feat: support Responses to Chat#5787
Merged
Merged
Conversation
Add a shared Responses-to-Chat stream state machine and use it from the OpenAI relay path. Preserve assistant text alongside tool calls, bind tool argument deltas by output_index, map incomplete finish reasons, support reasoning/custom tool events, and buffer upstream SSE for non-stream Chat clients. Add deterministic service tests and relay SSE tests for the conversion path. Related to #5745.
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (8)
🚧 Files skipped from review as they are similar to previous changes (6)
WalkthroughAdds a new ChangesResponses ↔ Chat Completions Bidirectional Conversion
Sequence Diagram(s)sequenceDiagram
participant Client
participant AdvancedCustomAdaptor
participant relayconvert
participant UpstreamChatAPI
Client->>AdvancedCustomAdaptor: POST /v1/responses (OpenAIResponsesRequest)
AdvancedCustomAdaptor->>relayconvert: ResponsesRequestToChatCompletionsRequest
relayconvert-->>AdvancedCustomAdaptor: GeneralOpenAIRequest
AdvancedCustomAdaptor->>UpstreamChatAPI: POST /v1/chat/completions
UpstreamChatAPI-->>AdvancedCustomAdaptor: ChatCompletions response/stream
AdvancedCustomAdaptor->>relayconvert: ChatCompletionsStreamChunkToResponsesEvents / ChatCompletionsResponseToResponsesResponse
relayconvert-->>AdvancedCustomAdaptor: ResponsesStreamEvents / OpenAIResponsesResponse
AdvancedCustomAdaptor-->>Client: Responses SSE stream or JSON response
Estimated code review effort🎯 5 (Critical) | ⏱️ ~120 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Calcium-Ion
changed the title
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 5
🧹 Nitpick comments (1)
relay/channel/gemini/relay_responses_test.go (1)
52-65: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winUse
assertfor the value checks in these tests.After the fatal setup passes, the remaining equality/substring checks should use
assertto match the repo’s Go test convention. As per coding guidelines, “New or substantially rewritten Go backend tests MUST usegithub.com/stretchr/testify/requirefor setup and fatal assertions, andgithub.com/stretchr/testify/assertfor non-fatal value checks.”Also applies to: 130-150
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@relay/channel/gemini/relay_responses_test.go` around lines 52 - 65, The test in relay_responses_test.go is using require for non-fatal value and substring checks after setup has already passed; switch those assertions in the affected test blocks to assert to match the Go test convention. Keep require only for setup/fatal checks like nil validation, and update the remaining equality/contains/not-contains checks in the relay response tests (including the later similar block) to use assert, referencing the same test functions so the intent stays consistent.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@relay/channel/gemini/relay_responses.go`:
- Around line 21-27: In GeminiResponsesHandler, the upstream response body is
only closed after io.ReadAll succeeds, so a read error can leak the connection.
Move the body cleanup to run before the read, ideally with a deferred call right
after entering GeminiResponsesHandler, and keep the rest of the response
parsing/logging logic unchanged.
In `@relay/chat_completions_via_responses.go`:
- Around line 148-150: The SSE detection in the
`chat_completions_via_responses.go` flow is doing a case-sensitive
`strings.HasPrefix` check on the upstream Content-Type, so valid
`text/event-stream` responses with different casing can be missed. Update the
logic around `info.IsStream` to normalize or parse the
`httpResp.Header.Get("Content-Type")` value before checking for SSE, using the
existing stream detection path in `chat_completions_via_responses.go` so
`upstreamStream` is set correctly regardless of media-type casing.
In `@service/convert.go`:
- Around line 631-636: The fallback in `convert.go` for `claudeContent.Input`
currently lets empty, invalid, or null `toolUse.Function.Arguments` become a
string, which breaks the expected object shape. Update the `tool_use` conversion
logic around `common.Unmarshal` so `claudeContent.Input` always remains an
object/map on fallback, and treat non-parseable or null arguments as an empty
object rather than copying the raw string.
In `@service/relayconvert/responses_request_to_chat.go`:
- Around line 169-172: The conversion logic in
responsesInputTypeFunctionCallOutput and the function-tool handling path should
fail fast when required identifiers are missing instead of emitting invalid Chat
Completions payloads. Add validation before appending dto.Message so
function_call_output without a non-empty call_id and type:"function" tools
without a non-empty name return a converter error immediately. Keep the check
close to the existing responseToolOutputToChatContent and tool-mapping logic so
malformed references are rejected before building the chat payload.
