This is an automated email from the ASF dual-hosted git repository.
shreemaan-abhishek pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/apisix-website.git
The following commit(s) were added to refs/heads/master by this push:
new 92db16e6d66 blog: add 2026 June monthly report (#2067)
92db16e6d66 is described below
commit 92db16e6d6674f13cff5221c9bcb5aa90da4f485
Author: Yilia Lin <[email protected]>
AuthorDate: Tue Jun 30 16:21:23 2026 +0800
blog: add 2026 June monthly report (#2067)
---
blog/en/blog/2026/06/30/2026-jun-monthly-report.md | 233 +++++++++++++++++++++
blog/zh/blog/2026/06/30/2026-jun-monthly-report.md | 233 +++++++++++++++++++++
2 files changed, 466 insertions(+)
diff --git a/blog/en/blog/2026/06/30/2026-jun-monthly-report.md
b/blog/en/blog/2026/06/30/2026-jun-monthly-report.md
new file mode 100644
index 00000000000..7be04d504be
--- /dev/null
+++ b/blog/en/blog/2026/06/30/2026-jun-monthly-report.md
@@ -0,0 +1,233 @@
+---
+title: "2026 Monthly Report (June 01 - June 30)"
+keywords: ["Apache APISIX", "API Gateway", "Monthly Report", "Contributor"]
+description: Our monthly Apache APISIX community report generates insights
into the project's monthly developments. The reports provide a pathway into the
Apache APISIX community, ensuring that you stay well-informed and actively
involved.
+tags: [Community]
+image:
https://static.api7.ai/uploads/2026/06/30/fx3aed6D_2026-06-monthly-report-cover-en.webp
+---
+
+> Recently, we've introduced and updated some new features, including SAML
authentication, GraphQL caching, route-level proxy buffering, richer AI
observability, new LLM security and cache plugins, safer request-body handling,
and enhanced rate limiting backends. For more details, please read this month's
newsletter.
+
+<!--truncate-->
+
+## Introduction
+
+From its inception, the Apache APISIX project has embraced the ethos of
open-source community collaboration, propelling it into the ranks of the most
active global open-source API gateway projects. The proverbial wisdom of
'teamwork makes the dream work' rings true in our way and is made possible by
the collective effort of our community.
+
+From June 1st to June 30th, 16 contributors made 137 commits to Apache APISIX.
We sincerely appreciate your contributions to Apache APISIX.
+
+## Contributor Statistics
+
+
+
+
+
+## Feature Highlights
+
+Here are the key updates from this month, grouped by capability area.
+
+### Authentication and Identity
+
+#### 1. Add `saml-auth` Plugin
+
+PR: https://github.com/apache/apisix/pull/13346
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR adds the `saml-auth` plugin, bringing SAML 2.0 authentication to the
APISIX gateway layer. Organizations that rely on SAML identity providers can
now protect upstream services through APISIX without implementing SAML handling
in each application.
+
+#### 2. Make `client_secret` Optional for Local JWT Verification in
`openid-connect`
+
+PR: https://github.com/apache/apisix/pull/13472
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR makes `client_secret` optional in `openid-connect` modes where APISIX
only validates bearer tokens locally, such as public-key or JWKS-based JWT
verification. Users no longer need to provide dummy secrets for flows that
never call the identity provider's token or introspection endpoints.
+
+#### 3. Remove Server-Side Token Generation Endpoint from `jwe-decrypt`
+
+PR: https://github.com/apache/apisix/pull/13464
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR removes the unfinished `/apisix/plugin/jwe/encrypt` helper endpoint
from the `jwe-decrypt` plugin. APISIX now focuses on decrypting JWEs at the
gateway, while token generation remains the responsibility of the service that
owns the consumer secret.
+
+#### 4. Add `max_req_body_size` to `hmac-auth`
+
+PR: https://github.com/apache/apisix/pull/13478
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR adds a `max_req_body_size` option to `hmac-auth` for routes that
enable request body validation. Oversized bodies can now be rejected before
APISIX buffers them into memory, improving protection for deployments where the
global NGINX body size limit is raised or disabled.
+
+#### 5. Update `openid-connect` Session Configuration for `lua-resty-session`
4.x
+
+PR: https://github.com/apache/apisix/pull/13178
+
+Contributor: [francescodedomenico](https://github.com/francescodedomenico)
+
+This PR updates the `openid-connect` session configuration schema to match
`lua-resty-session` 4.x. Session options are now exposed under their real names
and validated explicitly, so settings such as session expiry actually take
effect instead of being silently ignored.
