The best way to control Claude Code, Codex and more from any computer or phone

main
md 293 lines 10.7 KB
Rendered Raw
1 ---
2 title: 2.0.0-beta.62 - Build-Once Workers & GitHub Action
3 date: 2026-07-14T06:00:00Z
4 category: release
5 excerpt: v2.0.0-beta.62 makes Workers build their layer stack once per isolate instead of on every request. It also adds an official GitHub Action, PlanetScale OAuth login, a custom Worker entry for Vite websites, and a batch of `alchemy dev` fixes.
6 ---
7
8 beta.62 makes Workers build their layer stack **once per isolate**
9 instead of on every request. It also adds an official **GitHub
10 Action**, **PlanetScale OAuth login**, a **custom Worker entry**
11 for Vite websites, and a batch of `alchemy dev` fixes.
12
13 :::caution
14 **Breaking changes** — migration for each is below.
15
16 - **Worker layers build once per isolate, not once per request**
17 ([#774](https://github.com/alchemy-run/alchemy/pull/774)).
18 Connections and pools move from init to the handler.
19 - **effect `>=4.0.0-beta.97` is required**
20 ([#801](https://github.com/alchemy-run/alchemy/pull/801),
21 [#813](https://github.com/alchemy-run/alchemy/pull/813)).
22 effect removed several `Schedule` APIs. Alchemy is migrated and
23 the peer floor is raised to match.
24 :::
25
26 ## Layers build once per isolate
27
28 The Worker bridge used to rebuild your entire layer stack on
29 **every request**. Every `Layer.effect`, config decode, and client
30 constructor re-ran per event. Now layers build once, on the first
31 event. Every later event reuses them
32 ([#774](https://github.com/alchemy-run/alchemy/pull/774)):
33
34 ```typescript
35 Effect.gen(function* () {
36 // init: runs once per isolate. beta.61 ran this on every request.
37 const client = yield* HttpClient.HttpClient;
38 const response = yield* client.get("https://example.com/config.json");
39 const config = yield* response.json;
40
41 return {
42 fetch: Effect.gen(function* () {
43 // handler: runs per event, reusing the cached config
44 return yield* HttpServerResponse.json(config);
45 }),
46 };
47 })
48 ```
49
50 Init cost is paid once at cold start. After that, every request
51 goes straight to the handler. Durable Objects get the same win:
52 one layer build shared across activations.
53
54 Sharing one build across events needs cross-request async. One
55 event awaits a layer build that another event started. workerd
56 normally forbids this: a promise created in one request context
57 can't be resolved from another. So alchemy enables the
58 [`handle_cross_request_promise_resolution`](https://developers.cloudflare.com/workers/configuration/compatibility-flags/#handle-cross-request-promise-resolution-correctly)
59 compatibility flag automatically for Workers with a compatibility
60 date before 2024-10-14. Later dates enable it by default. Each
61 awaiting event also pins the in-flight build with `ctx.waitUntil`,
62 so workerd doesn't drop the continuation when the original request
63 ends. Opting out with `no_handle_cross_request_promise_resolution`
64 is a deploy-time error.
65
66 There is one migration. Anything **disposable** moves out of init
67 and into the handler. The handler's per-event `Scope` runs its
68 finalizers after the response:
69
70 ```typescript
71 // ❌ init: workerd has no isolate teardown, so this finalizer never runs
72 const pool = yield* acquirePool();
73
74 // ✅ handler: acquired per event, released when the event settles
75 fetch: Effect.gen(function* () {
76 const conn = yield* acquireConn();
77 // ...
78 }),
79 ```
80
81 You can still declare the resource in init and use it at runtime.
82 An Effect is only a description, so declaring one in init acquires
83 nothing:
84
85 ```typescript
86 import * as PgClient from "@effect/sql-pg/PgClient";
87
88 Effect.gen(function* () {
89 // init: declare the client. Nothing connects yet.
90 const client = PgClient.make({ url });
91
92 return {
93 fetch: Effect.gen(function* () {
94 // handler: connects on this event's Scope
95 const sql = yield* client;
96 const rows = yield* sql`select * from users`;
97 return yield* HttpServerResponse.json(rows);
98 }),
99 };
100 })
101 ```
102
103 The pool is acquired on the event's `Scope` and released when the
104 event settles.
