Release 3.0.0-beta.25

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

.changeset/angry-eagles-trade.md

@@ -0,0 +1,9 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+- Fix additionalFiles that aren't decendants
+- Stop swallowing uncaught exceptions in prod
+- Improve warnings and errors, fail early on critical warnings
+- New arg to --save-logs even for successful builds

.changeset/beige-pens-dance.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+v3 CLI update command and package manager detection fix

.changeset/big-tomatoes-deliver.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Hoist uncaughtException handler to the top of workers to better report error messages

.changeset/breezy-gorillas-mate.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+better handle task metadata parse errors, and display nicely formatted errors

.changeset/chilled-hornets-move.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Vastly improved dev command output

.changeset/clean-pianos-listen.md

@@ -0,0 +1,8 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core-apps": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+add machine config and secure zod connection

.changeset/clever-apes-collect.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Fix error stack traces

.changeset/clever-carrots-travel.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Implement task.onSuccess/onFailure and config.onSuccess/onFailure

.changeset/config.json

@@ -4,7 +4,8 @@
   "commit": false,
   "fixed": [
     [
-      "@trigger.dev/*"
+      "@trigger.dev/*",
+      "trigger.dev"
     ]
   ],
   "linked": [],
@@ -16,7 +17,10 @@
     "emails",
     "proxy",
     "yalt",
-    "@trigger.dev/database"
+    "@trigger.dev/database",
+    "coordinator",
+    "docker-provider",
+    "kubernetes-provider"
   ],
   "___experimentalUnsafeOptions_WILL_CHANGE_IN_PATCH": {
     "onlyUpdatePeerDependentsWhenOutOfRange": true

.changeset/cool-glasses-bake.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Configurable log levels in the config file and via env var

.changeset/cuddly-feet-approve.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Don’t swallow some error messages when deploying

.changeset/dry-walls-check.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Add option to print console logs in the dev CLI locally (issue #1014)

.changeset/eight-pumas-float.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Fixed batch otel flushing

.changeset/eleven-paws-join.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Remove the env var check during deploy (too many false negatives)

.changeset/few-students-share.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fix permissions inside node_modules

.changeset/funny-swans-destroy.md

@@ -0,0 +1,9 @@
+---
+"@trigger.dev/sdk": patch
+---
+
+Make msw a normal dependency (for now) to fix Module Not Found error in Next.js.
+
+It turns out that webpack will "hoist" dynamically imported modules and attempt to resolve them at build time, even though it's an optional peer dep:
+
+https://x.com/maverickdotdev/status/1782465214308319404

.changeset/green-bags-wink.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/react": patch
+---
+
+Fix for shared queryKey between useRunDetails and useRunStatuses

.changeset/khaki-apricots-design.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Retry 429, 500, and connection error API requests to the trigger.dev server

.changeset/khaki-poems-lay.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/sdk": patch
+---
+
+Export queue from the SDK

.changeset/late-icons-lie.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/core-apps": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Display errors for runs and deployments

.changeset/late-steaks-behave.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+v3: fix digest extraction

.changeset/lemon-jobs-repair.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/nestjs": patch
+---
+
+fix: [nestjs integration] fastify HTTP adapter detection now works correctly for response headers

.changeset/light-bulldogs-press.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Changed "Worker" to "Version" in the dev command key

.changeset/light-dragons-complain.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Correctly handle self-hosted deploy command errors

.changeset/little-crabs-cross.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Fix issues with consecutive waits

.changeset/loud-actors-remember.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Default to retrying enabled in dev when running init

.changeset/many-ligers-pump.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Handle string and non-stringifiable outputs like functions

.changeset/mighty-camels-joke.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Improve error messages during dev/deploy and handle deploy image build issues

.changeset/mighty-flowers-train.md

@@ -0,0 +1,83 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Adds support for `emitDecoratorMetadata: true` and `experimentalDecorators: true` in your tsconfig using the [`@anatine/esbuild-decorators`](https://github.com/anatine/esbuildnx/tree/main/packages/esbuild-decorators) package. This allows you to use libraries like TypeORM: 
+
+```ts orm/index.ts
+import "reflect-metadata";
+import { DataSource } from "typeorm";
+import { Entity, Column, PrimaryColumn } from "typeorm";
+
+@Entity()
+export class Photo {
+  @PrimaryColumn()
+  id!: number;
+
+  @Column()
+  name!: string;
+
+  @Column()
+  description!: string;
+
+  @Column()
+  filename!: string;
+
+  @Column()
+  views!: number;
+
+  @Column()
+  isPublished!: boolean;
+}
+
+export const AppDataSource = new DataSource({
+  type: "postgres",
+  host: "localhost",
+  port: 5432,
+  username: "postgres",
+  password: "postgres",
+  database: "v3-catalog",
+  entities: [Photo],
+  synchronize: true,
+  logging: false,
+});
+```
+
+And then in your trigger.config.ts file you can initialize the datasource using the new `init` option:
+
+```ts trigger.config.ts
+import type { TriggerConfig } from "@trigger.dev/sdk/v3";
+import { AppDataSource } from "@/trigger/orm";
+
+export const config: TriggerConfig = {
+  // ... other options here
+  init: async (payload, { ctx }) => {
+    await AppDataSource.initialize();
+  },
+};
+```
+
+Now you are ready to use this in your tasks:
+
+```ts
+import { task } from "@trigger.dev/sdk/v3";
+import { AppDataSource, Photo } from "./orm";
+
+export const taskThatUsesDecorators = task({
+  id: "taskThatUsesDecorators",
+  run: async (payload: { message: string }) => {
+    console.log("Creating a photo...");
+
+    const photo = new Photo();
+    photo.id = 2;
+    photo.name = "Me and Bears";
+    photo.description = "I am near polar bears";
+    photo.filename = "photo-with-bears.jpg";
+    photo.views = 1;
+    photo.isPublished = true;
+
+    await AppDataSource.manager.save(photo);
+  },
+});
+```

.changeset/nasty-jars-pump.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fixes an issue that was treating v2 trigger directories as v3

.changeset/new-rivers-tell.md

@@ -0,0 +1,6 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core": patch
+---
+
+Added replayRun function to the SDK

.changeset/ninety-pets-travel.md

@@ -0,0 +1,6 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core": patch
+---
+
+Improve the SDK function types and expose a new APIError instead of the APIResult type

.changeset/odd-poets-own.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Added a Node.js runtime check for the CLI

.changeset/polite-ducks-switch.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fix post start hooks

.changeset/polite-rockets-matter.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fix dev CLI output when not printing update messages

.changeset/poor-flowers-cross.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Use the dashboard url instead of the API url for the View logs link

.changeset/pre.json

@@ -0,0 +1,117 @@
+{
+  "mode": "pre",
+  "tag": "beta",
+  "initialVersions": {
+    "coordinator": "0.0.1",
+    "docker-provider": "0.0.1",
+    "kubernetes-provider": "0.0.1",
+    "proxy": "0.0.11",
+    "webapp": "1.0.0",
+    "yalt": "0.0.1",
+    "@trigger.dev/airtable": "2.3.18",
+    "@trigger.dev/github": "2.3.18",
+    "@trigger.dev/linear": "2.3.18",
+    "@trigger.dev/openai": "2.3.18",
+    "@trigger.dev/plain": "2.3.18",
+    "@trigger.dev/replicate": "2.3.18",
+    "@trigger.dev/resend": "2.3.18",
+    "@trigger.dev/sendgrid": "2.3.18",
+    "@trigger.dev/shopify": "2.3.18",
+    "@trigger.dev/slack": "2.3.18",
+    "@trigger.dev/stripe": "2.3.18",
+    "@trigger.dev/supabase": "2.3.18",
+    "@trigger.dev/typeform": "2.3.18",
+    "@trigger.dev/astro": "2.3.18",
+    "@trigger.dev/cli": "2.3.18",
+    "trigger.dev": "2.3.18",
+    "@trigger.dev/core": "2.3.18",
+    "@trigger.dev/core-apps": "0.0.0",
+    "@trigger.dev/core-backend": "2.3.18",
+    "@trigger.dev/database": "0.0.1",
+    "emails": "1.0.0",
+    "@trigger.dev/eslint-plugin": "2.3.18",
+    "@trigger.dev/express": "2.3.18",
+    "@trigger.dev/hono": "2.3.18",
+    "@trigger.dev/integration-kit": "2.3.18",
+    "@trigger.dev/nestjs": "2.3.18",
+    "@trigger.dev/nextjs": "2.3.18",
+    "@trigger.dev/otlp-importer": "2.3.11",
+    "@trigger.dev/react": "2.3.18",
+    "@trigger.dev/remix": "2.3.18",
+    "@trigger.dev/sveltekit": "2.3.18",
+    "@trigger.dev/testing": "2.3.18",
+    "@trigger.dev/sdk": "2.3.18",
+    "@trigger.dev/yalt": "2.3.18"
+  },
+  "changesets": [
+    "angry-eagles-trade",
+    "beige-pens-dance",
+    "big-tomatoes-deliver",
+    "breezy-gorillas-mate",
+    "chilled-hornets-move",
+    "clean-pianos-listen",
+    "clever-apes-collect",
+    "clever-carrots-travel",
+    "cool-glasses-bake",
+    "cuddly-feet-approve",
+    "dry-walls-check",
+    "eight-pumas-float",
+    "eleven-paws-join",
+    "few-students-share",
+    "funny-swans-destroy",
+    "green-bags-wink",
+    "khaki-apricots-design",
+    "khaki-poems-lay",
+    "late-icons-lie",
+    "late-steaks-behave",
+    "lemon-jobs-repair",
+    "light-bulldogs-press",
+    "light-dragons-complain",
+    "little-crabs-cross",
+    "loud-actors-remember",
+    "many-ligers-pump",
+    "mighty-camels-joke",
+    "mighty-flowers-train",
+    "nasty-jars-pump",
+    "new-rivers-tell",
+    "ninety-pets-travel",
+    "odd-poets-own",
+    "polite-ducks-switch",
+    "polite-rockets-matter",
+    "poor-flowers-cross",
+    "purple-garlics-shop",
+    "rare-lamps-promise",
+    "rare-roses-float",
+    "real-planets-stare",
+    "rich-kangaroos-unite",
+    "rotten-beers-refuse",
+    "rotten-dryers-exercise",
+    "selfish-ducks-sort",
+    "shaggy-spoons-taste",
+    "sharp-emus-compare",
+    "sharp-zebras-serve",
+    "shiny-coats-cry",
+    "silly-suits-switch",
+    "six-ligers-exist",
+    "slow-buses-own",
+    "smart-needles-move",
+    "smart-olives-eat",
+    "spicy-lamps-smoke",
+    "strange-ghosts-matter",
+    "strong-lemons-add",
+    "stupid-bulldogs-applaud",
+    "sweet-lizards-press",
+    "swift-dragons-peel",
+    "tall-bees-wave",
+    "tame-guests-know",
+    "tender-moose-tell",
+    "tender-oranges-rhyme",
+    "tidy-balloons-suffer",
+    "tidy-dryers-sleep",
+    "tiny-doors-type",
+    "tiny-elephants-scream",
+    "tricky-bulldogs-heal",
+    "tricky-ladybugs-unite",
+    "two-pumas-wait"
+  ]
+}

.changeset/purple-garlics-shop.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Add typescript as a dependency so the esbuild-decorator will work even when running in npx

.changeset/rare-lamps-promise.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Add a postInstall option to allow running scripts after dependencies have been installed in deployed images

.changeset/rare-roses-float.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Add openssl to prod worker image and allow passing auth token via env var for deploy

.changeset/real-planets-stare.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Fixed incorrect span timings around checkpoints by implementing a precise wall clock that resets after restores

.changeset/rich-kangaroos-unite.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fix a bug where revoking the CLI token would prevent you from ever logging in again with the CLI.

