✦VibeAPI

更新日志

版本日志信息 · 数据更新于 2026-5-16 02:21:24

如需查看全部历史版本，请访问 GitHub Releases 页面，本页面从该页面定时获取最新更新信息。

v1.0.0-rc.6

最新正式版本 · 发布于 2026-05-13 22:29:13 (中国时间)

Highlights

This release improves observability by recording upstream request IDs in logs and adds a compliance confirmation step for paid features. If your logs table is large, pre-apply the upstream_request_id column and index with online DDL before upgrading to avoid a potentially long migration.

New Features

  • Added upstream request ID tracking in request logs for easier upstream call tracing, while avoiding unintended response header overrides.
  • Added a compliance confirmation step before paid features can be used.

Bug Fixes

  • Fixed the channel combobox so focusing it no longer hides valid channel options (#4829).

Other

  • Upgrade notice: the logs table now includes a new upstream_request_id column and index; on large installations, run the appropriate online DDL below before upgrading to avoid a long or blocking automatic migration.

MySQL / InnoDB:

-- 1. Add the column using online DDL.
ALTER TABLE logs
  ADD COLUMN upstream_request_id VARCHAR(128) NOT NULL DEFAULT '',
  ALGORITHM=INPLACE, LOCK=NONE;

-- 2. Add the index using online DDL.
ALTER TABLE logs
  ADD INDEX idx_logs_upstream_request_id (upstream_request_id),
  ALGORITHM=INPLACE, LOCK=NONE;

PostgreSQL 11+:

-- 1. Add the column without rewriting the table on PG 11+.
ALTER TABLE logs
  ADD COLUMN IF NOT EXISTS upstream_request_id VARCHAR(128) NOT NULL DEFAULT '';

-- 2. Add the index concurrently to avoid blocking reads and writes.
CREATE INDEX CONCURRENTLY IF NOT EXISTS idx_logs_upstream_request_id
  ON logs (upstream_request_id);

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-rc.5...v1.0.0-rc.6

下载资源

v1.0.0-rc.5

正式版本 · 发布于 2026-05-12 16:53:57 (中国时间)

Highlights

The default dashboard has been refreshed with built-in model performance metrics and a clearer health view. This release also adds DeepChat deeplink support and fixes several default UI, wallet, playground, and ratio-display regressions.

New Features

  • Dashboard overview has been redesigned and now includes model performance metrics with a clearer performance health panel.
  • DeepChat deeplinks are now supported, making it easier to open supported chat links directly in DeepChat. (#4668)

Bug Fixes

  • Top-up gateway availability now uses the correct top-up configuration, so wallet payment options are shown accurately. (#4599)
  • Group ratio displays now honor configured group-to-group ratio settings. (#4772)
  • Default UI dropdown menu items now trigger their selection handlers reliably. (#4787)
  • The playground now handles legacy classic-format messages without failing. (#4650)
  • Default frontend pages now avoid several v1 UI regressions, including stray empty description areas and layout issues. (#4802)

Improvements

  • The default web UI and charts now follow shared theme presets for more consistent colors, controls, and corner radii.
  • Docker images now include project and third-party license notices for easier compliance review. (#4739)

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-rc.4...v1.0.0-rc.5

下载资源

v1.0.0-rc.4

正式版本 · 发布于 2026-05-06 22:25:59 (中国时间)

Highlights

OpenAI image editing is more reliable: image edit requests now preserve reference image data and related metadata, preventing edits from losing their intended context (#4646). This release also adds model performance badges and fixes several billing and dropdown UI issues.

New Features

  • Added model performance badges to model lists so users can compare available models at a glance.

Bug Fixes

  • Fixed OpenAI image edit requests so reference images and related edit metadata are preserved when forwarded (#4646).
  • Fixed billing settings forms so top-up and billing configuration inputs behave more reliably.
  • Fixed dropdown and select controls so options render and select correctly after the UI component update (#4655).

