feat(xai): 为xAI渠道添加/v1/responses支持 by funkpopo · Pull Request #2897 · QuantumNous/new-api

funkpopo · 2026-02-08T15:46:49Z

xAI官方文档提出可使用/v1/responses接口。

但本PR之前的实现仅兼容/v1/chat/completions接口。

参考 https://docs.x.ai/developers/quickstart#step-3-make-your-first-request ，实现了对于官方x.AI渠道的/v1/responses格式兼容

Summary by CodeRabbit

New Features
- Expanded Xai channel support with additional endpoint types.
- Refreshed available Grok models and added new image/video variants (e.g., grok-4-1-fast variants, grok-3-mini, grok-2-vision-1212, grok-imagine-*).
Bug Fixes / Behavior
- Improved response and streaming handling for Xai responses.
- Default upstream model is used when none is specified.
- Removed the previous "medium" model suffix option.

coderabbitai · 2026-02-08T15:47:06Z

Walkthrough

Adds xAI channel support for OpenAI "responses" relay mode, updates xAI model list, removes "-medium" reasoning-suffix handling, and replaces direct json.Unmarshal with a common JSON helper in streaming text handling.

Changes

Cohort / File(s)	Summary
Endpoint Type Configuration `common/endpoint_type.go`	Add `ChannelTypeXai` case to return `OpenAI` and `OpenAIResponse` endpoint types so xAI can route responses-mode traffic.
xAI Channel Adaptor `relay/channel/xai/adaptor.go`	Removed "-medium" reasoning-suffix trimming (keeps "-high"/"-low"), default `request.Model` to `info.UpstreamModelName` in `ConvertOpenAIResponsesRequest`, and route `RelayModeResponses` to `OaiResponsesHandler` or `OaiResponsesStreamHandler` in `DoResponse`.
xAI Model List `relay/channel/xai/constants.go`	Replaced legacy grok model entries with a refreshed, expanded set of grok variants (language, search, high/low, image/video models).
Stream Response Handler `relay/channel/xai/text.go`	Replaced direct `json.Unmarshal` usage with `common.UnmarshalJsonStr` for streaming JSON parsing in xAI stream handler.

Sequence Diagram(s)

sequenceDiagram
  participant Client as Client
  participant Adaptor as xAI Adaptor
  participant Handler as OAI Responses Handler
  participant Upstream as Upstream API

  Client->>Adaptor: Send /v1/responses request
  Adaptor->>Adaptor: ConvertOpenAIResponsesRequest (default model from info if empty)
  Adaptor->>Handler: Route to OaiResponsesHandler or OaiResponsesStreamHandler (based on IsStream)
  Handler->>Upstream: Forward adapted request
  Upstream-->>Handler: Streamed/Non-streamed response
  Handler-->>Adaptor: Return formatted response stream or JSON
  Adaptor-->>Client: Relay response

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

PR #1569: Adds responses-mode support and integrates ConvertOpenAIResponsesRequest with OaiResponsesHandler/OaiResponsesStreamHandler.
PR #1205: Adjusts RelayModeResponses handling and routing for channel adaptors (responses preview URL / flow changes).
PR #2889: Integrates Responses-specific request/response routing and handler selection similar to this change.

Suggested reviewers

seefs001
creamlike1024

"I'm a rabbit in the code-lush glade,
New grok models in my hopping parade.
Responses routed, old suffixes undone,
Streams parsed neat with a helper run.
I nibble bugs and celebrate the sun! 🐇✨"

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title clearly describes the main objective: adding support for the /v1/responses endpoint to the xAI channel, which aligns with the primary changes across all modified files.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

📝 Generate docstrings

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment

No actionable comments were generated in the recent review. 🎉

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

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.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@relay/channel/xai/constants.go`:
- Around line 3-18: Update the ModelList slice in constants.go to match the
current xAI API: remove the deprecated "grok-4-0709" and the unavailable
"grok-imagine-image-pro", and add the missing models "grok-4", "grok-2-1212",
"grok-vision-beta", "grok-beta", and the video model "grok-imagine-video";
ensure the list keeps language models grouped together and image/video models
grouped together and that the variable ModelList is updated accordingly so
callers of ModelList use only supported model names.