.changeset/rotten-beers-refuse.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Add git to prod worker image which fixes private package installs

.changeset/rotten-dryers-exercise.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Adding task with a triggerSource of schedule

.changeset/selfish-ducks-sort.md

@@ -0,0 +1,6 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+---
+
+Lock SDK and CLI deps on exact core version

.changeset/shaggy-spoons-taste.md

@@ -0,0 +1,56 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core": patch
+---
+
+Updates the `trigger`, `batchTrigger` and their `*AndWait` variants to use the first parameter for the payload/items, and the second parameter for options.
+
+Before:
+
+```ts
+await yourTask.trigger({ payload: { foo: "bar" }, options: { idempotencyKey: "key_1234" } });
+await yourTask.triggerAndWait({ payload: { foo: "bar" }, options: { idempotencyKey: "key_1234" } });
+
+await yourTask.batchTrigger({ items: [{ payload: { foo: "bar" } }, { payload: { foo: "baz" } }] });
+await yourTask.batchTriggerAndWait({ items: [{ payload: { foo: "bar" } }, { payload: { foo: "baz" } }] });
+```
+
+After:
+
+```ts
+await yourTask.trigger({ foo: "bar" }, { idempotencyKey: "key_1234" });
+await yourTask.triggerAndWait({ foo: "bar" }, { idempotencyKey: "key_1234" });
+
+await yourTask.batchTrigger([{ payload: { foo: "bar" } }, { payload: { foo: "baz" } }]);
+await yourTask.batchTriggerAndWait([{ payload: { foo: "bar" } }, { payload: { foo: "baz" } }]);
+```
+
+We've also changed the API of the `triggerAndWait` result. Before, if the subtask that was triggered finished with an error, we would automatically "rethrow" the error in the parent task.
+
+Now instead we're returning a `TaskRunResult` object that allows you to discriminate between successful and failed runs in the subtask:
+
+Before:
+
+```ts
+try {
+  const result = await yourTask.triggerAndWait({ foo: "bar" });
+
+  // result is the output of your task
+  console.log("result", result);
+
+} catch (error) {
+  // handle subtask errors here
+}
+```
+
+After:
+
+```ts
+const result = await yourTask.triggerAndWait({ foo: "bar" });
+
+if (result.ok) {
+  console.log(`Run ${result.id} succeeded with output`, result.output);
+} else {
+  console.log(`Run ${result.id} failed with error`, result.error);
+}
+```

.changeset/sharp-emus-compare.md

@@ -0,0 +1,13 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+When using idempotency keys, triggerAndWait and batchTriggerAndWait will still work even if the existing runs have already been completed (or even partially completed, in the case of batchTriggerAndWait)
+
+- TaskRunExecutionResult.id is now the run friendlyId, not the attempt friendlyId
+- A single TaskRun can now have many batchItems, in the case of batchTriggerAndWait while using idempotency keys
+- A run’s idempotencyKey is now added to the ctx as well as the TaskEvent and displayed in the span view
+- When resolving batchTriggerAndWait, the runtimes no longer reject promises, leading to an error in the parent task
+

.changeset/sharp-zebras-serve.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Update trigger.dev CLI for new batch otel support

.changeset/shiny-coats-cry.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Add additional logging around cleaning up dev workers, and always kill them after 5 seconds if they haven't already exited

.changeset/silly-suits-switch.md

@@ -0,0 +1,6 @@
+---
+"@trigger.dev/otlp-importer": patch
+"trigger.dev": patch
+---
+
+Fix package builds and CLI commands on Windows

.changeset/six-ligers-exist.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Improve the display of non-object return types in the run trace viewer

.changeset/slow-buses-own.md

@@ -0,0 +1,8 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+"@trigger.dev/cli": patch
+---
+
+Move to our global system from AsyncLocalStorage for the current task context storage

.changeset/smart-needles-move.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/sdk": patch
+---
+
+Remove unimplemented batchOptions

.changeset/smart-olives-eat.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fix CLI logout and add list-profiles command

.changeset/spicy-lamps-smoke.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fixing an issue with bundling @trigger.dev/core/v3 in dev when using pnpm

.changeset/strange-ghosts-matter.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Added DEBUG to the ignored env vars

.changeset/strong-lemons-add.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Make optional schedule object fields nullish

.changeset/stupid-bulldogs-applaud.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Fixed an issue where the trigger.dev package was not being built before publishing to npm

.changeset/sweet-lizards-press.md

@@ -0,0 +1,8 @@
+---
+"trigger.dev": major
+"@trigger.dev/core": major
+"@trigger.dev/otlp-importer": major
+"@trigger.dev/sdk": major
+---
+
+Updates to support Trigger.dev v3

.changeset/swift-dragons-peel.md

@@ -0,0 +1,5 @@
+---
+"@trigger.dev/core": patch
+---
+
+Added JSDocs to the schedule SDK types

.changeset/tall-bees-wave.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Dynamically import superjson and fix some bundling issues

.changeset/tame-guests-know.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Changed the binary name from trigger.dev to triggerdev to fix a Windows issue

.changeset/tender-moose-tell.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Use locked package versions when resolving dependencies in deployed workers

.changeset/tender-oranges-rhyme.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Stop swallowing deployment errors and display them better

.changeset/tidy-balloons-suffer.md

@@ -0,0 +1,5 @@
+---
+"trigger.dev": patch
+---
+
+Init command was failing on Windows because of bad template paths

.changeset/tidy-dryers-sleep.md

@@ -0,0 +1,10 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Fixes an issue that caused failed tasks when resuming after calling `triggerAndWait` or `batchTriggerAndWait` in prod/staging (this doesn't effect dev).
+
+The version of Node.js we use for deployed workers (latest 20) would crash with an out-of-memory error when the checkpoint was restored. This crash does not happen on Node 18x or Node21x, so we've decided to upgrade the worker version to Node.js21x, to mitigate this issue.
+
+You'll need to re-deploy to production to fix the issue.

.changeset/tiny-doors-type.md

@@ -0,0 +1,6 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core": patch
+---
+
+Added cancelRun to the SDK

.changeset/tiny-elephants-scream.md

@@ -0,0 +1,7 @@
+---
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+- Add graceful exit for prod workers
+- Prevent overflow in long waits

.changeset/tricky-bulldogs-heal.md

@@ -0,0 +1,7 @@
+---
+"@trigger.dev/sdk": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Added a new global - Task Catalog - to better handle task metadata

.changeset/tricky-ladybugs-unite.md

@@ -0,0 +1,8 @@
+---
+"@trigger.dev/sdk": patch
+"@trigger.dev/core-apps": patch
+"trigger.dev": patch
+"@trigger.dev/core": patch
+---
+
+Extracting out all the non-SDK related features from the main @trigger.dev/core/v3 export

.changeset/two-pumas-wait.md

@@ -0,0 +1,6 @@
+---
+"trigger.dev": patch
+"@trigger.dev/cli": patch
+---
+
+Add support for tasks located in subdirectories inside trigger dirs

.env.example

@@ -31,6 +31,9 @@ NODE_ENV=development
 # FROM_EMAIL=
 # REPLY_TO_EMAIL=
 
+# Remove the following line to enable logging telemetry traces to the console
+LOG_TELEMETRY="false"
+
 # CLOUD VARIABLES
 POSTHOG_PROJECT_KEY=
 PLAIN_API_KEY= 
@@ -42,4 +45,31 @@ CLOUD_LINEAR_CLIENT_ID=
 CLOUD_LINEAR_CLIENT_SECRET=
 CLOUD_SLACK_APP_HOST=
 CLOUD_SLACK_CLIENT_ID=
-CLOUD_SLACK_CLIENT_SECRET= 
+CLOUD_SLACK_CLIENT_SECRET= 
+
+# v3 variables
+PROVIDER_SECRET=provider-secret # generate the actual secret with `openssl rand -hex 32`
+COORDINATOR_SECRET=coordinator-secret # generate the actual secret with `openssl rand -hex 32`
+
+# Uncomment the following line to enable the registry proxy
+# ENABLE_REGISTRY_PROXY=true
+# DEPOT_TOKEN=<Depot org token>
+# DEPOT_PROJECT_ID=<Depot project id>
+# DEPLOY_REGISTRY_HOST=${APP_ORIGIN} # This is the host that the deploy CLI will use to push images to the registry
+# CONTAINER_REGISTRY_ORIGIN=<Container registry origin e.g. https://registry.digitalocean.com>
+# CONTAINER_REGISTRY_USERNAME=<Container registry username e.g. Digital ocean email address>
+# CONTAINER_REGISTRY_PASSWORD=<Container registry password e.g. Digital ocean PAT>
+# DEV_OTEL_EXPORTER_OTLP_ENDPOINT="http://0.0.0.0:4318"
+# These are needed for the object store (for handling large payloads/outputs)
+# OBJECT_STORE_BASE_URL="https://{bucket}.{accountId}.r2.cloudflarestorage.com"
+# OBJECT_STORE_ACCESS_KEY_ID=
+# OBJECT_STORE_SECRET_ACCESS_KEY=
+# RUNTIME_WAIT_THRESHOLD_IN_MS=10000
+
+# These control the server-side internal telemetry
+# INTERNAL_OTEL_TRACE_EXPORTER_URL=<URL to send traces to>
+# INTERNAL_OTEL_TRACE_EXPORTER_AUTH_HEADER_NAME=<Header name for the auth token>
+# INTERNAL_OTEL_TRACE_EXPORTER_AUTH_HEADER_VALUE=<Auth token value>
+# INTERNAL_OTEL_TRACE_LOGGING_ENABLED=1
+# INTERNAL_OTEL_TRACE_SAMPING_RATE=20 # this means 1/20 traces or 5% of traces will be sampled (sampled = recorded)
+# INTERNAL_OTEL_TRACE_INSTRUMENT_PRISMA_ENABLED=0,

.github/ISSUE_TEMPLATE/instrumentation_request.yml

@@ -0,0 +1,21 @@
+name: OpenTelemetry Auto-Instrumentation Request
+description: Suggest an SDK that you'd like to be auto-instrumented in the Run log view
+title: "auto-instrumentation: "
+labels: ["🌟 enhancement"]
+body:
+  - type: textarea
+    attributes:
+      label: What API or SDK would you to have automatic spans for?
+      description: A clear description of which API or SDK you'd like, and links to it.
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: Is there an existing OpenTelemetry auto-instrumentation package?
+      description: You can search for existing ones – https://opentelemetry.io/ecosystem/registry/?component=instrumentation&language=js
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: Additional information
+      description: Add any other information related to the feature here. If your feature request is related to any issues or discussions, link them here.