Breaking Changes

  • Integrations should now read the top-up link from the top-up information API instead of the status API.
Model availability rankingsRefreshed pricing and settings UI

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-rc.3...v1.0.0-rc.4

下载资源

v1.0.0-rc.3

正式版本 · 发布于 2026-05-06 18:43:33 (中国时间)

Highlights

v1.0.0-rc.3 adds a new model rankings experience with model availability and performance visibility, making it easier to compare reliability before choosing routes or providers. This release also refreshes key default UI areas, including pricing/settings pages and table controls.

New Features

  • Added a model rankings dashboard with real availability and performance data so users can compare model reliability at a glance (#4633).

Improvements

  • Refreshed the default web UI with Base UI components, updated theme presets, and cleaner table toolbar interactions (#4633).
  • Reorganized the system settings pricing interface to make pricing configuration easier to scan and manage.

Bug Fixes

  • Fixed channel table sorting so server-side sorting now works correctly (#4600).
  • Corrected subscription payment information display in the default UI.
Model availability rankingsRefreshed pricing and settings UI

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-rc.2...v1.0.0-rc.3

下载资源

v1.0.0-rc.2

正式版本 · 发布于 2026-04-30 20:42:48 (中国时间)

Highlights

v1.0.0-rc.2 officially opens up switching between the new and classic frontend, making it easier to test the redesigned UI, compare workflows, and fall back instantly when needed. This RC also refreshes key console pages and fixes several web issues affecting email binding, tiered pricing, API key grouping, and custom Vertex base URLs.

New Features

  • Added a built-in switch between the new and classic frontend, so teams can adopt the redesigned UI gradually without losing access to the previous workflow.

Improvements

  • Refreshed the web console with redesigned pricing pages, more consistent dashboard and log controls, better table and analytics filtering, smoother navigation, and improved mobile responsiveness.

Bug Fixes

  • Fixed email binding in the redesigned frontend so the action completes correctly (#4551).
  • Fixed tiered pricing number input behavior so editing values is more reliable (#4536).
  • Fixed tiered pricing display issues that could cause values and labels to appear incorrectly (#4530).
  • Fixed API key grouping in the default web theme so keys appear under the expected groups (#4548).
  • Fixed custom Vertex base URLs so requests route correctly when a gateway prefix is configured (#4200).
Refreshed pricing pageDashboard and analytics controls
Refreshed pricing pageDashboard and analytics controls

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-rc.1...v1.0.0-rc.2

下载资源

v1.0.0-rc.1

正式版本 · 发布于 2026-04-29 20:15:01 (中国时间)

v1.0.0-rc.1 is the first release candidate for the upcoming 1.0 release, and its main focus is the brand-new v1 frontend rebuilt from the ground up for a cleaner, faster, and more capable admin experience.

This RC refreshes the interface across navigation, configuration, analytics, and logs, making day-to-day management much easier as the project moves toward the final v1 launch.

New Features

  • Introduced the rebuilt v1 frontend with modern navigation and layouts, a visual channel editor for parameter overrides, and easier access to system settings and theme switching (#4265).

Improvements

  • Upgraded dashboards and charts with clearer model analytics, better data sorting, and more informative tooltips for easier usage analysis.
  • Improved usage logs with log type indicators, usernames and avatars, and cleaner tables that make activity easier to scan.
  • Polished shared layouts, tables, and form styling across the default frontend for a cleaner and more consistent experience.

Bug Fixes

  • Fixed themes loaded from the database so the selected frontend theme now applies correctly at runtime (#4518).
  • Fixed forwarded requests so empty reasoning fields are preserved instead of being silently dropped (#4520).
  • Fixed required-field markers so forms more consistently show which inputs are mandatory (#4533).