+
+### Traffic Handling and Request Safety
+
+#### 6. Add `graphql-proxy-cache` Plugin
+
+PR: https://github.com/apache/apisix/pull/13435
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR introduces the `graphql-proxy-cache` plugin, which caches GraphQL
query responses using APISIX's existing `proxy-cache` infrastructure. It helps
reduce repeated upstream work for cacheable GraphQL operations while supporting
both disk and memory cache strategies.
+
+#### 7. Add `proxy-buffering` Plugin
+
+PR: https://github.com/apache/apisix/pull/13446
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR adds the `proxy-buffering` plugin to control NGINX proxy buffering
behavior per route. Disabling buffering allows APISIX to stream responses
directly to clients, which is important for Server-Sent Events, streaming APIs,
and other real-time delivery scenarios.
+
+#### 8. Bound Request Body Reads in `forward-auth` and `ai-proxy`
+
+PR: https://github.com/apache/apisix/pull/13466
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR adds `max_req_body_size` to `forward-auth`, `ai-proxy`, and
`ai-proxy-multi`. The new limit prevents these plugins from buffering
arbitrarily large client bodies into worker memory and returns `413` for
oversized requests before parsing.
+
+#### 9. Bound Pipeline Item Count and Schema in `batch-requests`
+
+PR: https://github.com/apache/apisix/pull/13492
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR hardens the `batch-requests` plugin by adding a configurable
`max_pipeline_items` limit, requiring positive timeout values, and rejecting
undocumented fields in pipeline entries. A single batch request can no longer
fan out into an unbounded number of internal requests.
+
+#### 10. Support Per-Port PROXY Protocol for Stream TCP Proxy
+
+PR: https://github.com/apache/apisix/pull/13561
+
+Contributor: [nic-6443](https://github.com/nic-6443)
+
+This PR adds per-port PROXY protocol controls for stream TCP proxy entries.
Instead of applying PROXY protocol globally to every stream TCP port, users can
enable accepting or sending it on specific ports while preserving the existing
global defaults.
+
+#### 11. Add `max_post_args_readable_size` for `post_arg.*` Matching
+
+PR: https://github.com/apache/apisix/pull/13601
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR adds `apisix.max_post_args_readable_size` to bound request body reads
during `post_arg.*` route matching. Large JSON or multipart bodies are no
longer parsed into worker memory for route predicates; instead, oversized
bodies simply do not match those predicates.
+
+### AI Gateway Security, Caching, and Resilience
+
+#### 12. Add Fallback Retry Controls to `ai-proxy-multi`
+
+PR: https://github.com/apache/apisix/pull/13495
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR adds `max_retries` and `retry_on_failure_within_ms` to
`ai-proxy-multi` fallback behavior. Operators can bound how many alternate LLM
instances a request tries and avoid retrying slow failures that would only
increase latency and provider cost.
+
+#### 13. Add `fail_mode` for Consumer-Bound AI Plugin Handling
+
+PR: https://github.com/apache/apisix/pull/13489
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR adds a shared `fail_mode` mechanism for AI security plugins bound at
the Consumer or Service level. Plugins can now consistently skip, warn, or fail
closed when they receive plain HTTP, non-JSON, or non-AI requests that cannot
be inspected.
+
+#### 14. Add `ai-lakera-guard` Plugin
+
+PR: https://github.com/apache/apisix/pull/13570
+
+Contributor: [janiussyafiq](https://github.com/janiussyafiq)
+
+This PR adds the `ai-lakera-guard` plugin, integrating APISIX with Lakera
Guard to scan LLM requests at the gateway. It helps protect AI applications
from prompt injection, jailbreaks, PII leakage, policy violations, and
malicious links before traffic reaches backend LLM services.
+
+#### 15. Improve `ai-aliyun-content-moderation` Performance and Request Scope
+
+PR: https://github.com/apache/apisix/pull/13598
+
+Contributor: [nic-6443](https://github.com/nic-6443)
+
+This PR optimizes `ai-aliyun-content-moderation` by replacing O(n²) content
chunking with an O(n) approach, reducing signing overhead, and reusing HTTP
connections within a request. It also adds role-aware `request_check_mode`,
allowing moderation to focus on the latest user input instead of rechecking the
full conversation every turn.
