Major Updates
- Added disk request body caching to System Settings - 性能设置, which can significantly reduce memory usage for large request bodies (swapping memory to disk).
- Supports configuring request header pass-through, see configuration
{
"Authorization": "{client_header:Test_Header}"
}
- Compatible with Grok's charging for non-compliant requests, with configuration support (System Settings - Model Related Settings - Grok Settings)
- Supports codex context compression/v1/response/compact (requires adding a model, e.g., to use gpt-5.2, you need to configure gpt-5.2-openai-compact)
- Supports channel preference, prioritizing a certain channel based on a field in the request (System Settings - Model Related Settings - Channel Affinity)
- Displays detailed reasons for errors rejected due to policy violations in logs
主要更新
- 系统设置-性能设置 增加磁盘请求体缓存，可大幅减少大请求体下的内存占用（磁盘换内存）
- 支持配置请求头透传，配置参考
{
"Authorization": "{client_header:Test_Header}"
}
- 兼容Grok对不合规的请求进行收费操作，支持配置 （系统设置 - 模型相关设置 - Grok设置）
- 支持codex上下文压缩/v1/response/compact （需要添加模型，例如要使用gpt-5.2，就需要配置gpt-5.2-openai-compact）
- 支持渠道优选，根据请求某个字段优先选择某个渠道 (系统设置 - 模型相关设置 - 渠道亲和性)
- 日志详情展示由于违反政策被拒绝的错误原因
What's Changed
- feat: add regex pattern to mask API keys in sensitive information by @xiangsx in #2587
- fix(task): 修复使用 auto 分组时 Task Relay 不记录日志和不扣费的问题 by @147API in #2588
- fix: 设置默认max req body 为128MB by @xyfacai in #2590
- fix: remove Minimax from FETCHABLE channels by @RedwindA in #2619
- fix(gemini): fetch model list via native v1beta/models endpoint by @RedwindA in #2615
- feat: /v1/chat/completion -> /v1/response by @seefs001 in #2629
- fix: support snake_case fields in GeminiChatGenerationConfig by @deanxv in #2646
- feat: status code auto-disable configuration by @seefs001 in #2647
- fix: chat2response setting ui by @seefs001 in #2643
- fix: clean propertyNames for gemini function by @seefs001 in #2642
- feat: channel testing supports parameter overriding by @seefs001 in #2627
- 新增: 豆包视频1.5pro by @feitianbubu in #2632
- feat: codex channel by @seefs001 in #2652
- fix: log info input_token (chat -> claude model ) by @seefs001 in #2670
- fix: update warning threshold label from '5$' to '2$' by @feitianbubu in #2690
- fix: fix email send when special character(./+) included in email string by @Bliod-Cook in #2696
- fix: video content api Priority use url field by @feitianbubu in #2691
- fix: jimeng i2v support multi image by metadata by @feitianbubu in #2688
- Revert "fix: video content api Priority use url field" by @seefs001 in #2710
- fix: openAI function to gemini function field adjusted to whitelist mode by @seefs001 in #2667
- feat: TLS_INSECURE_SKIP_VERIFY env by @seefs001 in #2668
- fix: the login method cannot be displayed under the aff link. by @seefs001 in #2676
- fix: codex Unsupported parameter: max_output_tokens by @seefs001 in #2684
- feat: customizable automatic retry status codes by @seefs001 in #2663
- fix: calls to multiple tools in gemini all return index=0 by @seefs001 in #2701
- fix: replace Alibaba's Claude-compatible url with new url by @seefs001 in #2702
- feat: log shows request conversion by @seefs001 in #2703
- feat(qwen): support qwen image sync image model config by @xyfacai in #2717
- feat: requestId time string use UTC by @feitianbubu in #2713
- fix request pass-through aws channels can't test by @daggeryu in #2693
- feat: enhance Authorization header handling with Header Override support by @Li-Xingyu in #2738
- feat(gemini): 支持 tool_choice 参数转换，优化多个渠道错误处理 by @seefs001 in #2742
- 支持 tool_choice 参数转换，优化错误处理 by @laozuzhen in #2540
- 修复钱包页面刷新时显示错误余额数据的问题 by @wans10 in #2749
- feat: 可视化模型重定向设置中，允许用户从拉取的模型列表里直接选择模型 by @Bliod-Cook in #2610
- feat: header passthrough by @seefs001 in #2735
- fix: test using the correct path for rerank by @seefs001 in #2736
- feat: channel affinity by @seefs001 in #2669
- feat: openai response /v1/response/compact by @seefs001 in #2644
- feat: grok Usage Guidelines Violation Fee by @seefs001 in #2753
- feat(gemini): map OpenAI stop to Gemini stopSequences by @RedwindA in #2779
- fix: remove disable_parallel_tool_use if tool_choice=none by @seefs001 in #2765
- fix: /v1/responses/compact default billing by @seefs001 in #2766
- feat: disk request body cache by @Calcium-Ion in #2780
- feat: Support customizing the success and cancel url of Stripe. by @mehunk in #2745
- feat: doubao add first and last image to video by @feitianbubu in #2756
- feat: task pre consume modelPrice default use setting value by @feitianbubu in #2733
- feat: CodeViewer click link and auto wrap by @feitianbubu in #2783
- feat(gemini): support cached token billing by @RedwindA in #2798
- fix(ui): use distinct color palette for group tags by @KiGamji in #2759
- fix: channel affinity by @seefs001 in #2799
- fix: make channel Host override take effect by @lanfunoe in #2805
- fix: claude panic by @seefs001 in #2804
- feat: /v1/responses qwen3 max && perplexity by @seefs001 in #2803
- feat: auto-adapt video modal by @feitianbubu in #2764
- fix: openrouter claude cache usage by @seefs001 in #2811
- feat：default enable channel affinity by @seefs001 in #2809
- fix: vertex maas api addr by @seefs001 in #2810
New Contributors
- @xiangsx made their first contribution in #2587
- @147API made their first contribution in #2588
- @deanxv made their first contribution in #2646
- @Li-Xingyu made their first contribution in #2738
- @laozuzhen made their first contribution in #2540
- @mehunk made their first contribution in #2745
- @KiGamji made their first contribution in #2759
- @lanfunoe made their first contribution in #2805
Full Changelog: v0.10.5...v0.10.7