coderabbitai · 2026-02-08T15:50:55Z

 var ModelList = []string{
-	// grok-4
-	"grok-4", "grok-4-0709", "grok-4-0709-search",
-	// grok-3
-	"grok-3-beta", "grok-3-mini-beta",
-	// grok-3 mini
-	"grok-3-fast-beta", "grok-3-mini-fast-beta",
-	// extend grok-3-mini reasoning
-	"grok-3-mini-beta-high", "grok-3-mini-beta-low", "grok-3-mini-beta-medium",
-	"grok-3-mini-fast-beta-high", "grok-3-mini-fast-beta-low", "grok-3-mini-fast-beta-medium",
-	// image model
-	"grok-2-image",
-	// legacy models
-	"grok-2", "grok-2-vision",
-	"grok-beta", "grok-vision-beta",
+	// language models
+	"grok-4-1-fast-reasoning",
+	"grok-4-1-fast-non-reasoning",
+	"grok-code-fast-1",
+	"grok-4-fast-reasoning",
+	"grok-4-fast-non-reasoning",
+	"grok-4-0709",
+	"grok-3-mini",
+	"grok-3",
+	"grok-2-vision-1212",
+	// image generation models
+	"grok-imagine-image-pro",
+	"grok-imagine-image",
+	"grok-2-image-1212",
 }


⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🌐 Web query:

xAI grok API supported models 2026

💡 Result:

As of February 2026, the xAI (Grok) API supports these model IDs (as shown in xAI’s API models/pricing UI and docs):

Language (text) models

grok-4-1-fast-reasoning [1]

grok-4-1-fast-non-reasoning [1]

grok-4-fast-reasoning [1]

grok-4-fast-non-reasoning [1]

grok-code-fast-1 [1]

grok-4 [1]

grok-3 (aliases include grok-3-latest, grok-3-beta, grok-3-fast, grok-3-fast-latest, grok-3-fast-beta) [2]

grok-3-mini [1]

grok-2-vision-1212 [1]

grok-2-1212 [1]

grok-vision-beta [1]

grok-beta [1]

Image generation

grok-2-image-1212 [1]

grok-imagine-image [3]

Video generation

grok-imagine-video [4]

Source list: xAI API models/pricing page [1], xAI Grok-3 model doc (aliases) [2], xAI Image Generations guide [3], xAI Video Generations guide [4].

Remove unsupported models and update to match current xAI API.

Two models in this list are not currently supported by the xAI API and will cause request failures:

grok-4-0709 (deprecated)

grok-imagine-image-pro (not available; only grok-imagine-image is supported)

Additionally, the following currently available models are missing:

Language models: grok-4, grok-2-1212, grok-vision-beta, grok-beta

Video generation: grok-imagine-video

Update this list to reflect the current xAI API as of February 2026.

🤖 Prompt for AI Agents

In `@relay/channel/xai/constants.go` around lines 3 - 18, Update the ModelList slice in constants.go to match the current xAI API: remove the deprecated "grok-4-0709" and the unavailable "grok-imagine-image-pro", and add the missing models "grok-4", "grok-2-1212", "grok-vision-beta", "grok-beta", and the video model "grok-imagine-video"; ensure the list keeps language models grouped together and image/video models grouped together and that the variable ModelList is updated accordingly so callers of ModelList use only supported model names.

seefs001 · 2026-02-11T15:33:37Z

@funkpopo 为啥要移除思考预算后缀和 -search 的处理嘞

funkpopo · 2026-02-11T16:30:12Z

@funkpopo 为啥要移除思考预算后缀和 -search 的处理嘞

修改时主要参考的 grok 4 / grok 4.1 ，忽略了还有个使用到思考预算参数的 grok-3-mini ，很抱歉，我想我应该回退，让 grok-3-mini 支持使用 reasoning_effort: low / high