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v1.0.0-alpha.1...v1.0.0-rc.1

下载资源

v1.0.0-alpha.1 - License boundary release

预发布版本 · 发布于 2026-04-28 14:26:06 (中国时间)

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.13.2...v1.0.0-alpha.1

下载资源

v0.13.2

正式版本 · 发布于 2026-04-27 22:11:42 (中国时间)

New Features

  • Model sync now detects upstream models more accurately, refreshes pricing automatically from provider pricing endpoints, shows removed upstream models in the fetch-models dialog, and keeps pricing available during sync delays (#4452).
  • User records now include account creation and last login timestamps for easier auditing and administration (#4450).
  • Added configurable model matching for Ali native Messages APIs, making it easier to route requests to compatible upstream models (#4468).

Bug Fixes

  • Tool calls now accept raw JSON argument payloads, improving compatibility with providers that return tool arguments as JSON objects (#4469).
  • Configuration updates now handle JSON object fields more reliably, preventing some settings from being saved incorrectly.

Improvements

  • Retry messages now clearly explain when a channel was skipped because affinity is disabled (#4453).

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.13.1-patch.1...v0.13.2

下载资源

v0.13.1-patch.1

正式版本 · 发布于 2026-04-25 13:27:23 (中国时间)

Highlights

This patch adds support for using len in tier conditions and the LLM prompt helper, making it easier to build rules based on prompt length.

New Features

  • Added a len variable to tier conditions and the LLM prompt helper so you can define rules based on prompt length.

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.13.1...v0.13.1-patch.1

下载资源

v0.13.1

正式版本 · 发布于 2026-04-24 22:20:58 (中国时间)

New Features

  • Added support for handling DeepSeek V4 reasoning suffixes (#4428)

Bug Fixes

  • Fixed an issue where tiered billing models were excluded from the model listing (#4431)

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.13.0...v0.13.1

下载资源

v0.13.0

正式版本 · 发布于 2026-04-24 14:09:21 (中国时间)

Highlights

This release introduces powerful support for expr expression-based tiered billing, allowing administrators to configure complex, dynamic pricing rules. A simple use case for this feature is a pricing expression like tokens <= 128000 ? 0.002 : 0.004, where the system will automatically apply a lower rate for usage under 128K tokens and a higher rate for anything above that threshold.

New Features

  • Introduces support for evaluating expr expressions in the billing system, enabling dynamic tiered pricing with configurable presets (#4409).
  • Adds a UI for tool pricing settings and improves tool call quota calculations.
  • Displays the user ID in admin top-up bills for easier tracking (#4349).
  • Adds a 'last used time' column to the tokens table (#4372).
  • Relaxes token key column length requirements to support the migration of longer legacy token keys (#4401).

Bug Fixes

  • Ensures the pricing calculation logic only applies the N ratio to image models (#4412).
  • Preserves text tool surcharges properly during tiered billing settlement (#4162).
  • Corrects the completion ratio configuration for the gpt-5.5 model (#4414).
  • Fixes the model pricing UI to use the correct display type (#4426).
  • Resolves runtime crashes and edge cases within the tiered pricing editor and settlement logic.
  • Updates channel testing to properly support and evaluate tiered billing models.

Improvements

  • Adds preset pricing tiers for Doubao Seed 1.8 to enhance discount calculations.
  • Replaces Card components with divs to improve overall layout consistency.
  • Updates internal billing calculations to use quota per unit for better accuracy.

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.15...v0.13.0

下载资源

v0.12.15

正式版本 · 发布于 2026-04-22 23:04:03 (中国时间)

New Features

  • Added Waffo Pay support for account top-ups and payment workflows (#4089).

Bug Fixes

  • Fixed a recharge card issue in the top-up flow that could break parts of the card interface (#4388).

Improvements

  • Added support for Gemini's IncludeServerSideToolInvocations setting in ToolConfig, improving compatibility with newer Gemini tool-calling behavior (#4311).
  • Improved Codex channel reliability and usability by refreshing credentials for auto-disabled channels, using streaming for automatic tests, and refining the usage modal (#4324).

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.14...v0.12.15

下载资源

v0.12.14

正式版本 · 发布于 2026-04-18 00:41:21 (中国时间)

Improvements

  • Improved user-visible management and top-up logs with a cleaner display, preserved row expansion, and clearer handling of legacy records.

