-
#15819 cafec4e Thanks @delucis! - Upgrade to Vite v8
-
#16965 57ead0d Thanks @Princesseuh! - Makes 'jsx' the default value for compressHTML
Astro now strips whitespace from your HTML using JSX rules by default, the same way frameworks like React do. Whitespace and line breaks around elements are removed, but meaningful whitespace within a single line — like a space between two inline elements — is preserved. To keep a space that would otherwise be removed, write it explicitly in your source, for example with {" "}.
This can change rendered output where whitespace between inline elements was previously meaningful. To keep Astro’s earlier behavior, set compressHTML: true for HTML-aware compression, or compressHTML: false to preserve all whitespace.
-
#16610 c63e7e4 Thanks @matthewp! - Adds background dev server management for AI coding agents.
When an AI coding agent is detected, astro dev now automatically starts the dev server as a detached background process. This prevents the dev server from blocking the agent’s terminal and allows it to continue working while the server runs.
A lock file (.astro/dev.json) is written when the dev server starts, recording the server’s URL, port, and PID. This prevents duplicate servers from being started for the same project.
New flag and subcommands
-
astro dev --background — Start the dev server as a background process (this is what runs automatically when an agent is detected).
-
astro dev stop — Stop a running background dev server.
-
astro dev status — Check if a dev server is running and display its URL, PID, and uptime.
-
astro dev logs — View logs from a background dev server. Use --follow (-f) to stream new output as it’s written.
These allow you to start and manage dev servers programmatically and were designed with AI coding agents in mind.
What should I do?
No action is required. If you are not using an AI coding agent, astro dev behaves exactly as before. If you are using an agent, background mode is enabled automatically — the agent will receive the server URL and PID, and can use astro dev stop to shut it down.
To opt out of automatic background mode when an agent is detected, set the environment variable ASTRO_DEV_BACKGROUND=0 before running astro dev.
-
#17010 0606073 Thanks @ocavue! - Removes the @astrojs/db package as it is no longer maintained.
The @astrojs/db package were deprecated in v6.4.5 and is now removed. This means the astro db, astro login, astro logout, astro link, and astro init CLI commands have also been removed.
If you were using Astro DB in your project, remove @astrojs/db from your project’s dependencies and replace it with one of the following alternatives:
-
Node.js built-in SQLite: Node.js now includes a built-in
node:sqlite module (available since Node.js v22.5.0). This is a good option if you are using the Node.js adapter and were using @astrojs/db for local SQLite storage.
-
Drizzle ORM: If you were using
@astrojs/db for its Drizzle-based schema and query API, you can use Drizzle directly with any supported database.
-
Other database libraries: Use any database library that suits your deployment platform (e.g. Turso, PlanetScale, Neon).
-
#16462 c30a778 Thanks @Princesseuh! - Replaces the Go compiler with a Rust-based version.
The Rust-based Astro compiler (@astrojs/compiler-rs) is now the default compiler. This new compiler is faster and more reliable, leading to faster build times and iteration cycles during development.
This new compiler is more strict regarding invalid syntax. For example, unclosed HTML tags will now throw an error instead of being ignored. It also does not attempt to correct semantically invalid HTML anymore, instead leaving it to the browser to handle, similar to other tools or document.write() in JavaScript.
The previous Go-based compiler has been removed, along with the experimental.rustCompiler flag used to opt into the Rust compiler. If you were setting experimental.rustCompiler in your astro.config.mjs, you can now remove it. No other action is required.
-
#16966 6650ec2 Thanks @Princesseuh! - Makes Sätteri the default Markdown processor
Astro now renders .md files with satteri() from @astrojs/markdown-satteri, its native Markdown pipeline, instead of the remark/rehype pipeline. @astrojs/markdown-remark is no longer installed by default.
To keep using the remark/rehype pipeline, install @astrojs/markdown-remark and set it as your processor:
// astro.config.mjs
import { defineConfig } from 'astro/config';
import { unified } from '@astrojs/markdown-remark';
export default defineConfig({
markdown: {
processor: unified(),
},
});
The deprecated markdown.remarkPlugins, markdown.rehypePlugins, and markdown.remarkRehype options still work, but now require @astrojs/markdown-remark to be used.
-
#16877 3b7d76e Thanks @matthewp! - Enables advanced routing by default.
The advanced routing feature introduced behind a flag in v6.3.0 is no longer experimental and is now enabled by default.
This gives full control over how requests flow through your application, with first-class support for frameworks like Hono.
Advanced routing now uses src/fetch.ts as default entrypoint instead of src/app.ts.
If you were previously using this feature without a custom entrypoint, please configure fetchFile or rename your entrypoint to src/fetch.ts, and then remove the experimental flag from your Astro config:
import { defineConfig } from 'astro/config';
export default defineConfig({
experimental {
- advancedRouting: true,
},
+ fetchFile: 'app.ts' // optional, you only need this if you cannot rename your entrypoint.
