feat: respnoses api error cache handler

[zijiren233]

No description provided.

[Copilot]

Pull request overview

This PR improves the relay’s handling of OpenAI “Responses API” (and related adapters) by (1) rewriting outbound model fields back to the client-requested/origin model, and (2) strengthening streaming error handling—especially around lifecycle events and late-arriving error events.

Changes:

• Generalize stream scanner sizing to consider multiple model names (origin + actual), ensuring image-model streams get the larger buffer even when models are mapped.
• Add keepalive stream event support and parse/propagate error objects on Responses API SSE events; implement buffering logic to avoid emitting lifecycle events too early when an error is imminent.
• Rewrite top-level and nested model fields in Responses + Video outputs to return origin model names consistently; update adapter conversions and tests accordingly.

Reviewed changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
core/relay/utils/utils.go	NewStreamScanner now accepts multiple model names; uses image buffer if any matches image models.
core/relay/utils/utils_test.go	Adds coverage to ensure mapped image models get the larger scanner buffer.
core/relay/utils/reasoning.go	Refactors FirstMatchingModelName to accept variadic model names and updates internal callers.
core/relay/model/response.go	Adds keepalive event constant and error field to stream event model.
core/relay/adaptor/vertexai/adaptor.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/qianfan/adaptor.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/openai/video.go	Rewrites top-level model in video JSON responses back to origin model.
core/relay/adaptor/openai/response.go	Adds Responses API model rewriting + new streaming scan/buffer/error handling logic.
core/relay/adaptor/openai/response_test.go	Adds/updates tests for stream buffering/error behavior and model rewriting.
core/relay/adaptor/openai/reasoning.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/openai/image.go	Ensures scanner sizing considers origin + actual model names.
core/relay/adaptor/openai/gemini.go	Uses origin model name in Gemini responses; strengthens stream error handling.
core/relay/adaptor/openai/gemini_to_responses_test.go	Updates expected Gemini model version to origin model.
core/relay/adaptor/openai/gemini_test.go	Adds coverage for error handling before downstream writes.
core/relay/adaptor/openai/claude.go	Uses origin model name in Claude responses; strengthens stream error handling.
core/relay/adaptor/openai/claude_test.go	Updates expected Claude model and adds stream error test.
core/relay/adaptor/openai/chat.go	Introduces responseModelName helper; improves stream error handling and output filtering.
core/relay/adaptor/openai/chat_test.go	Updates/extends tests for stream done/error behavior and origin model propagation.
core/relay/adaptor/gemini/openai.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/gemini/adaptor.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/doubao/main.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/doubao/chat.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/baidu/adaptor.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/anthropic/utils.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/anthropic/openai.go	Updates FirstMatchingModelName call signature usage.
core/relay/adaptor/ali/chat.go	Updates FirstMatchingModelName call signature usage.

---

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

[Copilot]

Pull request overview

Copilot reviewed 26 out of 26 changed files in this pull request and generated no new comments.