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.13...v0.12.14

下载资源

v0.12.13

正式版本 · 发布于 2026-04-18 00:04:31 (中国时间)

This release improves admin workflows for top-ups and makes account status changes apply more consistently.

Bug Fixes

  • Disabling a user now takes effect immediately instead of being delayed by stale cached account data.

Improvements

  • Top-up logs now include additional admin-only audit details to make review and troubleshooting easier.

Breaking Changes

  • Top-up searches are now limited to a maximum 30-day date range to keep queries responsive and reliable.

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.12...v0.12.13

下载资源

v0.12.12

正式版本 · 发布于 2026-04-17 14:05:08 (中国时间)

This release adds support for Claude Opus 4.7, expanding the set of available models in the API.

New Features

  • Added support for the Claude Opus 4.7 model, making it available through the API alongside existing model options (#4293).

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.11...v0.12.12

下载资源

v0.12.11

正式版本 · 发布于 2026-04-16 20:13:49 (中国时间)

New Features

  • Added payment method tracking and validation for top-ups.

Bug Fixes

  • Improved handling of empty string content during OpenAI to Claude message conversion.

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.10...v0.12.11

下载资源

v0.12.10

正式版本 · 发布于 2026-04-15 21:02:01 (中国时间)

New Features

  • Added passthrough support for Claude cache_control and speed options, giving you more control over request behavior when using Claude models (#4247).

Bug Fixes

  • Fixed OpenAI Responses API compatibility issues when instructions contain structured content instead of plain text (#4260).

Improvements

  • Improved Stripe payment processing to better handle asynchronous webhook events, making delayed payment confirmations more reliable.
  • Quota adjustment logs now record the acting admin username, making administrative audit trails easier to follow (#4216).

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.9...v0.12.10

下载资源

v0.12.9

正式版本 · 发布于 2026-04-13 15:28:47 (中国时间)

New Features

  • Subscription cards now show the next quota reset time, making it easier to see when usage limits will refresh (#4181).

Bug Fixes

  • Claude requests are now more reliable by avoiding incompatible Top P settings when needed.
  • Azure channels now correctly support the /v1/responses/compact route (#4149).
  • Editing values in GroupTable no longer forces the cursor to jump to the end on each keystroke (#4208).

Improvements

  • Dashboard charts have been refined to present rankings and chart sizing more clearly for easier monitoring.

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.8...v0.12.9

下载资源

v0.12.8

正式版本 · 发布于 2026-04-12 23:58:29 (中国时间)

Bug Fixes

  • Correctly report the stream status in error logs instead of defaulting to false (#4195).

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.7...v0.12.8

下载资源

v0.12.7

正式版本 · 发布于 2026-04-12 17:47:25 (中国时间)

Improvements

  • Quota management now uses an amount-first input flow and applies changes atomically, making balance adjustments simpler and more reliable.
  • Model pricing errors now use cleaner, role-aware messages so it is easier to understand and resolve failed pricing actions.

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.6...v0.12.7

下载资源

v0.12.6

正式版本 · 发布于 2026-04-09 14:44:42 (中国时间)

New Features

  • Fill in custom fields for vllm-omini (#4154)

Bug Fixes

  • Accept string usage values in Alibaba task polling (#4155)
  • Use correct endpoint for coding plan image generation in Zhipu 4V (#4146)
  • Prefer explicit pricing for compact models (#4156)
  • Fix document rendering (#4153)

Improvements

  • Refine PR template and add PR submission checks (#4076)

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.5...v0.12.6

下载资源

nightly-20260409-4d2993e

正式版本 · 发布于 2026-04-09 17:14:49 (中国时间)

Highlights / 重点更新

This release introduces comprehensive support for tiered billing (阶梯计费). It allows for flexible pricing expressions, detailed tool call quota calculations, and channel test support for tiered models. 本次发布重点增加了对阶梯计费的全面支持，支持灵活的计费表达式解析、自定义工具调用额度计算，并完善了UI配置项和渠道测试支持。

Docker image: calciumion/new-api:nightly Docker 镜像: calciumion/new-api:nightly

New Features

  • Added comprehensive support for tiered billing, including flexible expression evaluation, tool pricing settings UI, and channel test support (#4145).
  • Added custom fields support for vllm-omini models (#4154).