In `@service/relayconvert/responses_to_chat.go`:
- Around line 784-794: The buffered argument accumulator in ProcessEvent is
duplicating tool arguments because
responsesEventFunctionArgsDelta/responsesEventCustomToolInputDelta appends the
same delta to both pendingByOutputIndex and pendingByItemID when both
identifiers are present. Update the ProcessEvent path in responses_to_chat.go to
mirror toolArgumentsDelta’s behavior by preferring OutputIndex and only falling
back to ItemID when OutputIndex is absent, so ensureTool does not drain the same
delta twice into tool.Arguments.
---
Nitpick comments:
In `@relay/channel/gemini/relay_responses_test.go`:
- Around line 52-65: The test in relay_responses_test.go is using require for
non-fatal value and substring checks after setup has already passed; switch
those assertions in the affected test blocks to assert to match the Go test
convention. Keep require only for setup/fatal checks like nil validation, and
update the remaining equality/contains/not-contains checks in the relay response
tests (including the later similar block) to use assert, referencing the same
test functions so the intent stays consistent.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: d640591e-3bc7-4b1a-b537-d417ef4e723e
📒 Files selected for processing (35)
.gitignoredto/channel_settings.godto/channel_settings_test.godto/openai_response.gorelay/channel/advancedcustom/adaptor.gorelay/channel/advancedcustom/adaptor_test.gorelay/channel/gemini/adaptor.gorelay/channel/gemini/adaptor_responses.gorelay/channel/gemini/adaptor_responses_test.gorelay/channel/gemini/relay_responses.gorelay/channel/gemini/relay_responses_test.gorelay/channel/openai/chat_via_responses.gorelay/channel/openai/chat_via_responses_test.gorelay/channel/openai/responses_via_chat.gorelay/chat_completions_via_responses.goservice/convert.goservice/openai_chat_responses_compat.goservice/openai_chat_responses_mode.goservice/openaicompat/responses_to_chat.goservice/relayconvert/chat_responses_compat_test.goservice/relayconvert/chat_to_responses.goservice/relayconvert/chat_to_responses_response.goservice/relayconvert/policy.goservice/relayconvert/regex.goservice/relayconvert/responses_request_to_chat.goservice/relayconvert/responses_request_to_chat_test.goservice/relayconvert/responses_to_chat.goweb/default/src/features/channels/lib/advanced-custom.tsweb/default/src/features/channels/types.tsweb/default/src/i18n/locales/en.jsonweb/default/src/i18n/locales/fr.jsonweb/default/src/i18n/locales/ja.jsonweb/default/src/i18n/locales/ru.jsonweb/default/src/i18n/locales/vi.jsonweb/default/src/i18n/locales/zh.json
💤 Files with no reviewable changes (1)
- service/openaicompat/responses_to_chat.go
Comment on lines
+169
to
+172
| case responsesInputTypeFunctionCallOutput: | ||
| callID := strings.TrimSpace(common.Interface2String(item["call_id"])) | ||
| content := responseToolOutputToChatContent(item["output"]) | ||
| return append(messages, dto.Message{Role: "tool", ToolCallId: callID, Content: content}), nil |
Contributor
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
Reject malformed tool references before building the chat payload.
function_call_output without call_id and type:"function" tools without name currently get converted into Chat Completions shapes the upstream API cannot validate. Failing fast here will return a precise converter error instead of leaking a later 400 from the chat endpoint.
Suggested fix
case responsesInputTypeFunctionCallOutput:
callID := strings.TrimSpace(common.Interface2String(item["call_id"]))
+ if callID == "" {
+ return nil, errors.New("function_call_output item is missing call_id")
+ }
content := responseToolOutputToChatContent(item["output"])
return append(messages, dto.Message{Role: "tool", ToolCallId: callID, Content: content}), nil
}
@@
for _, tool := range tools {
toolType := strings.TrimSpace(common.Interface2String(tool["type"]))
if toolType == "function" {
+ name := strings.TrimSpace(common.Interface2String(tool["name"]))
+ if name == "" {
+ return nil, errors.New("function tool is missing name")
+ }
out = append(out, dto.ToolCallRequest{
Type: "function",
Function: dto.FunctionRequest{
- Name: strings.TrimSpace(common.Interface2String(tool["name"])),
+ Name: name,
Description: common.Interface2String(tool["description"]),
Parameters: tool["parameters"],
},
})Also applies to: 321-330
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@service/relayconvert/responses_request_to_chat.go` around lines 169 - 172,
The conversion logic in responsesInputTypeFunctionCallOutput and the
function-tool handling path should fail fast when required identifiers are
missing instead of emitting invalid Chat Completions payloads. Add validation
before appending dto.Message so function_call_output without a non-empty call_id
and type:"function" tools without a non-empty name return a converter error
immediately. Keep the check close to the existing
responseToolOutputToChatContent and tool-mapping logic so malformed references
are rejected before building the chat payload.