});
fetchFile is now a top-level config option instead of being nested under experimental.advancedRouting. If you were using a custom entrypoint, please update your Astro config to move its configuration:
// astro.config.mjs
export default defineConfig({
- experimental: {
- advancedRouting: {
- fetchFile: 'my-custom-entrypoint.ts',
- },
- },
+ fetchFile: 'my-custom-entrypoint.ts',
})
You can also set fetchFile: null to disable the entrypoint if you are using src/fetch.ts for another purpose, or don’t need advanced routing features.
If you have been waiting for stabilization before using advanced routing, you can now do so.
Please see the advanced routing guide in docs for more about this feature.
-
#16725 10229f7 Thanks @ArmandPhilippot! - Removes deprecated APIs exported from astro:transitions.
In Astro 6.x, some helpers available in astro:transitions and astro:transitions/client were deprecated.
In Astro 7.0, the following APIs can no longer be used in your project:
-
TRANSITION_BEFORE_PREPARATION
-
TRANSITION_AFTER_PREPARATION
-
TRANSITION_BEFORE_SWAP
-
TRANSITION_AFTER_SWAP
-
TRANSITION_PAGE_LOAD
-
isTransitionBeforePreparationEvent()
-
isTransitionBeforeSwapEvent()
-
createAnimationScope()
What should I do?
Remove any occurrence of createAnimationScope():
-import { createAnimationScope } from 'astro:transitions';
Replace any occurrence of the other APIs using the lifecycle event names directly:
-import {
- TRANSITION_AFTER_SWAP,
- isTransitionBeforePreparationEvent,
-} from 'astro:transitions/client';
-console.log(isTransitionBeforePreparationEvent(event));
+console.log(event.type === 'astro:before-preparation');
-console.log(TRANSITION_AFTER_SWAP);
+console.log('astro:after-swap');
Learn more about all utilities available in the View Transitions Router API Reference.
-
#16998 57dcc31 Thanks @matthewp! - Exposes getFetchState() from astro/hono as a public API
The getFetchState() function retrieves or lazily creates a FetchState from a Hono context object. This allows third-party packages to build Hono middleware that interacts with Astro’s per-request state, giving the astro/hono API the same extensibility as astro/fetch.
import { Hono } from 'hono';
import { getFetchState, pages } from 'astro/hono';
const app = new Hono();
app.use(async (context, next) => {
const state = getFetchState(context);
state.locals.message = 'Hello from custom middleware';
await next();
});
app.use(pages());
export default app;
-
#16996 300641e Thanks @florian-lefebvre! - Adds a subset field to the FontData type exposed via fontData from astro:assets. When using multiple font subsets (e.g., subsets: ["latin", "korean"]), each font data entry now includes the subset name, making it possible to distinguish between font entries for different subsets that share the same weight and style.
-
#16745 f864a80 Thanks @ematipico! - The custom logger feature introduced behind a flag in v6.2.0 is no longer experimental and is available for general use.
This feature provides better control over Astro’s logging infrastructure by allowing you to replace the default console output with custom logging implementations (e.g., structured JSON). This is particularly useful for on-demand rendering when connecting to log aggregation services such as Kibana, Logstash, CloudWatch, Grafana, or Loki.
Astro provides three built-in log handlers (json, node, and console), and you can also create your own.
JSON logging
import { defineConfig, logHandlers } from 'astro/config';
export default defineConfig({
logger: logHandlers.json({
pretty: true,
level: 'warn',
}),
});
Custom logger
import { defineConfig } from 'astro/config';
export default defineConfig({
logger: {
entrypoint: '@org/custom-logger',
},
});
Additionally, context.logger is now always available in API routes and middleware, even without a custom logger configured.
If you were previously using this feature, please remove the experimental flag from your Astro config:
import { defineConfig } from 'astro/config';
export default defineConfig({
- experimental: {
- logger: {
- entrypoint: '@org/custom-logger',
- },
- },
+ logger: {
+ entrypoint: '@org/custom-logger',
+ },
});
If you have been waiting for stabilization before using custom loggers, you can now do so.
Please see the Logger docs for more about this feature.
-
#16981 0d6d644 Thanks @ematipico! - Removes the setting experimental.queuedRendering. The new rendering engine is now stable and replaces the old one.
As part of the stabilization, the queued rendering has been improved, and some features have been removed:
-
The construction of the queue has been removed, instead now Astro uses a streaming approach where components are rendered and flushed as they are encountered.
-
The node polling feature has been removed because it doesn’t yield concrete savings.
-
The content cache has been descoped, and how only tag names are cached.