Bug Fixes

  • Fixed an issue in Alibaba Cloud task polling to correctly accept string usage values (#4155).
  • Fixed the endpoint for Zhipu 4v coding plan image generation (#4146).
  • Fixed pricing calculation to prefer explicit pricing for compact models (#4156).

Improvements

  • Improved UI layout consistency by replacing card components with divs.
  • Added nightly branch trigger to Docker image CI workflows.
  • Refined PR templates and added PR submission checks (#4076).

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.5...nightly-20260409

下载资源

v0.12.5

正式版本 · 发布于 2026-04-08 17:13:32 (中国时间)

New Features

  • Add support for Minimax image generation relay (#4103).

Improvements

  • Redesign group ratio rules with a collapsible layout for better usability.
  • Enhance dashboard chart axes and sorting logic.
  • Add a convenient copy button next to the API link in the dashboard info panel.

Bug Fixes

  • Resolve email delivery issues for Outlook and similar providers (#4112).

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.4...v0.12.5

下载资源

v0.12.4

正式版本 · 发布于 2026-04-08 15:49:44 (中国时间)

New Features

  • Added an IncludeModelName option to channel affinity rules for per-model affinity tracking (#3488)
  • Added an ErrorBoundary component to the web UI to prevent full-page crashes (#3350)
  • Added support for Duration in TaskSubmitReq (#4068)
  • Added admin user analytics to the dashboard

Bug Fixes

  • Fixed pricing filtering based on usable groups (#4123)
  • Fixed an issue where Claude streaming interruptions incorrectly overwrote usage data instead of preserving cache billing fields (#4128)
  • Fixed an issue where the "do not retry after failure" configuration was incorrectly overwritten in memory (#4142)

Improvements

  • Added support for enabling error logging via environment variables (#4131)
  • Wrapped scope tag labels with t() for better i18n support
  • Fixed chart labels on the dashboard

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.3...v0.12.4

下载资源

v0.12.3

正式版本 · 发布于 2026-04-08 01:04:04 (中国时间)

This is a small polish release focused on making configuration easier to navigate by simplifying the settings layout and consolidating model pricing controls.

Improvements

  • Refined the settings interface with a unified model pricing view and a cleaner tab structure to make configuration easier to navigate.

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.2...v0.12.3

下载资源

v0.12.2

正式版本 · 发布于 2026-04-06 18:10:52 (中国时间)

New Features

  • Add support for PDF conversion between OpenAI and Claude APIs.
  • Support differential billing for Seedance 2.0 based on video vs. no-video input.
  • Enhance max_tokens handling and input sanitization in the playground (#4106).
  • Enhance stream status display with error tooltips in usage logs.

Bug Fixes

  • Fix usage calculation issues when converting between Claude and OpenAI APIs by emitting message_delta for the final stream chunk (#4090).
  • Fix Gemini stream detection by checking the :streamGenerateContent URL path (#4087).
  • Fix Claude request requirements by properly setting TopP to nil.

Improvements

  • Unify file source creation and enhance caching mechanisms.

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.1...v0.12.2

下载资源

v0.12.1

正式版本 · 发布于 2026-04-02 22:01:41 (中国时间)

This release adds Seedance 2.0 video support and broadens media compatibility across providers. It also makes failures easier to understand with clearer Seedance and performance-related error messages.

New Features

  • Added Seedance 2.0 video generation API support with duration control in seconds, clearer failure reporting, and better prompt handling for single-text requests (#4042).
  • Added HEIC/HEIF image support, including Gemini channel compatibility and more reliable handling of HEIF images (#4049).
  • Added basic inline file support for Claude relay so compatible requests can include media directly (#3505).