8020a18 to
b59eb9e
Compare
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@relay/chat_completions_via_responses.go`:
- Around line 183-184: The content-type check in
isResponsesEventStreamContentType should not rely on substring matching, since
it can misclassify non-SSE responses that merely contain “text/event-stream” in
a parameter. Update the helper to parse the header with the MIME media type
logic and compare only the actual media type token case-insensitively, keeping
the existing call sites in chat_completions_via_responses.go unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: fb5a6772-fc5b-4d41-9413-abd27bf1f70c
📒 Files selected for processing (8)
relay/channel/gemini/relay_responses.gorelay/channel/gemini/relay_responses_test.gorelay/chat_completions_via_responses.gorelay/chat_completions_via_responses_test.goservice/convert.goservice/convert_test.goservice/relayconvert/chat_responses_compat_test.goservice/relayconvert/responses_to_chat.go
🚧 Files skipped from review as they are similar to previous changes (4)
- relay/channel/gemini/relay_responses.go
- relay/channel/gemini/relay_responses_test.go
- service/relayconvert/responses_to_chat.go
- service/relayconvert/chat_responses_compat_test.go
Comment on lines
+183
to
+184
| func isResponsesEventStreamContentType(contentType string) bool { | ||
| return strings.Contains(strings.ToLower(contentType), "text/event-stream") |
Contributor
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Parse the media type instead of substring-matching it.
strings.Contains(strings.ToLower(contentType), "text/event-stream") will also return true for headers like application/json; note=text/event-stream, so a non-stream response can be routed through the SSE path. Compare the actual media type token case-insensitively (for example via mime.ParseMediaType) rather than searching the whole header string.
Suggested fix
+import "mime"
+
func isResponsesEventStreamContentType(contentType string) bool {
- return strings.Contains(strings.ToLower(contentType), "text/event-stream")
+ mediaType, _, err := mime.ParseMediaType(contentType)
+ return err == nil && strings.EqualFold(mediaType, "text/event-stream")
}📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| func isResponsesEventStreamContentType(contentType string) bool { | |
| return strings.Contains(strings.ToLower(contentType), "text/event-stream") | |
| import "mime" | |
| func isResponsesEventStreamContentType(contentType string) bool { | |
| mediaType, _, err := mime.ParseMediaType(contentType) | |
| return err == nil && strings.EqualFold(mediaType, "text/event-stream") | |
| } |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@relay/chat_completions_via_responses.go` around lines 183 - 184, The
content-type check in isResponsesEventStreamContentType should not rely on
substring matching, since it can misclassify non-SSE responses that merely
contain “text/event-stream” in a parameter. Update the helper to parse the
header with the MIME media type logic and compare only the actual media type
token case-insensitively, keeping the existing call sites in
chat_completions_via_responses.go unchanged.
jahnli
added a commit
to jahnli/new-api
that referenced
this pull request
Jun 28, 2026
…finement Upstream commits merged: - feat: support Responses to Chat (QuantumNous#5787) - fix: update section titles and improve layout in channel components - feat: refine channel management UI Changes applied: - Gemini Responses API adaptor and relay handlers - OpenAI Responses↔Chat conversion (responses_via_chat.go) - Package rename: service/openaicompat → service/relayconvert - Channel management UI refactor (batch mode, drawer layout) - Advanced custom converter: openai_responses_to_openai_chat_completions - Brand rename (NewAPIError → AIHubError) applied to all new files Conflicts resolved (8 files): - relay/channel/gemini/adaptor.go: manual merge (keep AIHubError + add Responses mode) - channel-mutate-drawer.tsx: adopt upstream + reapply brand rename - 6 i18n locale files: both sides' new keys retained Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Important
📝 变更描述 / Description
支持 Responses to Chat 兼容转换：新增 Responses request 转 Chat Completions request 的回退能力，并补齐 Chat Completions 响应回转为 Responses 的非流式与流式输出。Advanced Custom 新增 OpenAI Responses to OpenAI Chat 转换选项；Gemini 渠道在客户端请求 /v1/responses 时会将 Gemini 响应转换回 OpenAI Responses 格式，避免上游已有内容但 Responses 客户端无法渲染。
同时修复 Responses 流式工具参数归位中 output_index 与 item_id 双 pending 导致参数重复的问题，保持 output_index 作为主绑定键，item_id 作为兼容补充。
🚀 变更类型 / Type of change
🔗 关联任务 / Related Issue
✅ 提交前检查项 / Checklist
Bug fix，我已提交或关联对应 Issue，且不会将设计取舍、预期不一致或理解偏差直接归类为 bug。📸 运行证明 / Proof of Work
已通过以下验证：
Summary by CodeRabbit
New Features
Bug Fixes