QuantumNous/new-api v1.0.0-rc.42 on GitHub

Important

Search Billing & Task Plugins / 搜索计费与任务插件

重做了搜索函数计费。渠道可以按上游自己的方式进行网络搜索，并按上游回报的搜索次数计费。

任务插件性能大幅提升。支持插件决定是否保留 JSON 字段顺序，声明 json-order@1 即可；其他插件行为不变。

高级自定义渠道可以把 Claude Messages 和 Gemini 转到 OpenAI Responses。

新增 xAI Grok Imagine 视频任务插件。

Search-function billing has been reworked. Channels can encode Web search for their upstream and settle from the search counts the vendor reports.

Task Plugins are substantially faster. A plugin that must keep the client’s JSON field order can declare json-order@1. Other plugins are unchanged.

Advanced Custom channels can convert Claude Messages and Gemini to OpenAI Responses. An xAI Grok Imagine video Task Plugin is added.

New Features

Reworked search-function billing, including per-Channel Web search encoding and vendor-reported search counts.
Advanced Custom channels can convert Claude Messages and Gemini generateContent to OpenAI Responses.
Moonshot Channels forward /v1/responses unchanged. Kimi currently serves kimi-k3 on that path.
Added the xAI Grok Imagine video Task Plugin: grok-imagine-video, grok-imagine-video-1.5, and grok-imagine-video-1.5-lite.
Advanced parameter overrides support Regex conditions (#7698).
Users can be filtered by User Group.
Task Plugins that declare json-order@1 keep the client’s JSON field order through to the upstream.

Task Plugin requests are substantially faster, including passthrough submissions.
Token counting is much faster, including long text without punctuation.
Gemini stream usage is no longer multiplied by the number of frames (#7687).
Responses tool-output images are sent to Gemini as images, not as text. Claude Opus 5.5 defaults to medium effort when the client sets none.
Responses custom tools such as exec and apply_patch are kept on Claude and Gemini upstreams (#7636).
Turning off the Thinking Suffix Adapter now also leaves effort suffixes such as -high unchanged.
Completed image generations still bill when the Responses request later fails or is cancelled.
Frontend build files no longer count against the page rate limit, so a cold load does not return 429 for the next session on the same IP.
gpt-6-sol and gpt-6-luna use max_completion_tokens (#7559).
System Settings are hidden from non-root admins (#7513).
Fixed a blank Channels page on OriginOS 6. Fixed rounded clipping of Group names on API Keys (#7645).

fix: isolate TLS configs across HTTP transports by @seefs001 in #7627
fix: use max_completion_tokens for gpt-6-sol and gpt-6-luna by @feitianbubu in #7559
fix(web): Add class to wrap the long strings at auto-disable status tooltip content in channel manage page by @makisekuris in #7604
fix(web): hide system settings from non-root admins by @jimmyleocn in #7513
fix(i18n): keep cached zhTW interface code stable across page loads by @yenchieh in #7139
fix(web): correct translations, chart ordering, and font loading by @seefs001 in #7628
fix(relayconvert): keep Responses custom tools on Claude and Gemini upstreams by @Calcium-Ion in #7636
fix(keys): prevent rounded clipping of group names by @jimmyleocn in #7645
fix(relaykit): Gemini stream token details, Responses tool-output media on Gemini, Opus 5.5 default effort by @Calcium-Ion in #7687
feat(relay): web search encoding per channel and vendor-reported search billing by @Calcium-Ion in #7690
feat: support regex conditions in advanced parameter overrides by @seefs001 in #7698

Full Changelog: v1.0.0-rc.41...v1.0.0-rc.42