+
+#### 16. Add `ai-cache` Plugin
+
+PR: https://github.com/apache/apisix/pull/13578
+
+Contributor: [janiussyafiq](https://github.com/janiussyafiq)
+
+This PR adds the `ai-cache` plugin, which caches LLM responses for repeated
requests that resolve to the same prompt and AI instance. It can reduce
upstream token cost and latency for repetitive AI workloads such as FAQ bots,
document Q&A, and translation.
+
+#### 17. Add Response Scanning to `ai-lakera-guard`
+
+PR: https://github.com/apache/apisix/pull/13606
+
+Contributor: [janiussyafiq](https://github.com/janiussyafiq)
+
+This PR extends `ai-lakera-guard` with response scanning for both
non-streaming and streaming LLM traffic. Users can configure the plugin to
inspect input, output, or both, and flagged responses are replaced with
provider-compatible denial messages.
+
+### Observability and Logging
+
+#### 18. Add Built-In LLM Histograms to Prometheus Metrics
+
+PR: https://github.com/apache/apisix/pull/13487
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR enriches Prometheus metrics for AI Gateway traffic with clearer
latency and token distribution data. APISIX now distinguishes total LLM latency
from streaming time to first token and exposes prompt and completion token
histograms for quantile-based monitoring.
+
+#### 19. Add Built-In NGINX Variables for LLM Observability
+
+PR: https://github.com/apache/apisix/pull/13477
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR adds eight built-in `$llm_*` NGINX variables populated automatically
by `ai-proxy`. These variables expose LLM request and response metadata in
access logs and logger plugins without requiring custom parsing or additional
plugin configuration.
+
+#### 20. Support Disabling Prometheus Labels to Reduce Cardinality
+
+PR: https://github.com/apache/apisix/pull/13202
+
+Contributor: [janiussyafiq](https://github.com/janiussyafiq)
+
+This PR adds a `disabled_labels` option to Prometheus plugin metadata.
Operators can collapse selected built-in label values to reduce metric
cardinality while keeping the metric schema stable for dashboards, alerts, and
recording rules.
+
+#### 21. Add `log_format_extra` to Enrich Default Logger Output
+
+PR: https://github.com/apache/apisix/pull/13568
+
+Contributor: [shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+This PR adds `log_format_extra`, an additive logging format that preserves
APISIX's default rich log entry and overlays user-defined fields. It also
exposes `upstream_unresolved_host`, making it easier to log the configured
upstream hostname alongside the resolved upstream address.
+
+#### 22. Include AI Observability Variables in `llm_summary`
+
+PR: https://github.com/apache/apisix/pull/13609
+
+Contributor: [nic-6443](https://github.com/nic-6443)
+
+This PR adds AI observability fields such as stream mode, tool-call metadata,
cache token counts, reasoning tokens, and content risk level to `llm_summary`.
Logger plugins can now receive a fuller AI request summary automatically
instead of requiring every `$llm_*` variable to be added manually.
+
+### Rate Limiting and Credential Protection
+
+#### 23. Add Redis Sentinel, Sliding Window, and Delayed Sync to `limit-count`
+
+PR: https://github.com/apache/apisix/pull/13443
+
+Contributor: [AlinsRan](https://github.com/AlinsRan)
+
+This PR expands `limit-count` with a Redis Sentinel backend, sliding window
support, and delayed sync for Redis-based policies. These options improve high
availability, reduce fixed-window burstiness, and lower per-request Redis
overhead on high-traffic routes.
+
+#### 24. Encrypt Redis Passwords in Rate Limiting Plugins
+
+PR: https://github.com/apache/apisix/pull/13624
+
+Contributor: [nic-6443](https://github.com/nic-6443)
+
+This PR adds `redis_password` and `sentinel_password` fields to
`encrypt_fields` for `limit-count`, `limit-req`, and `limit-conn`. When data
encryption is enabled, Redis credentials used by rate limiting plugins are now
encrypted at rest in etcd and decrypted transparently at runtime.