.github/workflows/e2e.yml

@@ -4,31 +4,36 @@ on:
 jobs:
   e2e:
     name: "🧪 E2E Tests"
-    runs-on: buildjet-4vcpu-ubuntu-2204
+    runs-on: buildjet-16vcpu-ubuntu-2204
     steps:
       - name: 🐳 Login to Docker Hub
-        if: github.event_name == 'push'
         uses: docker/login-action@v2
         with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_TOKEN }}
+          username: ${{ secrets.DOCKERHUB_USERNAME || vars.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN || vars.DOCKERHUB_TOKEN }}
 
       - name: ⬇️ Checkout repo
         uses: actions/checkout@v3
         with:
           fetch-depth: 0
+          submodules: recursive
 
       - name: ⎔ Setup pnpm
         uses: pnpm/action-setup@v2.2.4
         with:
-          version: 7.18
+          version: 8.15.5
 
       - name: ⎔ Setup node
         uses: buildjet/setup-node@v3
         with:
-          node-version: 18
+          node-version: 20.11.1
           cache: "pnpm"
 
+      - name: Install Protoc
+        uses: arduino/setup-protoc@v3
+        with:
+          repo-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: 📥 Download deps
         run: pnpm install --frozen-lockfile
 
@@ -43,7 +48,10 @@ jobs:
 
           # Build packages
           pnpm run build --filter @references/nextjs-test^...
+          cd apps/webapp && pnpm run build:server
+          cd ../..
           pnpm --filter @trigger.dev/database generate
+          pnpm --filter @trigger.dev/otlp-importer generate
 
           # Move trigger-cli bin to correct place
           pnpm install --frozen-lockfile

.github/workflows/pr_checks.yml

@@ -27,6 +27,6 @@ jobs:
     uses: ./.github/workflows/unit-tests.yml
     secrets: inherit
 
-  e2e:
-    uses: ./.github/workflows/e2e.yml
-    secrets: inherit
+  # e2e:
+  #   uses: ./.github/workflows/e2e.yml
+  #   secrets: inherit

.github/workflows/publish-docker.yml

@@ -3,23 +3,20 @@ on:
   workflow_call:
 jobs:
   publish:
-    strategy:
-      fail-fast: true # when a job fails, all remaining ones will be cancelled
-      matrix:
-        runs-on: [buildjet-4vcpu-ubuntu-2204, buildjet-4vcpu-ubuntu-2204-arm]
-    name: ${{matrix.runs-on}}
-    runs-on: ${{matrix.runs-on}}
+    runs-on: ubuntu-latest
+    env:
+      PRISMA_ENGINES_CHECKSUM_IGNORE_MISSING: 1
     outputs:
       version: ${{ steps.get_version.outputs.version }}
       short_sha: ${{ steps.get_commit.outputs.sha_short }}
     steps:
-      - name: 🐳 Login to Docker Hub
-        uses: docker/login-action@v2
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Setup Depot CLI
+        uses: depot/setup-action@v1
+
       - name: ⬇️ Checkout repo
         uses: actions/checkout@v3
+        with:
+          submodules: recursive
 
       - name: 🆚 Get the version
         id: get_version
@@ -41,19 +38,12 @@ jobs:
             echo "Invalid reference: ${GITHUB_REF}"
             exit 1
           fi
-          if [[ ${{matrix.runs-on}} == *-arm ]]; then
-            IMAGE_TAG="${IMAGE_TAG}-arm"
-          fi
           echo "::set-output name=version::${IMAGE_TAG}"
       - name: 🔢 Get the commit hash
         id: get_commit
         run: |
           echo ::set-output name=sha_short::$(echo ${{ github.sha }} | cut -c1-7)
 
-      - name: 🐳 Build Docker Image
-        run: |
-          docker build -t release_build_image -f ./docker/Dockerfile .
-
       - name: 🐙 Login to GitHub Container Registry
         uses: docker/login-action@v2
         with:
@@ -61,24 +51,11 @@ jobs:
           username: ${{ github.repository_owner }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
-      - name: 🐙 Push to GitHub Container Registry
-        run: |
-          docker tag release_build_image $REGISTRY/$REPOSITORY:$IMAGE_TAG
-          docker push $REGISTRY/$REPOSITORY:$IMAGE_TAG
-        env:
-          REGISTRY: ghcr.io/triggerdotdev
-          REPOSITORY: trigger.dev
-          IMAGE_TAG: ${{ steps.get_version.outputs.version }}
-
-      - name: 🐙 Push 'latest' to GitHub Container Registry
-        if: startsWith(github.ref, 'refs/tags/v.docker')
-        run: |
-          LATEST=latest
-          if [[ ${{matrix.runs-on}} == *-arm ]]; then
-            LATEST="${LATEST}-arm"
-          fi
-          docker tag release_build_image $REGISTRY/$REPOSITORY:$LATEST
-          docker push $REGISTRY/$REPOSITORY:$LATEST
-        env:
-          REGISTRY: ghcr.io/triggerdotdev
-          REPOSITORY: trigger.dev
+      - name: 🐳 Build image and push to GitHub Container Registry
+        uses: depot/build-push-action@v1
+        with:
+          file: ./docker/Dockerfile
+          platforms: linux/amd64,linux/arm64
+          tags: |
+            ghcr.io/triggerdotdev/trigger.dev:${{ steps.get_version.outputs.version }}
+          push: true

.github/workflows/publish-infra.yml

@@ -0,0 +1,94 @@
+name: "🚢 Publish Infra Images"
+
+on:
+  push:
+    tags:
+      - "infra-dev-*"
+      - "infra-test-*"
+      - "infra-prod-*"
+    paths:
+      - ".github/workflows/publish.yml"
+      - "packages/**"
+      - "!packages/**/*.md"
+      - "!packages/**/*.eslintrc"
+      - "apps/**"
+      - "!apps/**/*.md"
+      - "!apps/**/*.eslintrc"
+      - "integrations/**"
+      - "!integrations/**/*.md"
+      - "!integrations/**/*.eslintrc"
+      - "pnpm-lock.yaml"
+      - "pnpm-workspace.yaml"
+      - "turbo.json"
+      - "docker/Dockerfile"
+      - "docker/scripts/**"
+      - "tests/**"
+
+permissions:
+  id-token: write
+  packages: write
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+
+env:
+  AWS_REGION: us-east-1
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        package: [coordinator, kubernetes-provider]
+    runs-on: buildjet-16vcpu-ubuntu-2204
+    env:
+      DOCKER_BUILDKIT: "1"
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Generate image reference
+        id: prep
+        # WARNING: This step expects the workflow to have been triggered by a specific tag format of: infra-${env}-*
+        run: |
+          env=$(echo ${{ github.ref_name }} | cut -d- -f2)
+          sha=${GITHUB_SHA::7}
+          ts=$(date +%s)
+          if [[ "${{ matrix.package }}" == *-provider ]]; then
+            provider_type=$(echo ${{ matrix.package }} | cut -d- -f1)
+            repository=provider/${provider_type}
+          else
+            repository=${{ matrix.package }}
+          fi
+          echo "IMAGE_TAG=${env}-${sha}-${ts}" >> "$GITHUB_OUTPUT"
+          echo "REPOSITORY=${repository}" >> "$GITHUB_OUTPUT"
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      # ..to avoid rate limits when pulling images
+      - name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: 🚢 Build Container Image
+        run: |
+          docker build -t infra_image -f ./apps/${{ matrix.package }}/Containerfile .
+
+      # ..to push image
+      - name: 🐙 Login to GitHub Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: 🐙 Push to GitHub Container Registry
+        run: |
+          docker tag infra_image $REGISTRY/$REPOSITORY:$IMAGE_TAG
+          docker push $REGISTRY/$REPOSITORY:$IMAGE_TAG
+        env:
+          REGISTRY: ghcr.io/triggerdotdev
+          REPOSITORY: ${{ steps.prep.outputs.REPOSITORY }}
+          IMAGE_TAG: ${{ steps.prep.outputs.IMAGE_TAG }}

.github/workflows/publish.yml

@@ -9,6 +9,10 @@ on:
       - "build-*"
     paths:
       - ".github/workflows/publish.yml"
+      - ".github/workflows/typecheck.yml"
+      - ".github/workflows/unit-tests.yml"
+      - ".github/workflows/e2e.yml"
+      - ".github/workflows/publish-docker.yml"
       - "packages/**"
       - "!packages/**/*.md"
       - "!packages/**/*.eslintrc"
@@ -45,11 +49,11 @@ jobs:
     uses: ./.github/workflows/unit-tests.yml
     secrets: inherit
 
-  e2e:
-    uses: ./.github/workflows/e2e.yml
-    secrets: inherit
+  # e2e:
+  #   uses: ./.github/workflows/e2e.yml
+  #   secrets: inherit
 
   publish:
-    needs: [typecheck, units, e2e]
+    needs: [typecheck, units]
     uses: ./.github/workflows/publish-docker.yml
     secrets: inherit

.github/workflows/release.yml

@@ -8,12 +8,11 @@ on:
       - "**.md"
       - ".github/CODEOWNERS"
       - ".github/ISSUE_TEMPLATE/**"
-    
 
 jobs:
   release:
     name: 🦋 Changesets Release
-    runs-on: buildjet-4vcpu-ubuntu-2204
+    runs-on: buildjet-8vcpu-ubuntu-2204
     if: |
       github.repository == 'triggerdotdev/trigger.dev'
     outputs:
@@ -31,14 +30,19 @@ jobs:
       - name: ⎔ Setup pnpm
         uses: pnpm/action-setup@v2.2.4
         with:
-          version: 7.18
+          version: 8.15.5
 
       - name: ⎔ Setup node
         uses: buildjet/setup-node@v3
         with:
-          node-version: 18
+          node-version: 20.11.1
           cache: "pnpm"
 
+      - name: Install Protoc
+        uses: arduino/setup-protoc@v3
+        with:
+          repo-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: 📥 Download deps
         run: pnpm install --frozen-lockfile
 
@@ -46,7 +50,7 @@ jobs:
         run: pnpm run generate
 
       - name: 🔎 Type check
-        run: pnpm run typecheck
+        run: pnpm run typecheck --filter "@trigger.dev/*" --filter "trigger.dev"
 
       - name: 🔐 Setup npm auth
         run: |

.github/workflows/typecheck.yml

@@ -3,7 +3,7 @@ on:
   workflow_call:
 jobs:
   typecheck:
-    runs-on: buildjet-4vcpu-ubuntu-2204
+    runs-on: buildjet-8vcpu-ubuntu-2204
 
     steps:
       - name: ⬇️ Checkout repo
@@ -14,14 +14,19 @@ jobs:
       - name: ⎔ Setup pnpm
         uses: pnpm/action-setup@v2.2.4
         with:
-          version: 7.18
+          version: 8.15.5
 