105
106 `proxyChain` removes the extra `yield*`. It wraps the deferred
107 client in a chainable handle. Yielding a query resolves the client
108 on the current event's `Scope`, then replays the chain against it:
109
110 ```typescript
111 import { proxyChain } from "alchemy/Util/proxy-chain";
112
113 Effect.gen(function* () {
114 // init: a chainable handle over the deferred client
115 const sql = proxyChain(PgClient.make({ url }));
116
117 return {
118 fetch: Effect.gen(function* () {
119 const rows = yield* sql`select * from users`;
120 return yield* HttpServerResponse.json(rows);
121 }),
122 };
123 })
124 ```
125
126 `Drizzle.postgres` is this pattern plus `Effect.cached` to memoize
127 the acquisition per event. The first query opens the pool, every
128 later query in the same event reuses it, and the pool closes when
129 the event settles:
130
131 ```typescript
132 Effect.gen(function* () {
133 // init: declares the handle. No connection is made.
134 const db = yield* Drizzle.postgres(hyperdrive.connectionString);
135
136 return {
137 fetch: Effect.gen(function* () {
138 // first query builds the pool, memoized on this event's Scope
139 const rows = yield* db.select().from(users);
140 return yield* HttpServerResponse.json(rows);
141 }),
142 };
143 })
144 ```
145
146 See
147 [Postgres.ts](https://github.com/alchemy-run/alchemy/blob/main/packages/alchemy/src/Drizzle/Postgres.ts)
148 for the full implementation. It is also why `Drizzle.postgres` now
149 works inside Durable Object methods.
150
151 Lambda gets a real shutdown from the same change. The generated
152 entry registers an internal extension, so the instance receives
153 SIGTERM and a ~500 ms window to run init-level finalizers.
154
155 Docs: [Instance scope vs request
156 scope](/infrastructure-as-effects/runtime#instance-scope-vs-request-scope).
157
158 ## Deploy from GitHub Actions
159
160 One step resolves the Alchemy lifecycle from the GitHub event.
161 Opening a PR deploys a `staging-{pr}` stage, closing it destroys
162 the stage, and pushing to main deploys `prod`
163 ([#699](https://github.com/alchemy-run/alchemy/pull/699)).
164 Thanks **Harry Solovay**!
165
166 ```yml
167 - uses: alchemy-run/alchemy@main
168 ```
169
170 A new `GitHub.GitHubEnv` config exposes the Actions metadata to
171 your stack. It resolves to `undefined` outside CI, so the same
172 stack runs unchanged locally:
173
174 ```typescript
175 import * as GitHub from "alchemy/GitHub";
176
177 const github = yield* GitHub.GitHubEnv;
178
179 if (github?.pr) {
180 yield* GitHub.Comment("preview-comment", {
181 owner: github.owner,
182 repository: github.repository,
183 issueNumber: github.pr,
184 body: `Preview for ${github.sha}`,
185 });
186 }
187 ```
188
189 Docs: [CI](/environments/ci).
190
191 ## PlanetScale OAuth login
192
193 `alchemy login` now offers browser-based OAuth for PlanetScale,
194 the same flow as Cloudflare
195 ([#467](https://github.com/alchemy-run/alchemy/pull/467)):
196
197 ```sh
198 alchemy login
199 # authorize in the browser → token stored, refreshed proactively
200 ```
201
202 Env vars and service tokens keep working. Both providers now land
203 on shared `/auth/success` and `/auth/error` pages
204 ([#780](https://github.com/alchemy-run/alchemy/pull/780)).
205
206 ## Custom Worker entry for Vite websites
207
208 `Cloudflare.Website.Vite` accepts a `main` prop that forwards a
209 custom Worker entry to the Vite plugin, in dev and deploy
210 ([#779](https://github.com/alchemy-run/alchemy/pull/779)).
211 Thanks **Daniel Gangl**!
212
213 ```typescript
214 // worker/index.ts: export more than the framework's fetch handler
215 import handler from "virtual:react-router/unstable_rsc/server";
216 export { MyDurableObject } from "./my-durable-object";
217 export default handler;
218 ```
219
220 ```typescript
221 yield* Cloudflare.Website.Vite("Site", {
222 rootDir: "./site",
223 main: "./site/worker/index.ts",
224 });
225 ```
226
227 React Router / RSC setups need this. The framework plugin owns the
228 entry, so the old `rollupOptions.input` workaround doesn't apply.