+
+## Conclusion
+
+The [official website](https://apisix.apache.org/) and [GitHub
Issues](https://github.com/apache/apisix/issues) of Apache APISIX provide a
wealth of documentation of tutorials, and real-world use cases. If you
encounter any issues, you can refer to the documentation, search for keywords
in Issues, or participate in discussions on Issues to share your ideas and
practical experiences.
diff --git a/blog/zh/blog/2026/06/30/2026-jun-monthly-report.md
b/blog/zh/blog/2026/06/30/2026-jun-monthly-report.md
new file mode 100644
index 00000000000..1e1ea86b9ee
--- /dev/null
+++ b/blog/zh/blog/2026/06/30/2026-jun-monthly-report.md
@@ -0,0 +1,233 @@
+---
+title: "2026 社区月报 (06.01 - 06.30)"
+keywords: ["Apache APISIX", "API 网关", "社区月报", "贡献者"]
+description: Apache APISIX 社区的月报旨在帮助社区成员更全面地了解社区的最新动态，方便大家参与到 Apache APISIX
社区中来。
+tags: [Community]
+image:
https://static.api7.ai/uploads/2026/06/30/jAdrrmmG_2026-06-monthly-report-cover-cn.webp
+---
+
+> 最近，我们引入并更新了一些新功能，包括 SAML 认证、GraphQL 缓存、路由级代理缓冲控制、更丰富的 AI 可观测性、新增 LLM
安全与缓存插件、更安全的请求体读取，以及增强的限流后端等。有关更多细节，请阅读本期月报。
+
+<!--truncate-->
+
+## 导语
+
+Apache APISIX 项目始终秉承着开源社区协作的精神，自问世起便崭露头角，如今已经成为全球最活跃的开源 API
网关项目之一。正如谚语所言，"众人拾柴火焰高"，这一辉煌成就，得益于整个社区伙伴的协同努力。
+
+从 2026.06.01 至 2026.06.30，有 16 名开发者提交了 137 个 commits，为 Apache APISIX
做出了重要贡献。感谢这些伙伴们对 Apache APISIX 的无私支持！正是因为你们的付出，才能让 Apache APISIX 项目不断改进、提升和壮大。
+
+## 贡献者统计
+
+
+
+
+
+## 近期亮点功能
+
+以下是本月的重点更新，并按功能方向进行归类。
+
+### 认证与身份管理
+
+#### 1. 新增 `saml-auth` 插件
+
+相关 PR：https://github.com/apache/apisix/pull/13346
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 新增 `saml-auth` 插件，将 SAML 2.0 认证能力引入 APISIX 网关层。依赖 SAML 身份提供方的组织，可以通过
APISIX 保护上游服务，而无需在每个业务应用中分别实现 SAML 处理逻辑。
+
+#### 2. `openid-connect` 本地 JWT 校验模式下 `client_secret` 可选
+
+相关 PR：https://github.com/apache/apisix/pull/13472
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 让 `openid-connect` 在仅进行本地 Bearer Token 校验的模式下不再强制要求
`client_secret`，例如基于公钥或 JWKS 的 JWT 校验。对于不会调用身份提供方 token 或 introspection
端点的流程，用户无需再配置无意义的占位密钥。
+
+#### 3. 移除 `jwe-decrypt` 的服务端 Token 生成端点
+
+相关 PR：https://github.com/apache/apisix/pull/13464
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 移除了 `jwe-decrypt` 插件中未完成的 `/apisix/plugin/jwe/encrypt` 辅助端点。APISIX
现在更专注于在网关层解密 JWE，而 Token 生成则由持有消费者密钥的服务负责。
+
+#### 4. `hmac-auth` 新增 `max_req_body_size`
+
+相关 PR：https://github.com/apache/apisix/pull/13478
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 为启用请求体验证的 `hmac-auth` 路由新增 `max_req_body_size` 选项。超大请求体可以在被 APISIX
缓冲进内存之前被拒绝，从而提升全局 NGINX 请求体限制被调高或关闭时的安全性。
+
+#### 5. `openid-connect` Session 配置适配 `lua-resty-session` 4.x
+
+相关 PR：https://github.com/apache/apisix/pull/13178
+
+贡献者：[francescodedomenico](https://github.com/francescodedomenico)
+
+本 PR 更新 `openid-connect` 的 session 配置 schema，使其与 `lua-resty-session` 4.x