       - name: ⎔ Setup node
         uses: buildjet/setup-node@v3
         with:
-          node-version: 18
+          node-version: 20.11.1
           cache: "pnpm"
 
+      - name: Install Protoc
+        uses: arduino/setup-protoc@v3
+        with:
+          repo-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: 📥 Download deps
         run: pnpm install --frozen-lockfile
 

.github/workflows/unit-tests.yml

@@ -4,7 +4,7 @@ on:
 jobs:
   unitTests:
     name: "🧪 Unit Tests"
-    runs-on: buildjet-4vcpu-ubuntu-2204
+    runs-on: buildjet-8vcpu-ubuntu-2204
     steps:
       - name: ⬇️ Checkout repo
         uses: actions/checkout@v3
@@ -14,14 +14,19 @@ jobs:
       - name: ⎔ Setup pnpm
         uses: pnpm/action-setup@v2.2.4
         with:
-          version: 7.18
+          version: 8.15.5
 
       - name: ⎔ Setup node
         uses: buildjet/setup-node@v3
         with:
-          node-version: 18
+          node-version: 20.11.1
           cache: "pnpm"
 
+      - name: Install Protoc
+        uses: arduino/setup-protoc@v3
+        with:
+          repo-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: ⎔ Setup Deno
         uses: denoland/setup-deno@v1
         with:

.gitignore

@@ -53,4 +53,5 @@ apps/**/public/build
 /playwright-report/
 /playwright/.cache/
 
-.cosine
+.cosine
+.trigger/

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "packages/otlp-importer/protos"]
+	path = packages/otlp-importer/protos
+	url = https://github.com/open-telemetry/opentelemetry-proto.git

.nvmrc

@@ -1 +1 @@
-v18.18.0
+v20.11.1

.vscode/launch.json

@@ -9,7 +9,7 @@
       "request": "launch",
       "name": "Debug WebApp",
       "command": "pnpm run dev --filter webapp",
-      "envFile": "${workspaceFolder}/apps/webapp/.env",
+      "envFile": "${workspaceFolder}/.env",
       "cwd": "${workspaceFolder}",
       "sourceMaps": true
     },
@@ -23,11 +23,35 @@
     {
       "type": "node-terminal",
       "request": "launch",
-      "name": "Debug BYO Auth",
+      "name": "Debug v2 job catalog",
       "command": "pnpm run byo-auth",
       "envFile": "${workspaceFolder}/references/job-catalog/.env",
       "cwd": "${workspaceFolder}/references/job-catalog",
       "sourceMaps": true
+    },
+    {
+      "type": "node-terminal",
+      "request": "launch",
+      "name": "Debug V3 Dev CLI",
+      "command": "pnpm exec triggerdev dev --log-level debug",
+      "cwd": "${workspaceFolder}/references/v3-catalog",
+      "sourceMaps": true
+    },
+    {
+      "type": "node-terminal",
+      "request": "launch",
+      "name": "Debug V3 Deploy CLI",
+      "command": "pnpm exec triggerdev deploy",
+      "cwd": "${workspaceFolder}/references/v3-catalog",
+      "sourceMaps": true
+    },
+    {
+      "type": "node",
+      "request": "attach",
+      "name": "Attach to Trigger.dev CLI (v3)",
+      "port": 9229,
+      "restart": true,
+      "skipFiles": ["<node_internals>/**"]
     }
   ]
 }

CHANGESETS.md

@@ -47,7 +47,7 @@ pnpm exec changeset version --snapshot prerelease
 3. Build the packages:
 
 ```sh
-pnpm run build --filter "@trigger.dev/*"
+pnpm run build --filter "@trigger.dev/*" --filter "trigger.dev"
 ```
 
 4. Publish the snapshot (replace "dev" with your tag)

README.md

@@ -7,7 +7,7 @@
   
 ### The open source background jobs framework
 
-[Discord](https://discord.gg/JtBAxBr2m3) | [Website](https://trigger.dev) | [Issues](https://github.com/triggerdotdev/trigger.dev/issues) | [Docs](https://trigger.dev/docs)
+[Discord](https://trigger.dev/discord) | [Website](https://trigger.dev) | [Issues](https://github.com/triggerdotdev/trigger.dev/issues) | [Docs](https://trigger.dev/docs)
 
 [![Twitter](https://img.shields.io/twitter/url/https/twitter.com/triggerdotdev.svg?style=social&label=Follow%20%40trigger.dev)](https://twitter.com/triggerdotdev)
 [![GitHub Repo stars](https://img.shields.io/github/stars/triggerdotdev/trigger.dev?style=social)](https://github.com/triggerdotdev/trigger.dev)

apps/coordinator/.env.example

@@ -0,0 +1,4 @@
+HTTP_SERVER_PORT=8020
+PLATFORM_ENABLED=true
+PLATFORM_WS_PORT=3030
+SECURE_CONNECTION=false

apps/coordinator/.gitignore

@@ -0,0 +1,3 @@
+dist/
+node_modules/
+.env

apps/coordinator/Containerfile

@@ -0,0 +1,60 @@
+# syntax=docker/dockerfile:labs
+
+FROM node:18-bullseye-slim@sha256:a4edd54dcfdcacc8a4100fee71498e8671d99556a1acf5614539214a70092426 AS node-18
+
+WORKDIR /app
+
+FROM node-18 AS pruner
+
+COPY --chown=node:node . .
+RUN npx -q turbo@1.10.9 prune --scope=coordinator --docker
+RUN find . -name "node_modules" -type d -prune -exec rm -rf '{}' +
+
+FROM node-18 AS base
+
+RUN apt-get update \
+  && apt-get install -y buildah ca-certificates dumb-init \
+  && rm -rf /var/lib/apt/lists/*
+
+COPY --chown=node:node .gitignore .gitignore
+COPY --from=pruner --chown=node:node /app/out/json/ .
+COPY --from=pruner --chown=node:node /app/out/pnpm-lock.yaml ./pnpm-lock.yaml
+COPY --from=pruner --chown=node:node /app/out/pnpm-workspace.yaml ./pnpm-workspace.yaml
+
+FROM base AS dev-deps
+RUN corepack enable
+ENV NODE_ENV development
+
+RUN --mount=type=cache,id=pnpm,target=/root/.local/share/pnpm/store pnpm fetch --no-frozen-lockfile
+RUN --mount=type=cache,id=pnpm,target=/root/.local/share/pnpm/store pnpm install --ignore-scripts --no-frozen-lockfile
+
+FROM base AS builder
+RUN corepack enable
+
+COPY --from=pruner --chown=node:node /app/out/full/ .
+COPY --from=dev-deps --chown=node:node /app/ .
+COPY --chown=node:node turbo.json turbo.json
+
+RUN pnpm run -r --filter coordinator build:bundle
+
+FROM alpine AS cri-tools
+
+WORKDIR /cri-tools
+
+ARG CRICTL_VERSION=v1.29.0
+ARG CRICTL_CHECKSUM=sha256:d16a1ffb3938f5a19d5c8f45d363bd091ef89c0bc4d44ad16b933eede32fdcbb
+ADD --checksum=${CRICTL_CHECKSUM} \
+  https://github.com/kubernetes-sigs/cri-tools/releases/download/${CRICTL_VERSION}/crictl-${CRICTL_VERSION}-linux-amd64.tar.gz .
+RUN tar zxvf crictl-${CRICTL_VERSION}-linux-amd64.tar.gz
+
+FROM base AS runner
+
+RUN corepack enable
+ENV NODE_ENV production
+
+COPY --from=cri-tools --chown=node:node /cri-tools/crictl /usr/local/bin
+COPY --from=builder --chown=node:node /app/apps/coordinator/dist/index.mjs ./index.mjs
+
+EXPOSE 8000
+
+CMD [ "/usr/bin/dumb-init", "--", "/usr/local/bin/node", "./index.mjs" ]

apps/coordinator/README.md

@@ -0,0 +1,3 @@
+# Coordinator
+
+Sits between the platform and tasks. Facilitates communication and checkpointing, amongst other things.

apps/coordinator/package.json

@@ -0,0 +1,34 @@
+{
+  "name": "coordinator",
+  "private": true,
+  "version": "0.0.1",
+  "description": "",
+  "main": "dist/index.cjs",
+  "scripts": {
+    "build": "npm run build:bundle",
+    "build:bundle": "esbuild src/index.ts --bundle --outfile=dist/index.mjs --platform=node --format=esm --target=esnext --banner:js=\"import { createRequire } from 'module';const require = createRequire(import.meta.url);\"",
+    "build:image": "docker build -f Containerfile . -t coordinator",
+    "dev": "tsx --no-warnings=ExperimentalWarning --require dotenv/config --watch src/index.ts",
+    "start": "tsx src/index.ts",
+    "typecheck": "tsc --noEmit"
+  },
+  "keywords": [],
+  "author": "",
+  "license": "MIT",
+  "dependencies": {
+    "@trigger.dev/core": "workspace:*",
+    "@trigger.dev/core-apps": "workspace:*",
+    "execa": "^8.0.1",
+    "nanoid": "^5.0.6",
+    "prom-client": "^15.1.0",
+    "socket.io": "4.7.4",
+    "socket.io-client": "4.7.4"
+  },
+  "devDependencies": {
+    "@types/node": "^18",
+    "dotenv": "^16.4.2",
+    "esbuild": "^0.19.11",
+    "tsx": "^4.7.0",
+    "typescript": "^5.3.3"
+  }
+}