229
230 ## `alchemy dev` reliability
231
232 - **Hot reload no longer kills service bindings**
233 ([#812](https://github.com/alchemy-run/alchemy/pull/812)) —
234 reloading a worker used to delete its replacement's registration,
235 so RPC calls failed with `Worker not found` after every backend
236 change. Thanks **utopy**!
237 - **Container environment variables work in dev**
238 ([#785](https://github.com/alchemy-run/alchemy/pull/785)).
239 - **Dev container images persist across restarts**
240 ([#727](https://github.com/alchemy-run/alchemy/pull/727)).
241 - **Remote images and external Dockerfiles work in
242 `LocalContainerProvider`**
243 ([#790](https://github.com/alchemy-run/alchemy/pull/790)).
244 - **The provider RPC session is cached and reused**
245 ([#789](https://github.com/alchemy-run/alchemy/pull/789)).
246 - **Dev-URL detection favors localhost/IP URLs**
247 ([#787](https://github.com/alchemy-run/alchemy/pull/787)).
248
249 ## Also in this release
250
251 - **`plan` no longer claims ownership of adopted resources**
252 ([#794](https://github.com/alchemy-run/alchemy/pull/794)) —
253 a read-only dry run could silently adopt an unowned cloud
254 resource, and a later unrelated deploy would orphan-delete it.
255 Thanks **Daniel Gangl**!
256 - **State-store credentials are bound to the account**
257 ([#760](https://github.com/alchemy-run/alchemy/pull/760)) —
258 after `alchemy login --configure` switched accounts, state
259 reads/writes silently kept going to the old account's store.
260 - **Provider handlers receive the resource `fqn`**
261 ([#783](https://github.com/alchemy-run/alchemy/pull/783)) —
262 the fully-qualified name joins `id` in every lifecycle handler,
263 so custom providers can stamp collision-free ownership metadata.
264 Thanks **Daniel Gangl**!
265 - **Cron handlers keep their requirements**
266 ([#791](https://github.com/alchemy-run/alchemy/pull/791)) —
267 services required by a `Workers.cron` handler surface on the
268 Worker's init effect again. Thanks **Jack Bisceglia**!
269 - **Vite build output is read after the build resolves**
270 ([#795](https://github.com/alchemy-run/alchemy/pull/795)) —
271 fixes a bundling race in `Website` deploys.
272 - **Container healthcheck durations are unit-suffixed**
273 ([#778](https://github.com/alchemy-run/alchemy/pull/778)) —
274 Docker rejects bare numbers. Thanks **Matthew Aylward**!
275 - **`alchemy state clear` respects `ALCHEMY_PROFILE`**
276 ([ba705](https://github.com/alchemy-run/alchemy/commit/ba70585f)).
277 - **TypeScript 7 (stable) toolchain**
278 ([#804](https://github.com/alchemy-run/alchemy/pull/804)).
279 - **New guides**: [Worker Loader](/cloudflare/compute/worker-loader),
280 [Rate limiting](/cloudflare/compute/rate-limiting), and
281 [Workers Cache](/cloudflare/compute/cache)
282 ([#805](https://github.com/alchemy-run/alchemy/pull/805)),
283 plus a TanStack + Atom RPC + Drizzle example
284 ([#634](https://github.com/alchemy-run/alchemy/pull/634)).
285
286 ## Where to go next
287
288 - [Instance scope vs request scope](/infrastructure-as-effects/runtime#instance-scope-vs-request-scope)
289 - [CI](/environments/ci)
290 - [Workers](/cloudflare/compute/workers)
291 - [Workers Cache](/cloudflare/compute/cache)
292 - [CHANGELOG](https://github.com/alchemy-run/alchemy/blob/main/CHANGELOG.md#v200-beta62)
293 - [Compare v2.0.0-beta.61 → v2.0.0-beta.62](https://github.com/alchemy-run/alchemy/compare/v2.0.0-beta.61...v2.0.0-beta.62)