If you were previously using this experimental feature, you must remove this experimental flag from your configuration as it no longer exists:
// astro.config.mjs
import { defineConfig } from "astro/config";
export default defineConfig({
experimental: {
- queuedRendering: {}
}
});
-
#17116 f95e58e Thanks @ascorbic! - Stabilizes route caching, removing the experimental.cache and experimental.routeRules flags and replacing them with the top-level cache and routeRules configuration options.
Route caching, introduced experimentally in v6.0.0, is now stable. It gives you a platform-agnostic way to cache responses from on-demand rendered pages and endpoints, based on standard HTTP caching semantics.
Update your config to move cache and routeRules out of the experimental block:
// astro.config.mjs
import { defineConfig, memoryCache } from 'astro/config';
export default defineConfig({
- experimental: {
- cache: {
- provider: memoryCache(),
- },
- routeRules: {
- '/blog/[...path]': { maxAge: 300, swr: 60 },
- },
- },
+ cache: {
+ provider: memoryCache(),
+ },
+ routeRules: {
+ '/blog/[...path]': { maxAge: 300, swr: 60 },
+ },
});
Set caching directives in your routes with Astro.cache (in .astro pages) or context.cache (in API routes and middleware), and Astro translates them into the appropriate headers or runtime behavior depending on your configured cache provider. You can also define cache rules for routes declaratively in your config using routeRules, without modifying route code.
See the route caching guide for more information.
-
#16980 1f07343 Thanks @matthewp! - Removes state.provide(), state.resolve(), state.finalizeAll(), and App.Providers from the public advanced routing API. These context provider extension points are now internal-only. If you were using them in an integration, use locals to share per-request state instead.
-
#17111 c0f33ed Thanks @ematipico! - Harden the limits on the number of decoding on the URL.
-
#16982 1e000e2 Thanks @matthewp! - Improves the warning when accessing Astro.session without session storage configured. The session property is now always defined on the context object, and accessing it without configuration logs a helpful message instead of silently returning undefined.
-
#16335 9a53f77 Thanks @ascorbic! - Adds shared helper utilities for CDN cache provider authors for route caching
Exports astro/cache/provider-utils with helpers for building platform-specific cache-control headers, generating path-based invalidation tags, and normalizing invalidation options. These are used internally by the first-party Netlify, Vercel, and Cloudflare cache providers.
-
#17095 e84ebc0 Thanks @matthewp! - Improves build performance by removing an unfiltered transform hook from the astro:head-metadata-build plugin. Head propagation modules are now identified by their module ID (?astroPropagatedAssets) instead of scanning every module’s source code.
-
#17041 4c4a91c Thanks @iseraph-dev! - Fixes a bug where the advanced routing astro/hono / astro/fetch pages() handler returned the host framework’s default Internal Server Error response instead of rendering the custom 500.astro page when a page threw during render. Unmatched requests with a prerendered (or absent) custom 404 page now render the 404 error page instead of failing the same way.
-
#17097 5e340d7 Thanks @iseraph-dev! - Fixes a bug where the advanced routing astro/hono / astro/fetch middleware() handler returned the host framework’s default Internal Server Error response instead of rendering the custom 500.astro page when middleware threw. Unmatched requests with a prerendered (or absent) custom 404 page now render the 404 error page instead of failing the same way. Errors surfaced through next (the host framework’s downstream chain) still propagate to the host’s own error handler.
-
#15819 cafec4e Thanks @delucis! - Fixes --port flag being ignored after a Vite-triggered server restart (e.g. when a .env file changes)
-
#17104 b074a37 Thanks @iseraph-dev! - Fixes the custom 500.astro page receiving an empty error prop when the error originated in middleware.
-
#17078 04547ec Thanks @astrobot-houston! - Fixes a spurious Astro.request.headers warning on prerendered pages when security.allowedDomains is configured. The internal allowedDomains header validation now skips prerendered routes, since they use synthetic requests with no real headers.
-
#16603 deaaf3f Thanks @alexanderniebuhr! - Removes the warning that Astro does not support vite v8, since Astro v7 does support vite v8
-
#16335 9a53f77 Thanks @ascorbic! - Passes the Request object to CacheProvider.setHeaders() for route caching
Cache providers now receive the incoming Request as a second argument to setHeaders(options, request). This allows CDN providers to read the request URL, headers, and other properties when generating cache response headers, for example to auto-tag responses with their pathname for path-based invalidation.
-
#17098 637a1b6 Thanks @matthewp! - Fixes internal Astro headers leaking from direct pages() handler responses
-
#17090 3cf76c0 Thanks @matthewp! - Fixes Vite and Rolldown build warnings
-
#16434 ee079d4 Thanks @ematipico! - Fixes an issue where i18n domains would return 404 when trailingSlash is set to never.
-
Updated dependencies [7e7ab87, ff7b718, 241250b]:
-
@astrojs/markdown-satteri@0.3.1