apps/coordinator/src/index.ts

@@ -0,0 +1,995 @@
+import { createServer } from "node:http";
+import { $ } from "execa";
+import { nanoid } from "nanoid";
+import { Server } from "socket.io";
+import {
+  CoordinatorToPlatformMessages,
+  CoordinatorToProdWorkerMessages,
+  PlatformToCoordinatorMessages,
+  ProdWorkerSocketData,
+  ProdWorkerToCoordinatorMessages,
+} from "@trigger.dev/core/v3";
+import { ZodNamespace } from "@trigger.dev/core/v3/zodNamespace";
+import { ZodSocketConnection } from "@trigger.dev/core/v3/zodSocket";
+import { HttpReply, getTextBody, SimpleLogger } from "@trigger.dev/core-apps";
+
+import { collectDefaultMetrics, register, Gauge } from "prom-client";
+collectDefaultMetrics();
+
+const HTTP_SERVER_PORT = Number(process.env.HTTP_SERVER_PORT || 8020);
+const NODE_NAME = process.env.NODE_NAME || "coordinator";
+const DEFAULT_RETRY_DELAY_THRESHOLD_IN_MS = 30_000;
+
+const REGISTRY_HOST = process.env.REGISTRY_HOST || "localhost:5000";
+const CHECKPOINT_PATH = process.env.CHECKPOINT_PATH || "/checkpoints";
+const REGISTRY_TLS_VERIFY = process.env.REGISTRY_TLS_VERIFY === "false" ? "false" : "true";
+
+const PLATFORM_ENABLED = ["1", "true"].includes(process.env.PLATFORM_ENABLED ?? "true");
+const PLATFORM_HOST = process.env.PLATFORM_HOST || "127.0.0.1";
+const PLATFORM_WS_PORT = process.env.PLATFORM_WS_PORT || 3030;
+const PLATFORM_SECRET = process.env.PLATFORM_SECRET || "coordinator-secret";
+const SECURE_CONNECTION = ["1", "true"].includes(process.env.SECURE_CONNECTION ?? "false");
+
+const logger = new SimpleLogger(`[${NODE_NAME}]`);
+
+type CheckpointerInitializeReturn = {
+  canCheckpoint: boolean;
+  willSimulate: boolean;
+};
+
+type CheckpointAndPushOptions = {
+  runId: string;
+  leaveRunning?: boolean;
+  projectRef: string;
+  deploymentVersion: string;
+};
+
+type CheckpointData = {
+  location: string;
+  docker: boolean;
+};
+
+class Checkpointer {
+  #initialized = false;
+  #canCheckpoint = false;
+  #dockerMode = !process.env.KUBERNETES_PORT;
+
+  #logger = new SimpleLogger("[checkptr]");
+  #abortControllers = new Map<string, AbortController>();
+
+  constructor(private opts = { forceSimulate: false }) {}
+
+  async initialize(): Promise<CheckpointerInitializeReturn> {
+    if (this.#initialized) {
+      return this.#getInitializeReturn();
+    }
+
+    this.#logger.log(`${this.#dockerMode ? "Docker" : "Kubernetes"} mode`);
+
+    if (this.#dockerMode) {
+      try {
+        await $`criu --version`;
+      } catch (error) {
+        this.#logger.error("No checkpoint support: Missing CRIU binary");
+        this.#logger.error("Will simulate instead");
+        this.#canCheckpoint = false;
+        this.#initialized = true;
+
+        return this.#getInitializeReturn();
+      }
+
+      try {
+        await $`docker checkpoint`;
+      } catch (error) {
+        this.#logger.error(
+          "No checkpoint support: Docker needs to have experimental features enabled"
+        );
+        this.#logger.error("Will simulate instead");
+        this.#canCheckpoint = false;
+        this.#initialized = true;
+
+        return this.#getInitializeReturn();
+      }
+    } else {
+      try {
+        await $`buildah login --get-login ${REGISTRY_HOST}`;
+      } catch (error) {
+        this.#logger.error(`No checkpoint support: Not logged in to registry ${REGISTRY_HOST}`);
+        this.#canCheckpoint = false;
+        this.#initialized = true;
+
+        return this.#getInitializeReturn();
+      }
+    }
+
+    this.#logger.log(
+      `Full checkpoint support${
+        this.#dockerMode && this.opts.forceSimulate ? " with forced simulation enabled." : "!"
+      }`
+    );
+
+    this.#initialized = true;
+    this.#canCheckpoint = true;
+
+    return this.#getInitializeReturn();
+  }
+
+  #getInitializeReturn(): CheckpointerInitializeReturn {
+    return {
+      canCheckpoint: this.#canCheckpoint,
+      willSimulate: this.#dockerMode && (!this.#canCheckpoint || this.opts.forceSimulate),
+    };
+  }
+
+  #getImageRef(projectRef: string, deploymentVersion: string, shortCode: string) {
+    return `${REGISTRY_HOST}/trigger/${projectRef}:${deploymentVersion}.prod-${shortCode}`;
+  }
+
+  #getExportLocation(projectRef: string, deploymentVersion: string, shortCode: string) {
+    const basename = `${projectRef}-${deploymentVersion}-${shortCode}`;
+
+    if (this.#dockerMode) {
+      return basename;
+    } else {
+      return `${CHECKPOINT_PATH}/${basename}.tar`;
+    }
+  }
+
+  async checkpointAndPush(opts: CheckpointAndPushOptions): Promise<CheckpointData | undefined> {
+    const start = performance.now();
+    logger.log(`checkpointAndPush() start`, { start, opts });
+
+    const result = await this.#checkpointAndPush(opts);
+
+    const end = performance.now();
+    logger.log(`checkpointAndPush() end`, {
+      start,
+      end,
+      diff: end - start,
+      opts,
+      success: !!result,
+    });
+
+    return result;
+  }
+
+  isCheckpointing(runId: string) {
+    return this.#abortControllers.has(runId);
+  }
+
+  cancelCheckpoint(runId: string): boolean {
+    const controller = this.#abortControllers.get(runId);
+
+    if (!controller) {
+      logger.debug("Nothing to cancel", { runId });
+      return false;
+    }
+
+    controller.abort("cancelCheckpointing()");
+    this.#abortControllers.delete(runId);
+
+    return true;
+  }
+
+  async #checkpointAndPush({
+    runId,
+    leaveRunning = true, // This mirrors kubernetes behaviour more accurately
+    projectRef,
+    deploymentVersion,
+  }: CheckpointAndPushOptions): Promise<CheckpointData | undefined> {
+    await this.initialize();
+
+    if (!this.#dockerMode && !this.#canCheckpoint) {
+      this.#logger.error("No checkpoint support. Simulation requires docker.");
+      return;
+    }
+
+    if (this.#abortControllers.has(runId)) {
+      logger.error("Checkpoint procedure already in progress", {
+        options: {
+          runId,
+          leaveRunning,
+          projectRef,
+          deploymentVersion,
+        },
+      });
+      return;
+    }
+
+    const controller = new AbortController();
+    this.#abortControllers.set(runId, controller);
+
+    const $$ = $({ signal: controller.signal });
+
+    try {
+      const shortCode = nanoid(8);
+      const imageRef = this.#getImageRef(projectRef, deploymentVersion, shortCode);
+      const exportLocation = this.#getExportLocation(projectRef, deploymentVersion, shortCode);
+
+      this.#logger.log("Checkpointing:", {
+        options: {
+          runId,
+          leaveRunning,
+          projectRef,
+          deploymentVersion,
+        },
+      });
+
+      const containterName = this.#getRunContainerName(runId);
+
+      // Create checkpoint (docker)
+      if (this.#dockerMode) {
+        try {
+          if (this.opts.forceSimulate || !this.#canCheckpoint) {
+            this.#logger.log("Simulating checkpoint");
+            this.#logger.debug(await $$`docker pause ${containterName}`);
+          } else {
+            if (leaveRunning) {
+              this.#logger.debug(
+                await $$`docker checkpoint create --leave-running ${containterName} ${exportLocation}`
+              );
+            } else {
+              this.#logger.debug(
+                await $$`docker checkpoint create ${containterName} ${exportLocation}`
+              );
+            }
+          }
+        } catch (error: any) {
+          this.#logger.error(error.stderr);
+          return;
+        }
+
+        this.#logger.log("checkpoint created:", {
+          runId,
+          location: exportLocation,
+        });
+
+        return {
+          location: exportLocation,
+          docker: true,
+        };
+      }
+
+      // Create checkpoint (CRI)
+      if (!this.#canCheckpoint) {
+        throw new Error("No checkpoint support in kubernetes mode.");
+      }
+
+      const containerId = this.#logger.debug(
+        // @ts-expect-error
+        await $$`crictl ps`
+          .pipeStdout($$({ stdin: "pipe" })`grep ${containterName}`)
+          .pipeStdout($$({ stdin: "pipe" })`cut -f1 ${"-d "}`)
+      );
+
+      if (!containerId.stdout) {
+        throw new Error("could not find container id");
+      }
+
+      this.#logger.debug(await $$`crictl checkpoint --export=${exportLocation} ${containerId}`);
+
+      // Create image from checkpoint
+      const container = this.#logger.debug(await $$`buildah from scratch`);
+      this.#logger.debug(await $$`buildah add ${container} ${exportLocation} /`);
+      this.#logger.debug(
+        await $$`buildah config --annotation=io.kubernetes.cri-o.annotations.checkpoint.name=counter ${container}`
+      );
+      this.#logger.debug(await $$`buildah commit ${container} ${imageRef}`);
+      this.#logger.debug(await $$`buildah rm ${container}`);
+
+      // Push checkpoint image
+      this.#logger.debug(await $$`buildah push --tls-verify=${REGISTRY_TLS_VERIFY} ${imageRef}`);
+
+      this.#logger.log("Checkpointed and pushed image to:", { location: imageRef });
+
+      try {
+        await $$`rm ${exportLocation}`;
+        this.#logger.log("Deleted checkpoint archive", { exportLocation });
+
+        // Disabled for now as this will increase restore time by having to pull the image again
+        // await $`buildah rmi ${imageRef}`;
+        // this.#logger.log("Deleted checkpoint image", { imageRef });
+      } catch (error) {
+        this.#logger.error("Failed during checkpoint cleanup", { exportLocation });
+        this.#logger.debug(error);
+      }
+
+      return {
+        location: imageRef,
+        docker: false,
+      };
+    } catch (error) {
+      this.#logger.error("checkpoint failed", {
+        options: {
+          runId,
+          leaveRunning,
+          projectRef,
+          deploymentVersion,
+        },
+        error,
+      });
+      return;
+    } finally {
+      this.#abortControllers.delete(runId);
+    }
+  }
+
+  #getRunContainerName(suffix: string) {
+    return `task-run-${suffix}`;
+  }
+}
+
+class TaskCoordinator {
+  #httpServer: ReturnType<typeof createServer>;
+  #checkpointer = new Checkpointer({ forceSimulate: true });
+
+  #prodWorkerNamespace: ZodNamespace<
+    typeof ProdWorkerToCoordinatorMessages,
+    typeof CoordinatorToProdWorkerMessages,
+    typeof ProdWorkerSocketData
+  >;
+  #platformSocket?: ZodSocketConnection<
+    typeof CoordinatorToPlatformMessages,
+    typeof PlatformToCoordinatorMessages
+  >;
+
+  #checkpointableTasks = new Map<
+    string,
+    { resolve: (value: void) => void; reject: (err?: any) => void }
+  >();
+
+  #delayThresholdInMs: number;
+
+  constructor(
+    private port: number,
+    private host = "0.0.0.0"
+  ) {
+    this.#httpServer = this.#createHttpServer();
+    this.#checkpointer.initialize();
+    this.#delayThresholdInMs = this.#getDelayThreshold();
+
+    if (process.env.DELAY_THRESHOLD_IN_MS) {
+      this.#delayThresholdInMs = this.#getDelayThreshold();
+    }
+
+    const io = new Server(this.#httpServer);
+    this.#prodWorkerNamespace = this.#createProdWorkerNamespace(io);
+
+    this.#platformSocket = this.#createPlatformSocket();
+
+    const connectedTasksTotal = new Gauge({
+      name: "daemon_connected_tasks_total", // don't change this without updating dashboard config
+      help: "The number of tasks currently connected.",
+      collect: () => {
+        connectedTasksTotal.set(this.#prodWorkerNamespace.namespace.sockets.size);
+      },
+    });
+    register.registerMetric(connectedTasksTotal);
+  }
+
+  #getDelayThreshold() {
+    if (!process.env.RETRY_DELAY_THRESHOLD_IN_MS) {
+      return DEFAULT_RETRY_DELAY_THRESHOLD_IN_MS;
+    }
+
+    const threshold = parseInt(process.env.RETRY_DELAY_THRESHOLD_IN_MS);
+
+    if (isNaN(threshold)) {
+      logger.log(
+        "RETRY_DELAY_THRESHOLD_IN_MS parses as NaN, must supply integer. Will use default instead.",
+        {
+          RETRY_DELAY_THRESHOLD_IN_MS: process.env.RETRY_DELAY_THRESHOLD_IN_MS,
+          DEFAULT_DELAY_THRESHOLD_IN_MS: DEFAULT_RETRY_DELAY_THRESHOLD_IN_MS,
+        }
+      );
+      return DEFAULT_RETRY_DELAY_THRESHOLD_IN_MS;
+    }
+
+    return threshold;
+  }
+
+  #createPlatformSocket() {
+    if (!PLATFORM_ENABLED) {
+      console.log("INFO: platform connection disabled");
+      return;
+    }
+
+    const platformConnection = new ZodSocketConnection({
+      namespace: "coordinator",
+      host: PLATFORM_HOST,
+      port: Number(PLATFORM_WS_PORT),
+      secure: SECURE_CONNECTION,
+      clientMessages: CoordinatorToPlatformMessages,
+      serverMessages: PlatformToCoordinatorMessages,
+      authToken: PLATFORM_SECRET,
+      handlers: {
+        RESUME_AFTER_DEPENDENCY: async (message) => {
+          const taskSocket = await this.#getAttemptSocket(message.attemptFriendlyId);
+
+          if (!taskSocket) {
+            logger.log("Socket for attempt not found", {
+              attemptFriendlyId: message.attemptFriendlyId,
+            });
+            return;
+          }
+
+          // In case the task resumed faster than we could checkpoint
+          this.#cancelCheckpoint(message.runId);
+
+          taskSocket.emit("RESUME_AFTER_DEPENDENCY", message);
+        },
+        RESUME_AFTER_DURATION: async (message) => {
+          const taskSocket = await this.#getAttemptSocket(message.attemptFriendlyId);
+
+          if (!taskSocket) {
+            logger.log("Socket for attempt not found", {
+              attemptFriendlyId: message.attemptFriendlyId,
+            });
+            return;
+          }
+
+          taskSocket.emit("RESUME_AFTER_DURATION", message);
+        },
+        REQUEST_ATTEMPT_CANCELLATION: async (message) => {
+          const taskSocket = await this.#getAttemptSocket(message.attemptFriendlyId);
+
+          if (!taskSocket) {
+            logger.log("Socket for attempt not found", {
+              attemptFriendlyId: message.attemptFriendlyId,
+            });
+            return;
+          }
+
+          taskSocket.emit("REQUEST_ATTEMPT_CANCELLATION", message);
+        },
+        READY_FOR_RETRY: async (message) => {
+          const taskSocket = await this.#getRunSocket(message.runId);
+
+          if (!taskSocket) {
+            logger.log("Socket for attempt not found", {
+              runId: message.runId,
+            });
+            return;
+          }
+
+          taskSocket.emit("READY_FOR_RETRY", message);
+        },
+      },
+    });
+
+    return platformConnection;
+  }
+
+  async #getRunSocket(runId: string) {
+    const sockets = await this.#prodWorkerNamespace.fetchSockets();
+
+    for (const socket of sockets) {
+      if (socket.data.runId === runId) {
+        return socket;
+      }
+    }
+  }
+
+  async #getAttemptSocket(attemptFriendlyId: string) {
+    const sockets = await this.#prodWorkerNamespace.fetchSockets();
+
+    for (const socket of sockets) {
+      if (socket.data.attemptFriendlyId === attemptFriendlyId) {
+        return socket;
+      }
+    }
+  }
+
+  #createProdWorkerNamespace(io: Server) {
+    const provider = new ZodNamespace({
+      io,
+      name: "prod-worker",
+      clientMessages: ProdWorkerToCoordinatorMessages,
+      serverMessages: CoordinatorToProdWorkerMessages,
+      socketData: ProdWorkerSocketData,
+      postAuth: async (socket, next, logger) => {
+        function setSocketDataFromHeader(
+          dataKey: keyof typeof socket.data,
+          headerName: string,
+          required: boolean = true
+        ) {
+          const value = socket.handshake.headers[headerName];
+
+          if (value) {
+            socket.data[dataKey] = Array.isArray(value) ? value[0] : value;
+            return;