Improvements

  • Improved performance-related error messages to make request failures easier to understand.

What's Changed

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.0...v0.12.1

下载资源

v0.12.0

正式版本 · 发布于 2026-04-02 00:22:05 (中国时间)

v0.12.0 adds Wan 2.7 image generation support and streamlines channel setup in the admin UI. It also improves billing accuracy and fixes several model and dashboard issues.

New Features

  • Added support for Wan 2.7 image generation, including control over how many images are generated per request (#3526).

Bug Fixes

  • Fixed image-generation billing and usage statistics so request counts and costs are reported more accurately without double-counting (#3512).
  • Fixed a UI issue that caused the consumption distribution chart to make the page scrollbar flicker on hover (#3474).
  • Fixed xAI Grok-3-mini compatibility when no completion token limit is set, reducing failed requests.

Improvements

  • Made channel setup faster and smoother by letting you create a channel from a copied token, remembering advanced settings in the edit dialog, and improving clipboard handling.

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.11.9...v0.12.0

下载资源

v0.12.0-alpha.2

正式版本 · 发布于 2026-03-31 21:43:12 (中国时间)

Release Highlights

This release includes minor improvements to the EditChannelModal component to enhance clipboard handling reliability.

Bug Fixes

  • EditChannelModal: Enhance clipboard handling with error checks (670abee)

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.12.0-alpha.1...v0.12.0-alpha.2

下载资源

v0.12.0-alpha.1

正式版本 · 发布于 2026-03-31 19:40:48 (中国时间)

Highlights

This release introduces a new clipboard magic string feature for quick channel creation and includes important security and UI improvements.

New Features

  • Add clipboard magic string for quick channel creation from token copy

Bug Fixes

  • 修复消耗分布图表悬浮时滚动条闪烁 (Fix dashboard scrollbar flickering) (#3474)
  • Fix xAI MaxTokens mapping for grok-3-mini model

Improvements

  • Security improvements
  • Enhance footer layout and styling
  • Update Traditional Chinese README with missing content and partner logo (#3462)
  • Refactor account binding endpoints to use POST and normalize reset responses

What's Changed

New Contributors

Full Changelog: https://github.com/QuantumNous/new-api/compare/v0.11.9...v0.12.0-alpha.1

下载资源

目录

v1.0.0-rc.6HighlightsNew FeaturesBug FixesOtherWhat's Changedv1.0.0-rc.5HighlightsNew FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv1.0.0-rc.4HighlightsNew FeaturesBug FixesBreaking ChangesGalleryWhat's Changedv1.0.0-rc.3HighlightsNew FeaturesImprovementsBug FixesGalleryWhat's Changedv1.0.0-rc.2HighlightsNew FeaturesImprovementsBug FixesGalleryWhat's ChangedNew Contributorsv1.0.0-rc.1New FeaturesImprovementsBug FixesWhat's ChangedNew Contributorsv1.0.0-alpha.1 - License boundary releasev0.13.2New FeaturesBug FixesImprovementsWhat's Changedv0.13.1-patch.1HighlightsNew Featuresv0.13.1New FeaturesBug FixesWhat's Changedv0.13.0HighlightsNew FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.15New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.14Improvementsv0.12.13Bug FixesImprovementsBreaking Changesv0.12.12New FeaturesWhat's Changedv0.12.11New FeaturesBug Fixesv0.12.10New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.9New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.8Bug FixesWhat's Changedv0.12.7Improvementsv0.12.6New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsnightly-20260409-4d2993eHighlights / 重点更新New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.5New FeaturesImprovementsBug FixesWhat's ChangedNew Contributorsv0.12.4New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.3ImprovementsWhat's Changedv0.12.2New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.1New FeaturesImprovementsWhat's Changedv0.12.0New FeaturesBug FixesImprovementsWhat's ChangedNew Contributorsv0.12.0-alpha.2Release HighlightsBug Fixesv0.12.0-alpha.1HighlightsNew FeaturesBug FixesImprovementsWhat's ChangedNew Contributors