CherryStudio V2访问grok-4.20-multi-agent-0309模型报错，有佬知道如何解决吗？

用的是中转站grok模型，之前CherryStudio V1版本；一直是好的，今天升级V2版本后，后端会返回失败，

错误信息：

错误名称: AI_APICallError

错误信息: Client-side tools for multi-agent models require beta access

堆栈信息: AI_APICallError: Client-side tools for multi-agent models require beta access

    at D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:54951:12

    at process.processTicksAndRejections (node:internal/process/task_queues:104:5)

    at async postToApi$2 (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:54880:25)

    at async OpenAICompatibleChatLanguageModel.doStream (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:119233:68)

    at async wrapStream (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:274375:21)

    at async wrapStream (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:62486:43)

    at async fn (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:64667:18)

    at async D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:59202:21

    at async _retryWithExponentialBackoff (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:59389:10)

    at async streamStep (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:64636:110)

    at async fn (D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:64953:6)

    at async D:\CherryStudio\Cherry Studio\resources\app.asar\out\main\main.js:59202:21

状态码: 400

请求路径: https://xxx/v1/chat/completions

请求体: {"model":"grok-4.20-multi-agent-0309","messages":[{"role":"system","content":""},{"role":"user","content":"你是什么模型？"}],"tools":[{"type":"function","function":{"name":"fs_read","description":"Read a text file by absolute path.\n\nPrimary use: retrieving the full content behind a <persisted-output> marker — call with the path shown after \"Full output saved to:\". Markers from earlier turns work too (persisted outputs live as long as their message does). Only paths from this conversation's markers are readable; reads elsewhere return access-denied.\n\nPagination is line-based: pass `offset` (1-indexed line) + `limit` for large files; results include `totalLines`. Lines are returned in full (never truncated mid-line); the per-call output is bounded as a whole, and oversized pages return an `output-too-large` error with a file-specific recommended `limit`. The one input it can't subdivide is a single physical line larger than that cap (e.g. heavily minified JSON) — line paging can't split one line and there is no byte-range read, so that case is reported as `output-too-large`; reason from the inline head/tail excerpt for such inputs.\n\nWhen reading a persisted output to summarize, analyze, or act on it, read sequential pages (advance `offset` to the returned `endLine` + 1) until you have covered 100% of the content. Before summarizing or drawing conclusions, state what fraction you actually read — and if you did not read all of it (including the single-oversized-line case), say so explicitly rather than implying full coverage.\n\nThe persistence layer applies to non-read tools only; this tool never persists its own output — narrow the read (smaller `limit`) instead.","parameters":{"type":"object","properties":{"path":{"type":"string","minLength":1,"description":"Absolute file path. Relative paths are rejected."},"offset":{"description":"1-indexed line number to start reading at. Default: 1.","type":"integer","minimum":1,"maximum":9007199254740991},"limit":{"description":"Maximum number of lines to return. Default: 2000. No schema upper bound — the per-call output cap (the configured tool-output threshold, 50000 chars by default) is the real gate; an oversized page reports the exact cap and a recommended limit.","type":"integer","minimum":1,"maximum":9007199254740991}},"required":["path"],"additionalProperties":false}}}],"tool_choice":"auto","stream":true,"stream_options":{"include_usage":true}}

响应首部: {

  "alt-svc": "h3=\":443\"; ma=86400",

  "cache-control": "no-cache",

  "cf-cache-status": "DYNAMIC",

  "cf-ray": "xxxad-HKG",

  "connection": "keep-alive",

  "content-type": "text/event-stream",

  "date": "Tue, 11 Aug 2026 09:59:24 GMT",

  "document-policy": "include-js-call-stacks-in-crash-reports",

  "nel": "{\"report_to\":\"cf-nel\",\"success_fraction\":0.0,\"max_age\":604800}",

  "report-to": "{\"group\":\"cf-nel\",\"max_age\":604800,\"endpoints\":[{\"url\":\"https://a.nel.cloudflare.com/report/v4?s=xxx\"}]}",

  "server": "cloudflare",

  "server-timing": "cfCacheStatus;desc=\"DYNAMIC\", cfEdge;dur=4,cfOrigin;dur=1738",

  "transfer-encoding": "chunked",

  "x-content-type-options": "nosniff",

  "x-frame-options": "SAMEORIGIN",

  "x-new-api-version": "v1.0.0-rc.21",

  "x-oneapi-request-id": "xxx"

}

响应内容: {"error":{"message":"Client-side tools for multi-agent models require beta access","type":"bad_response_status_code","param":"","code":"bad_response_status_code"}}

数据: {"error":{"message":"Client-side tools for multi-agent models require beta access","type":"bad_response_status_code","param":"","code":"bad_response_status_code"}}