+          }
+
+          if (required) {
+            logger.error("missing required header", { headerName });
+            throw new Error("missing header");
+          }
+        }
+
+        try {
+          setSocketDataFromHeader("podName", "x-pod-name");
+          setSocketDataFromHeader("contentHash", "x-trigger-content-hash");
+          setSocketDataFromHeader("projectRef", "x-trigger-project-ref");
+          setSocketDataFromHeader("runId", "x-trigger-run-id");
+          setSocketDataFromHeader("attemptFriendlyId", "x-trigger-attempt-friendly-id", false);
+          setSocketDataFromHeader("envId", "x-trigger-env-id");
+          setSocketDataFromHeader("deploymentId", "x-trigger-deployment-id");
+          setSocketDataFromHeader("deploymentVersion", "x-trigger-deployment-version");
+        } catch (error) {
+          logger.error("setSocketDataFromHeader error", { error });
+          socket.disconnect(true);
+          return;
+        }
+
+        logger.debug("success", socket.data);
+
+        next();
+      },
+      onConnection: async (socket, handler, sender) => {
+        const logger = new SimpleLogger(`[prod-worker][${socket.id}]`);
+
+        const checkpointInProgress = () => {
+          return this.#checkpointableTasks.has(socket.data.runId);
+        };
+
+        const readyToCheckpoint = async (): Promise<
+          { success: true } | { success: false; reason?: string }
+        > => {
+          if (checkpointInProgress()) {
+            return {
+              success: false,
+              reason: "checkpoint in progress",
+            };
+          }
+
+          const isCheckpointable = new Promise((resolve, reject) => {
+            // We set a reasonable timeout to prevent waiting forever
+            // TODO: We may also want to cancel the task as it's unlikely to recover
+            setTimeout(() => reject("timeout"), 10_000);
+
+            this.#checkpointableTasks.set(socket.data.runId, { resolve, reject });
+          });
+
+          try {
+            await isCheckpointable;
+            this.#checkpointableTasks.delete(socket.data.runId);
+
+            return {
+              success: true,
+            };
+          } catch (error) {
+            logger.error("Error while waiting for checkpointable state", { error });
+
+            return {
+              success: false,
+              reason: typeof error === "string" ? error : "unknown",
+            };
+          }
+        };
+
+        this.#platformSocket?.send("LOG", {
+          metadata: socket.data,
+          text: "connected",
+        });
+
+        socket.on("LOG", (message, callback) => {
+          logger.log("[LOG]", message.text);
+
+          callback();
+
+          this.#platformSocket?.send("LOG", {
+            version: "v1",
+            metadata: socket.data,
+            text: message.text,
+          });
+        });
+
+        socket.on("READY_FOR_EXECUTION", async (message) => {
+          logger.log("[READY_FOR_EXECUTION]", message);
+
+          try {
+            const executionAck = await this.#platformSocket?.sendWithAck(
+              "READY_FOR_EXECUTION",
+              message
+            );
+
+            if (!executionAck) {
+              logger.error("no execution ack", { runId: socket.data.runId });
+
+              socket.emit("REQUEST_EXIT", {
+                version: "v1",
+              });
+
+              return;
+            }
+
+            if (!executionAck.success) {
+              logger.error("failed to get execution payload", { runId: socket.data.runId });
+
+              socket.emit("REQUEST_EXIT", {
+                version: "v1",
+              });
+
+              return;
+            }
+
+            socket.emit("EXECUTE_TASK_RUN", {
+              version: "v1",
+              executionPayload: executionAck.payload,
+            });
+
+            socket.data.attemptFriendlyId = executionAck.payload.execution.attempt.id;
+          } catch (error) {
+            logger.error("Error", { error });
+          }
+        });
+
+        socket.on("READY_FOR_RESUME", async (message) => {
+          logger.log("[READY_FOR_RESUME]", message);
+
+          socket.data.attemptFriendlyId = message.attemptFriendlyId;
+          this.#platformSocket?.send("READY_FOR_RESUME", message);
+        });
+
+        socket.on("TASK_RUN_COMPLETED", async ({ completion, execution }, callback) => {
+          logger.log("completed task", { completionId: completion.id });
+
+          const completeWithoutCheckpoint = (shouldExit: boolean) => {
+            this.#platformSocket?.send("TASK_RUN_COMPLETED", {
+              version: "v1",
+              execution,
+              completion,
+            });
+            callback({ willCheckpointAndRestore: false, shouldExit });
+          };
+
+          if (completion.ok) {
+            completeWithoutCheckpoint(true);
+            return;
+          }
+
+          if (
+            completion.error.type === "INTERNAL_ERROR" &&
+            completion.error.code === "TASK_RUN_CANCELLED"
+          ) {
+            completeWithoutCheckpoint(true);
+            return;
+          }
+
+          if (completion.retry === undefined) {
+            completeWithoutCheckpoint(true);
+            return;
+          }
+
+          if (completion.retry.delay < this.#delayThresholdInMs) {
+            completeWithoutCheckpoint(false);
+            return;
+          }
+
+          const { canCheckpoint, willSimulate } = await this.#checkpointer.initialize();
+
+          const willCheckpointAndRestore = canCheckpoint || willSimulate;
+
+          if (!willCheckpointAndRestore) {
+            completeWithoutCheckpoint(false);
+            return;
+          }
+
+          // The worker will then put itself in a checkpointable state
+          callback({ willCheckpointAndRestore: true, shouldExit: false });
+
+          const ready = await readyToCheckpoint();
+
+          if (!ready.success) {
+            logger.error("Failed to become checkpointable", {
+              runId: socket.data.runId,
+              reason: ready.reason,
+            });
+            return;
+          }
+
+          const checkpoint = await this.#checkpointer.checkpointAndPush({
+            runId: socket.data.runId,
+            projectRef: socket.data.projectRef,
+            deploymentVersion: socket.data.deploymentVersion,
+          });
+
+          if (!checkpoint) {
+            logger.error("Failed to checkpoint", { runId: socket.data.runId });
+            completeWithoutCheckpoint(false);
+            return;
+          }
+
+          this.#platformSocket?.send("TASK_RUN_COMPLETED", {
+            version: "v1",
+            execution,
+            completion,
+            checkpoint,
+          });
+
+          if (!checkpoint.docker || !willSimulate) {
+            socket.emit("REQUEST_EXIT", {
+              version: "v1",
+            });
+          }
+        });
+
+        socket.on("READY_FOR_CHECKPOINT", async (message) => {
+          logger.log("[READY_FOR_CHECKPOINT]", message);
+
+          const checkpointable = this.#checkpointableTasks.get(socket.data.runId);
+
+          if (!checkpointable) {
+            logger.error("No checkpoint scheduled", { runId: socket.data.runId });
+            return;
+          }
+
+          checkpointable.resolve();
+        });
+
+        socket.on("CANCEL_CHECKPOINT", async (message, callback) => {
+          logger.log("[CANCEL_CHECKPOINT]", message);
+
+          if (message.version === "v1") {
+            this.#cancelCheckpoint(socket.data.runId);
+            // v1 has no callback
+            return;
+          }
+
+          const checkpointCanceled = this.#cancelCheckpoint(socket.data.runId);
+
+          callback({ version: "v2", checkpointCanceled });
+        });
+
+        socket.on("WAIT_FOR_DURATION", async (message, callback) => {
+          logger.log("[WAIT_FOR_DURATION]", message);
+
+          if (checkpointInProgress()) {
+            logger.error("Checkpoint already in progress", { runId: socket.data.runId });
+            callback({ willCheckpointAndRestore: false });
+            return;
+          }
+
+          const { canCheckpoint, willSimulate } = await this.#checkpointer.initialize();
+
+          const willCheckpointAndRestore = canCheckpoint || willSimulate;
+
+          callback({ willCheckpointAndRestore });
+
+          if (!willCheckpointAndRestore) {
+            return;
+          }
+
+          const ready = await readyToCheckpoint();
+
+          if (!ready.success) {
+            logger.error("Failed to become checkpointable", {
+              runId: socket.data.runId,
+              reason: ready.reason,
+            });
+            return;
+          }
+
+          const checkpoint = await this.#checkpointer.checkpointAndPush({
+            runId: socket.data.runId,
+            projectRef: socket.data.projectRef,
+            deploymentVersion: socket.data.deploymentVersion,
+          });
+
+          if (!checkpoint) {
+            // The task container will keep running until the wait duration has elapsed
+            logger.error("Failed to checkpoint", { runId: socket.data.runId });
+            return;
+          }
+
+          if (!checkpoint.docker || !willSimulate) {
+            socket.emit("REQUEST_EXIT", {
+              version: "v1",
+            });
+          }
+
+          this.#platformSocket?.send("CHECKPOINT_CREATED", {
+            version: "v1",
+            attemptFriendlyId: message.attemptFriendlyId,
+            docker: checkpoint.docker,
+            location: checkpoint.location,
+            reason: {
+              type: "WAIT_FOR_DURATION",
+              ms: message.ms,
+              now: message.now,
+            },
+          });
+        });
+
+        socket.on("WAIT_FOR_TASK", async (message, callback) => {
+          logger.log("[WAIT_FOR_TASK]", message);
+
+          const { canCheckpoint, willSimulate } = await this.#checkpointer.initialize();
+
+          const willCheckpointAndRestore = canCheckpoint || willSimulate;
+
+          callback({ willCheckpointAndRestore });
+
+          if (!willCheckpointAndRestore) {
+            return;
+          }
+
+          const checkpoint = await this.#checkpointer.checkpointAndPush({
+            runId: socket.data.runId,
+            projectRef: socket.data.projectRef,
+            deploymentVersion: socket.data.deploymentVersion,
+          });
+
+          if (!checkpoint) {
+            logger.error("Failed to checkpoint", { runId: socket.data.runId });
+            return;
+          }
+
+          if (!checkpoint.docker || !willSimulate) {
+            socket.emit("REQUEST_EXIT", {
+              version: "v1",
+            });
+          }
+
+          this.#platformSocket?.send("CHECKPOINT_CREATED", {
+            version: "v1",
+            attemptFriendlyId: message.attemptFriendlyId,
+            docker: checkpoint.docker,
+            location: checkpoint.location,
+            reason: {
+              type: "WAIT_FOR_TASK",
+              friendlyId: message.friendlyId,
+            },
+          });
+        });
+
+        socket.on("WAIT_FOR_BATCH", async (message, callback) => {
+          logger.log("[WAIT_FOR_BATCH]", message);
+
+          const { canCheckpoint, willSimulate } = await this.#checkpointer.initialize();
+
+          const willCheckpointAndRestore = canCheckpoint || willSimulate;
+
+          callback({ willCheckpointAndRestore });
+
+          if (!willCheckpointAndRestore) {
+            return;
+          }
+
+          const checkpoint = await this.#checkpointer.checkpointAndPush({
+            runId: socket.data.runId,
+            projectRef: socket.data.projectRef,
+            deploymentVersion: socket.data.deploymentVersion,
+          });
+
+          if (!checkpoint) {
+            logger.error("Failed to checkpoint", { runId: socket.data.runId });
+            return;
+          }
+
+          if (!checkpoint.docker || !willSimulate) {
+            socket.emit("REQUEST_EXIT", {
+              version: "v1",
+            });
+          }
+
+          this.#platformSocket?.send("CHECKPOINT_CREATED", {
+            version: "v1",
+            attemptFriendlyId: message.attemptFriendlyId,
+            docker: checkpoint.docker,
+            location: checkpoint.location,
+            reason: {
+              type: "WAIT_FOR_BATCH",
+              batchFriendlyId: message.batchFriendlyId,
+              runFriendlyIds: message.runFriendlyIds,
+            },
+          });
+        });
+
+        socket.on("INDEX_TASKS", async (message, callback) => {
+          logger.log("[INDEX_TASKS]", message);
+
+          const workerAck = await this.#platformSocket?.sendWithAck("CREATE_WORKER", {
+            version: "v1",
+            projectRef: socket.data.projectRef,
+            envId: socket.data.envId,
+            deploymentId: message.deploymentId,
+            metadata: {
+              contentHash: socket.data.contentHash,
+              packageVersion: message.packageVersion,
+              tasks: message.tasks,
+            },
+          });
+
+          if (!workerAck) {
+            logger.debug("no worker ack while indexing", message);
+          }
+
+          callback({ success: !!workerAck?.success });
+        });
+
+        socket.on("INDEXING_FAILED", async (message) => {
+          logger.log("[INDEXING_FAILED]", message);
+
+          this.#platformSocket?.send("INDEXING_FAILED", {
+            version: "v1",
+            deploymentId: message.deploymentId,
+            error: message.error,
+          });
+        });
+      },
+      onDisconnect: async (socket, handler, sender, logger) => {
+        this.#platformSocket?.send("LOG", {
+          metadata: socket.data,
+          text: "disconnect",
+        });
+      },
+      handlers: {
+        TASK_HEARTBEAT: async (message) => {
+          this.#platformSocket?.send("TASK_HEARTBEAT", message);
+        },
+      },
+    });
+
+    return provider;
+  }
+
+  #cancelCheckpoint(runId: string) {
+    const checkpointWait = this.#checkpointableTasks.get(runId);
+
+    if (checkpointWait) {
+      // Stop waiting for task to reach checkpointable state
+      checkpointWait.reject("Checkpoint cancelled");
+    }
+
+    // Cancel checkpointing procedure
+    const checkpointCanceled = this.#checkpointer.cancelCheckpoint(runId);
+
+    return checkpointCanceled;
+  }
+
+  #createHttpServer() {
+    const httpServer = createServer(async (req, res) => {
+      logger.log(`[${req.method}]`, req.url);
+
+      const reply = new HttpReply(res);
+
+      switch (req.url) {
+        case "/health": {
+          return reply.text("ok");
+        }
+        case "/metrics": {
+          return reply.text(await register.metrics(), 200, register.contentType);
+        }
+        case "/whoami": {
+          return reply.text(NODE_NAME);
+        }
+        case "/checkpoint": {
+          const body = await getTextBody(req);
+          // await this.#checkpointer.checkpointAndPush(body);
+          return reply.text(`sent restore request: ${body}`);
+        }
+        default: {
+          return reply.empty(404);
+        }
+      }
+    });
+
+    httpServer.on("clientError", (err, socket) => {
+      socket.end("HTTP/1.1 400 Bad Request\r\n\r\n");
+    });
+
+    httpServer.on("listening", () => {
+      logger.log("server listening on port", HTTP_SERVER_PORT);
+    });
+
+    return httpServer;
+  }
+
+  listen() {
+    this.#httpServer.listen(this.port, this.host);
+  }
+}
+
+const coordinator = new TaskCoordinator(HTTP_SERVER_PORT);
+coordinator.listen();