保持一致。Session 选项现在以真实字段名暴露并进行显式校验，避免诸如过期时间等配置被静默忽略。
+
+### 流量处理与请求安全
+
+#### 6. 新增 `graphql-proxy-cache` 插件
+
+相关 PR：https://github.com/apache/apisix/pull/13435
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 引入 `graphql-proxy-cache` 插件，基于 APISIX 现有的 `proxy-cache` 基础设施缓存 GraphQL
查询响应。对于可缓存的 GraphQL 操作，该插件可以减少重复上游计算，并支持磁盘和内存两种缓存策略。
+
+#### 7. 新增 `proxy-buffering` 插件
+
+相关 PR：https://github.com/apache/apisix/pull/13446
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 新增 `proxy-buffering` 插件，用于按路由控制 NGINX 代理缓冲行为。关闭缓冲后，APISIX
可以将响应直接流式传输给客户端，这对 Server-Sent Events、流式 API 和实时数据传递等场景非常重要。
+
+#### 8. `forward-auth` 与 `ai-proxy` 限制请求体读取大小
+
+相关 PR：https://github.com/apache/apisix/pull/13466
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 为 `forward-auth`、`ai-proxy` 和 `ai-proxy-multi` 新增
`max_req_body_size`。这些插件在解析前即可拒绝超大请求体，避免客户端让 worker 将任意大小的请求体缓冲到内存中。
+
+#### 9. `batch-requests` 限制 Pipeline 数量并收紧 Schema
+
+相关 PR：https://github.com/apache/apisix/pull/13492
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 强化 `batch-requests` 插件，新增可配置的 `max_pipeline_items` 限制，要求 timeout 为正数，并拒绝
pipeline 条目中的未文档化字段。单个批量请求不再能扩展为无限数量的内部请求。
+
+#### 10. Stream TCP Proxy 支持按端口配置 PROXY Protocol
+
+相关 PR：https://github.com/apache/apisix/pull/13561
+
+贡献者：[nic-6443](https://github.com/nic-6443)
+
+本 PR 为 stream TCP proxy 条目新增按端口控制 PROXY Protocol 的能力。用户可以在特定端口上启用接收或发送 PROXY
Protocol，而不是只能对所有 stream TCP 端口应用全局配置。
+
+#### 11. 新增 `max_post_args_readable_size` 限制 `post_arg.*` 匹配读取
+
+相关 PR：https://github.com/apache/apisix/pull/13601
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 新增 `apisix.max_post_args_readable_size`，限制 `post_arg.*`
路由匹配过程中读取请求体的大小。对于过大的 JSON 或 multipart 请求体，APISIX 不再为了路由谓词将其解析进 worker
内存，而是让相关谓词不匹配。
+
+### AI Gateway 安全、缓存与弹性
+
+#### 12. `ai-proxy-multi` 新增回退重试控制
+
+相关 PR：https://github.com/apache/apisix/pull/13495
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 为 `ai-proxy-multi` 的 fallback 行为新增 `max_retries` 和
`retry_on_failure_within_ms`。运维人员可以限制单个请求尝试的备用 LLM
实例数量，并避免对慢失败进行重试，从而减少额外延迟和模型调用成本。
+
+#### 13. AI 插件新增 Consumer 绑定场景的 `fail_mode`
+
+相关 PR：https://github.com/apache/apisix/pull/13489
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 为绑定在 Consumer 或 Service 级别的 AI 安全插件新增统一的 `fail_mode` 机制。当插件收到无法检查的普通
HTTP、非 JSON 或非 AI 请求时，可以一致地选择跳过、告警或失败关闭。
+
+#### 14. 新增 `ai-lakera-guard` 插件
+
+相关 PR：https://github.com/apache/apisix/pull/13570
+
+贡献者：[janiussyafiq](https://github.com/janiussyafiq)
+
+本 PR 新增 `ai-lakera-guard` 插件，将 APISIX 与 Lakera Guard 集成，在网关层扫描 LLM 请求。该插件可帮助
AI 应用在请求到达后端模型服务前识别 prompt injection、越狱、PII 泄露、内容策略违规和恶意链接等风险。
+
+#### 15. 优化 `ai-aliyun-content-moderation` 性能和请求检查范围
+
+相关 PR：https://github.com/apache/apisix/pull/13598
+
+贡献者：[nic-6443](https://github.com/nic-6443)
+