参考x.AI文档说明关于搜索需要转向参数控制 searchParameters: {mode: "on"} ，是我考虑不周。我单方面以为应该由客户端通过标准的 Tool Use 或参数来控制，但忽略了各类不支持自定义参数的客户端，为了保持向下兼容性，我会 revert 掉移除后缀处理的代码。

非常感谢您的提醒，我会立即更改，并在未来提交时更注意项目的向下兼容性。

* main: (21 commits) fix: normalize search pagination params to avoid [object Object] fix: ignore header passthrough during channel tests fix(token-search): use TrimPrefix for sk- token normalization fix: rename bulk test action to skip manually disabled channels fix: support numeric status code mapping in ResetStatusCode 优化: 任务日志查询速度并显示用户详情 (QuantumNous#2905) Merge pull request QuantumNous#2916 from worryzyy/feature/add-quota-amount-input feat: Improve backend multilingual support feat: add OpenRouter pricing support to upstream ratio sync feat: refactor request body handling to use BodyStorage for improved efficiency feat(xai): 为xAI渠道添加/v1/responses支持 (QuantumNous#2897) chore: remove deprecated Docker badge from README feat: refactor extra_body handling for improved configuration parsing Update README feat: logs cache field (QuantumNous#2920) feat(localization): added zh_TW (QuantumNous#2913) fix: update README files to improve link formatting and readability feat: add Aion UI link to README files chore(deps): bump axios from 1.12.0 to 1.13.5 in /web simplify language selector display to use text-only labels ... # Conflicts: # web/src/components/layout/headerbar/LanguageSelector.jsx

* feat(xai): 为xAI渠道添加/v1/responses支持 * Add video generation model to constants * fix: 修正先前更改中对于grok-3-mini的思考预算和"-search"设计

feat(xai): 为xAI渠道添加/v1/responses支持

edf6226

coderabbitai Bot reviewed Feb 8, 2026

View reviewed changes

Add video generation model to constants

dd91b8e

Calcium-Ion previously approved these changes Feb 8, 2026

View reviewed changes

funkpopo added 3 commits February 10, 2026 09:21

Merge branch 'QuantumNous:main' into main

384d9a1

Merge branch 'QuantumNous:main' into main

cac8ec9

Merge branch 'QuantumNous:main' into main

f3a185a

fix: 修正先前更改中对于grok-3-mini的思考预算和"-search"设计

1a9e9eb

funkpopo dismissed Calcium-Ion’s stale review via 1a9e9eb February 11, 2026 16:37

Calcium-Ion merged commit 75e533e into QuantumNous:main Feb 11, 2026
1 check passed

coderabbitai Bot mentioned this pull request Mar 21, 2026

Add WebSocket relay support for /v1/responses #3370

Closed

This was referenced Apr 22, 2026

Feat responses api #4381

Closed

feat(openai): add stream-to-nonstream chat handling #4447

Open

coderabbitai Bot mentioned this pull request Apr 30, 2026

fix: aggregate SSE upstream into chat.completion JSON for non-stream clients #4557

Open

Navigation Menu

Uh oh!

feat(xai): 为xAI渠道添加/v1/responses支持 - #2897

feat(xai): 为xAI渠道添加/v1/responses支持#2897
Calcium-Ion merged 6 commits into
QuantumNous:mainfrom
funkpopo:main

funkpopo commented Feb 8, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented Feb 8, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Feb 8, 2026

Uh oh!

seefs001 commented Feb 11, 2026

Uh oh!

funkpopo commented Feb 11, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

funkpopo commented Feb 8, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Feb 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

Possibly related PRs

Suggested reviewers

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Feb 8, 2026

Choose a reason for hiding this comment

Language (text) models

Image generation

Video generation

Uh oh!

seefs001 commented Feb 11, 2026

Uh oh!

funkpopo commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

funkpopo commented Feb 8, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Feb 8, 2026 •

edited

Loading

funkpopo commented Feb 11, 2026 •

edited

Loading