apps/coordinator/tsconfig.json

@@ -0,0 +1,19 @@
+{
+  "include": ["./src/**/*.ts"],
+  "exclude": ["node_modules"],
+  "compilerOptions": {
+    "target": "es2016",
+    "module": "commonjs",
+    "esModuleInterop": true,
+    "resolveJsonModule": true,
+    "forceConsistentCasingInFileNames": true,
+    "strict": true,
+    "skipLibCheck": true,
+    "paths": {
+      "@trigger.dev/core/v3": ["../../packages/core/src/v3"],
+      "@trigger.dev/core/v3/*": ["../../packages/core/src/v3/*"],
+      "@trigger.dev/core-apps": ["../../packages/core-apps/src"],
+      "@trigger.dev/core-apps/*": ["../../packages/core-apps/src/*"]
+    }
+  }
+}

apps/docker-provider/.env.example

@@ -0,0 +1,9 @@
+HTTP_SERVER_PORT=8050
+
+PLATFORM_WS_PORT=3030
+PLATFORM_SECRET=provider-secret
+SECURE_CONNECTION=false
+
+# Use this if you are on macOS
+# COORDINATOR_HOST="host.docker.internal"
+# OTEL_EXPORTER_OTLP_ENDPOINT="http://host.docker.internal:4318"

apps/docker-provider/.gitignore

@@ -0,0 +1,3 @@
+dist/
+node_modules/
+.env

apps/docker-provider/Containerfile

@@ -0,0 +1,16 @@
+# syntax=docker/dockerfile:labs
+
+FROM node:18-slim AS base
+
+RUN apt-get update \
+  && apt-get install -y dumb-init
+
+FROM base
+
+WORKDIR /app
+
+COPY --chown=node dist/index.mjs /app/
+
+EXPOSE 8000
+
+ENTRYPOINT [ "/usr/bin/dumb-init", "--", "/usr/local/bin/node", "/app/index.mjs" ]

apps/docker-provider/README.md

@@ -0,0 +1,3 @@
+# Docker provider
+
+The `docker-provider` allows the platform to be orchestrator-agnostic. The platform can perform actions such as `INDEX_TASKS` or `INVOKE_TASK` which the provider translates into Docker actions.

