relay/model/completions.go	Add Chat Completions request fields needed for conversion (modalities, n, parallel_tool_calls).
core/relay/adaptor/openai/chat.go	Convert more Chat Completions fields into Responses API shape (legacy functions/tool_choice flattening, image parts, response_format, logprobs, parallel tool calls).
core/relay/adaptor/openai/chat_test.go	Add coverage for new conversions (named tool choice, legacy functions, images, response_format flattening, logprobs include, parallel_tool_calls, and n handling).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.