+本 PR 优化 `ai-aliyun-content-moderation`，将 O(n²) 的内容切分改为 O(n)，降低签名开销，并在单个请求内复用
HTTP 连接。同时新增按角色感知的 `request_check_mode`，让审核可以只关注最新用户输入，而不必每轮都重新审核完整对话。
+
+#### 16. 新增 `ai-cache` 插件
+
+相关 PR：https://github.com/apache/apisix/pull/13578
+
+贡献者：[janiussyafiq](https://github.com/janiussyafiq)
+
+本 PR 新增 `ai-cache` 插件，用于缓存解析为相同 prompt 和 AI 实例的 LLM 响应。对于 FAQ 机器人、文档问答、翻译等重复性
AI 工作负载，该插件可以减少上游 token 成本和响应延迟。
+
+#### 17. `ai-lakera-guard` 支持响应扫描
+
+相关 PR：https://github.com/apache/apisix/pull/13606
+
+贡献者：[janiussyafiq](https://github.com/janiussyafiq)
+
+本 PR 为 `ai-lakera-guard` 增加非流式和流式 LLM
响应扫描能力。用户可以配置插件检查输入、输出或两者；当响应被判定存在风险时，插件会用兼容提供商协议的拒绝消息替换原始响应。
+
+### 可观测性与日志
+
+#### 18. Prometheus 指标新增内置 LLM 直方图
+
+相关 PR：https://github.com/apache/apisix/pull/13487
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 为 AI Gateway 流量增强 Prometheus 指标，提供更清晰的延迟和 token 分布数据。APISIX 现在可以区分 LLM
总延迟和流式响应首 token 时间，并暴露 prompt 与 completion token 直方图，便于基于分位数进行监控。
+
+#### 19. 新增 LLM 可观测性 NGINX 内置变量
+
+相关 PR：https://github.com/apache/apisix/pull/13477
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 新增 8 个由 `ai-proxy` 自动填充的 `$llm_*` NGINX 变量。这些变量可以在访问日志和各类 logger 插件中暴露
LLM 请求与响应元数据，无需额外解析或插件配置。
+
+#### 20. Prometheus 支持禁用标签值以降低基数
+
+相关 PR：https://github.com/apache/apisix/pull/13202
+
+贡献者：[janiussyafiq](https://github.com/janiussyafiq)
+
+本 PR 在 Prometheus 插件元数据中新增 `disabled_labels`
选项。运维人员可以将选定内置标签的值折叠为空字符串，以降低指标基数，同时保持指标 schema 对仪表盘、告警和 recording rules 的兼容性。
+
+#### 21. 新增 `log_format_extra` 以扩展默认日志输出
+
+相关 PR：https://github.com/apache/apisix/pull/13568
+
+贡献者：[shreemaan-abhishek](https://github.com/shreemaan-abhishek)
+
+本 PR 新增 `log_format_extra`，允许在保留 APISIX 默认丰富日志字段的基础上叠加用户自定义字段。同时新增
`upstream_unresolved_host` 变量，便于记录配置中的上游主机名以及解析后的上游地址。
+
+#### 22. `llm_summary` 纳入 AI 可观测性变量
+
+相关 PR：https://github.com/apache/apisix/pull/13609
+
+贡献者：[nic-6443](https://github.com/nic-6443)
+
+本 PR 将流式模式、工具调用信息、缓存 token、reasoning token 和内容风险等级等 AI 可观测性字段加入
`llm_summary`。Logger 插件现在可以自动获得更完整的 AI 请求摘要，而无需逐个手动添加 `$llm_*` 变量。
+
+### 限流与凭证保护
+
+#### 23. `limit-count` 支持 Redis Sentinel、滑动窗口和延迟同步
+
+相关 PR：https://github.com/apache/apisix/pull/13443
+
+贡献者：[AlinsRan](https://github.com/AlinsRan)
+
+本 PR 为 `limit-count` 扩展 Redis Sentinel 后端、滑动窗口和 Redis
策略的延迟同步能力。这些能力可以提升高可用性，减少固定窗口边界突发，并降低高流量路由上的逐请求 Redis 开销。
+
+#### 24. 限流插件中的 Redis 密码支持加密存储
+
+相关 PR：https://github.com/apache/apisix/pull/13624
+
+贡献者：[nic-6443](https://github.com/nic-6443)
+
+本 PR 为 `limit-count`、`limit-req` 和 `limit-conn` 的 `redis_password` 与
`sentinel_password` 字段加入 `encrypt_fields`。启用数据加密后，限流插件使用的 Redis 凭证会在 etcd
中加密存储，并在运行时透明解密。
+
+## 结语
+
+Apache APISIX 的项目[官网](https://apisix.apache.org/zh/)和 GitHub 上的
[Issues](https://github.com/apache/apisix/issues)
上已经积累了比较丰富的文档教程和使用经验，如果您遇到问题可以翻阅文档，用关键词在 Issues 中搜索，也可以参与 Issues
上的讨论，提出自己的想法和实践经验。