apps/docker-provider/package.json

@@ -0,0 +1,31 @@
+{
+  "name": "docker-provider",
+  "private": true,
+  "version": "0.0.1",
+  "description": "",
+  "main": "dist/index.cjs",
+  "scripts": {
+    "build": "npm run build:bundle",
+    "build:bundle": "esbuild src/index.ts --bundle --outfile=dist/index.mjs --platform=node --format=esm --target=esnext --banner:js=\"import { createRequire } from 'module';const require = createRequire(import.meta.url);\"",
+    "build:image": "docker build -f Containerfile . -t docker-provider",
+    "dev": "tsx --no-warnings=ExperimentalWarning --require dotenv/config --watch src/index.ts",
+    "start": "tsx src/index.ts",
+    "typecheck": "tsc --noEmit"
+  },
+  "keywords": [],
+  "author": "",
+  "license": "MIT",
+  "dependencies": {
+    "@trigger.dev/core": "workspace:*",
+    "@trigger.dev/core-apps": "workspace:*",
+    "execa": "^8.0.1",
+    "socket.io-client": "^4.7.4"
+  },
+  "devDependencies": {
+    "@types/node": "^18.19.8",
+    "dotenv": "^16.4.2",
+    "esbuild": "^0.19.11",
+    "tsx": "^4.7.0",
+    "typescript": "^5.3.3"
+  }
+}

apps/docker-provider/src/index.ts

@@ -0,0 +1,302 @@
+import { $, type ExecaChildProcess, execa } from "execa";
+import {
+  SimpleLogger,
+  TaskOperations,
+  ProviderShell,
+  TaskOperationsRestoreOptions,
+  TaskOperationsCreateOptions,
+  TaskOperationsIndexOptions,
+} from "@trigger.dev/core-apps";
+import { setTimeout } from "node:timers/promises";
+import { PostStartCauses, PreStopCauses } from "@trigger.dev/core/v3";
+
+const MACHINE_NAME = process.env.MACHINE_NAME || "local";
+const COORDINATOR_PORT = process.env.COORDINATOR_PORT || 8020;
+const COORDINATOR_HOST = process.env.COORDINATOR_HOST || "127.0.0.1";
+const OTEL_EXPORTER_OTLP_ENDPOINT =
+  process.env.OTEL_EXPORTER_OTLP_ENDPOINT || "http://0.0.0.0:4318";
+
+const logger = new SimpleLogger(`[${MACHINE_NAME}]`);
+
+type InitializeReturn = {
+  canCheckpoint: boolean;
+  willSimulate: boolean;
+};
+
+function isExecaChildProcess(maybeExeca: unknown): maybeExeca is Awaited<ExecaChildProcess> {
+  return typeof maybeExeca === "object" && maybeExeca !== null && "escapedCommand" in maybeExeca;
+}
+
+class DockerTaskOperations implements TaskOperations {
+  #initialized = false;
+  #canCheckpoint = false;
+
+  constructor(private opts = { forceSimulate: false }) {}
+
+  async #initialize(): Promise<InitializeReturn> {
+    if (this.#initialized) {
+      return this.#getInitializeReturn();
+    }
+
+    logger.log("Initializing task operations");
+
+    if (this.opts.forceSimulate) {
+      logger.log("Forced simulation enabled. Will simulate regardless of checkpoint support.");
+    }
+
+    try {
+      await $`criu --version`;
+    } catch (error) {
+      logger.error("No checkpoint support: Missing CRIU binary. Will simulate instead.");
+      this.#canCheckpoint = false;
+      this.#initialized = true;
+
+      return this.#getInitializeReturn();
+    }
+
+    try {
+      await $`docker checkpoint`;
+    } catch (error) {
+      logger.error("No checkpoint support: Docker needs to have experimental features enabled");
+      logger.error("Will simulate instead");
+      this.#canCheckpoint = false;
+      this.#initialized = true;
+
+      return this.#getInitializeReturn();
+    }
+
+    logger.log("Full checkpoint support!");
+
+    this.#initialized = true;
+    this.#canCheckpoint = true;
+
+    return this.#getInitializeReturn();
+  }
+
+  #getInitializeReturn(): InitializeReturn {
+    return {
+      canCheckpoint: this.#canCheckpoint,
+      willSimulate: !this.#canCheckpoint || this.opts.forceSimulate,
+    };
+  }
+
+  async index(opts: TaskOperationsIndexOptions) {
+    await this.#initialize();
+
+    const containerName = this.#getIndexContainerName(opts.shortCode);
+
+    logger.log(`Indexing task ${opts.imageRef}`, {
+      host: COORDINATOR_HOST,
+      port: COORDINATOR_PORT,
+    });
+
+    try {
+      logger.debug(
+        await execa("docker", [
+          "run",
+          "--network=host",
+          "--rm",
+          `--env=INDEX_TASKS=true`,
+          `--env=TRIGGER_SECRET_KEY=${opts.apiKey}`,
+          `--env=TRIGGER_API_URL=${opts.apiUrl}`,
+          `--env=TRIGGER_ENV_ID=${opts.envId}`,
+          `--env=OTEL_EXPORTER_OTLP_ENDPOINT=${OTEL_EXPORTER_OTLP_ENDPOINT}`,
+          `--env=POD_NAME=${containerName}`,
+          `--env=COORDINATOR_HOST=${COORDINATOR_HOST}`,
+          `--env=COORDINATOR_PORT=${COORDINATOR_PORT}`,
+          `--name=${containerName}`,
+          `${opts.imageRef}`,
+        ])
+      );
+    } catch (error: any) {
+      if (!isExecaChildProcess(error)) {
+        throw error;
+      }
+
+      logger.error("Index failed:", {
+        opts,
+        exitCode: error.exitCode,
+        escapedCommand: error.escapedCommand,
+        stdout: error.stdout,
+        stderr: error.stderr,
+      });
+    }
+  }
+
+  async create(opts: TaskOperationsCreateOptions) {
+    await this.#initialize();
+
+    const containerName = this.#getRunContainerName(opts.runId);
+
+    try {
+      logger.debug(
+        await execa("docker", [
+          "run",
+          "--network=host",
+          "--detach",
+          `--env=TRIGGER_ENV_ID=${opts.envId}`,
+          `--env=TRIGGER_RUN_ID=${opts.runId}`,
+          `--env=OTEL_EXPORTER_OTLP_ENDPOINT=${OTEL_EXPORTER_OTLP_ENDPOINT}`,
+          `--env=POD_NAME=${containerName}`,
+          `--env=COORDINATOR_HOST=${COORDINATOR_HOST}`,
+          `--env=COORDINATOR_PORT=${COORDINATOR_PORT}`,
+          `--name=${containerName}`,
+          `${opts.image}`,
+        ])
+      );
+    } catch (error) {
+      if (!isExecaChildProcess(error)) {
+        throw error;
+      }
+
+      logger.error("Create failed:", {
+        opts,
+        exitCode: error.exitCode,
+        escapedCommand: error.escapedCommand,
+        stdout: error.stdout,
+        stderr: error.stderr,
+      });
+    }
+  }
+
+  async restore(opts: TaskOperationsRestoreOptions) {
+    await this.#initialize();
+
+    const containerName = this.#getRunContainerName(opts.runId);
+
+    if (!this.#canCheckpoint || this.opts.forceSimulate) {
+      logger.log("Simulating restore");
+
+      const unpause = logger.debug(await $`docker unpause ${containerName}`);
+
+      if (unpause.exitCode !== 0) {
+        throw new Error("docker unpause command failed");
+      }
+
+      await this.#sendPostStart(containerName);
+      return;
+    }
+
+    const { exitCode } = logger.debug(
+      await $`docker start --checkpoint=${opts.checkpointRef} ${containerName}`
+    );
+
+    if (exitCode !== 0) {
+      throw new Error("docker start command failed");
+    }
+
+    await this.#sendPostStart(containerName);
+  }
+
+  async delete(opts: { runId: string }) {
+    await this.#initialize();
+
+    const containerName = this.#getRunContainerName(opts.runId);
+    await this.#sendPreStop(containerName);
+
+    logger.log("noop: delete");
+  }
+
+  async get(opts: { runId: string }) {
+    await this.#initialize();
+
+    logger.log("noop: get");
+  }
+
+  #getIndexContainerName(suffix: string) {
+    return `task-index-${suffix}`;
+  }
+
+  #getRunContainerName(suffix: string) {
+    return `task-run-${suffix}`;
+  }
+
+  async #sendPostStart(containerName: string): Promise<void> {
+    try {
+      const port = await this.#getHttpServerPort(containerName);
+      logger.debug(await this.#runLifecycleCommand(containerName, port, "postStart", "restore"));
+    } catch (error) {
+      logger.error("postStart error", { error });
+      throw new Error("postStart command failed");
+    }
+  }
+
+  async #sendPreStop(containerName: string): Promise<void> {
+    try {
+      const port = await this.#getHttpServerPort(containerName);
+      logger.debug(await this.#runLifecycleCommand(containerName, port, "preStop", "terminate"));
+    } catch (error) {
+      logger.error("preStop error", { error });
+      throw new Error("preStop command failed");
+    }
+  }
+
+  async #getHttpServerPort(containerName: string): Promise<number> {
+    // We first get the correct port, which is random during dev as we run with host networking and need to avoid clashes
+    // FIXME: Skip this in prod
+    const logs = logger.debug(await $`docker logs ${containerName}`);
+    const matches = logs.stdout.match(/http server listening on port (?<port>[0-9]+)/);
+
+    const port = Number(matches?.groups?.port);
+
+    if (!port) {
+      throw new Error("failed to extract port from logs");
+    }
+
+    return port;
+  }
+
+  async #runLifecycleCommand<THookType extends "postStart" | "preStop">(
+    containerName: string,
+    port: number,
+    type: THookType,
+    cause: THookType extends "postStart" ? PostStartCauses : PreStopCauses,
+    retryCount = 0
+  ): Promise<ExecaChildProcess> {
+    try {
+      return await execa("docker", [
+        "exec",
+        containerName,
+        "busybox",
+        "wget",
+        "-q",
+        "-O-",
+        `127.0.0.1:${port}/${type}?cause=${cause}`,
+      ]);
+    } catch (error: any) {
+      if (type === "postStart" && retryCount < 6) {
+        logger.debug(`retriable ${type} error`, { retryCount, message: error?.message });
+        await setTimeout(exponentialBackoff(retryCount + 1, 2, 50, 1150, 50));
+
+        return this.#runLifecycleCommand(containerName, port, type, cause, retryCount + 1);
+      }
+
+      logger.error(`final ${type} error`, { message: error?.message });
+      throw new Error(`${type} command failed after ${retryCount - 1} retries`);
+    }
+  }
+}
+
+const provider = new ProviderShell({
+  tasks: new DockerTaskOperations({ forceSimulate: true }),
+  type: "docker",
+});
+
+provider.listen();
+
+function exponentialBackoff(
+  retryCount: number,
+  exponential: number,
+  minDelay: number,
+  maxDelay: number,
+  jitter: number
+): number {
+  // Calculate the delay using the exponential backoff formula
+  const delay = Math.min(Math.pow(exponential, retryCount) * minDelay, maxDelay);
+
+  // Calculate the jitter
+  const jitterValue = Math.random() * jitter;
+
+  // Return the calculated delay with jitter
+  return delay + jitterValue;
+}