Trigger.dev
@triggerdev.bsky.social
Open source background jobs and AI infrastructure (YC W23).
→ http://trigger.dev
→ http://github.com/triggerdotdev
→ http://trigger.dev/discord
The official 🦋 account for Trigger.dev
→ http://trigger.dev
→ http://github.com/triggerdotdev
→ http://trigger.dev/discord
The official 🦋 account for Trigger.dev
Update to 4.1.1 to access the new Prisma extension. Read the full changelog here: trigger.dev/changelog/p...
Prisma 6 & 7 support | Trigger.dev
Complete redesign of our Prisma extension with three modes to support Prisma 6.x, 7.x, and everything in between.
trigger.dev
November 19, 2025 at 3:00 PM
Update to 4.1.1 to access the new Prisma extension. Read the full changelog here: trigger.dev/changelog/p...
③ Modern Mode (Prisma 7+)
The future is here. Prisma 7's new prisma-client removes Rust dependencies entirely, enabling faster installs and simpler deployments.
The future is here. Prisma 7's new prisma-client removes Rust dependencies entirely, enabling faster installs and simpler deployments.
November 19, 2025 at 3:00 PM
③ Modern Mode (Prisma 7+)
The future is here. Prisma 7's new prisma-client removes Rust dependencies entirely, enabling faster installs and simpler deployments.
The future is here. Prisma 7's new prisma-client removes Rust dependencies entirely, enabling faster installs and simpler deployments.
② Engine-Only Mode (Prisma 6.7+ with custom output)
For teams using Prisma 6.7+ with a custom output directory but still using the Rust-based engine. This mode gives you control over the client generation process, but will take care of installing the engine binaries for you when deploying.
For teams using Prisma 6.7+ with a custom output directory but still using the Rust-based engine. This mode gives you control over the client generation process, but will take care of installing the engine binaries for you when deploying.
November 19, 2025 at 3:00 PM
② Engine-Only Mode (Prisma 6.7+ with custom output)
For teams using Prisma 6.7+ with a custom output directory but still using the Rust-based engine. This mode gives you control over the client generation process, but will take care of installing the engine binaries for you when deploying.
For teams using Prisma 6.7+ with a custom output directory but still using the Rust-based engine. This mode gives you control over the client generation process, but will take care of installing the engine binaries for you when deploying.
① Legacy Mode (Prisma 6.x and earlier)
Use this when you're running Prisma 6.x or earlier with the standard prisma-client-js provider.
Use this when you're running Prisma 6.x or earlier with the standard prisma-client-js provider.
November 19, 2025 at 3:00 PM
① Legacy Mode (Prisma 6.x and earlier)
Use this when you're running Prisma 6.x or earlier with the standard prisma-client-js provider.
Use this when you're running Prisma 6.x or earlier with the standard prisma-client-js provider.
If you’re streaming AI completions or any long-running job from tasks, Realtime Streams v2 is more reliable, more flexible, and way harder to break. Upgrade to the latest SDK and you’re set.
Changelog: trigger.dev/changelog/r...
Changelog: trigger.dev/changelog/r...
Realtime Streams v2 | Trigger.dev
Introducing Realtime Streams v2: Unlimited, Reliable, and Type-Safe
trigger.dev
November 18, 2025 at 5:15 PM
If you’re streaming AI completions or any long-running job from tasks, Realtime Streams v2 is more reliable, more flexible, and way harder to break. Upgrade to the latest SDK and you’re set.
Changelog: trigger.dev/changelog/r...
Changelog: trigger.dev/changelog/r...
→Dashboard visibility:
Streams show up as first-class citizens now. You can inspect chunks in order, replay them, and debug long-running tasks without digging through logs or guessing what happened.
Streams show up as first-class citizens now. You can inspect chunks in order, replay them, and debug long-running tasks without digging through logs or guessing what happened.
November 18, 2025 at 5:15 PM
→Dashboard visibility:
Streams show up as first-class citizens now. You can inspect chunks in order, replay them, and debug long-running tasks without digging through logs or guessing what happened.
Streams show up as first-class citizens now. You can inspect chunks in order, replay them, and debug long-running tasks without digging through logs or guessing what happened.
→Pipe from your tasks:
Tasks can push chunked output directly into a stream without blocking:
Tasks can push chunked output directly into a stream without blocking:
November 18, 2025 at 5:15 PM
→Pipe from your tasks:
Tasks can push chunked output directly into a stream without blocking:
Tasks can push chunked output directly into a stream without blocking:
→React integration: useRealtimeStream hook
Frontend consumption is much more straightforward:
Frontend consumption is much more straightforward:
November 18, 2025 at 5:15 PM
→React integration: useRealtimeStream hook
Frontend consumption is much more straightforward:
Frontend consumption is much more straightforward:
→ Define your streams:
You define a stream once and use it everywhere with full type safety built-in: No per-run setup, no ad-hoc keys, no guessing the shape of chunks later.
You define a stream once and use it everywhere with full type safety built-in: No per-run setup, no ad-hoc keys, no guessing the shape of chunks later.
November 18, 2025 at 5:15 PM
→ Define your streams:
You define a stream once and use it everywhere with full type safety built-in: No per-run setup, no ad-hoc keys, no guessing the shape of chunks later.
You define a stream once and use it everywhere with full type safety built-in: No per-run setup, no ad-hoc keys, no guessing the shape of chunks later.
The DX is much cleaner: Same definition works in tasks, API routes, or React via useRealtimeStream:
November 18, 2025 at 5:15 PM
The DX is much cleaner: Same definition works in tasks, API routes, or React via useRealtimeStream:
The core change: unlimited chunks, unlimited active streams, longer retention, and 300 MiB max size. v1 had tiny caps everywhere; v2 removes them:
November 18, 2025 at 5:15 PM
The core change: unlimited chunks, unlimited active streams, longer retention, and 300 MiB max size. v1 had tiny caps everywhere; v2 removes them:
Streaming long-running AI outputs is inherently fragile. Network hiccups happen. Connections drop.
With v2, streams now auto-resume on both reader and writer. If a connection drops mid-generation, it picks up from the last confirmed chunk without you doing anything.
With v2, streams now auto-resume on both reader and writer. If a connection drops mid-generation, it picks up from the last confirmed chunk without you doing anything.
November 18, 2025 at 5:15 PM
Streaming long-running AI outputs is inherently fragile. Network hiccups happen. Connections drop.
With v2, streams now auto-resume on both reader and writer. If a connection drops mid-generation, it picks up from the last confirmed chunk without you doing anything.
With v2, streams now auto-resume on both reader and writer. If a connection drops mid-generation, it picks up from the last confirmed chunk without you doing anything.
Read how Pallet uses Trigger.dev to automate global logistics workflows → trigger.dev/customers/p...
Accelerating global logistics workflows using AI copilots | Trigger.dev
Karl Kaiser, Engineer at Pallet, shares how his team uses Trigger.dev to power their logistics AI copilot, automating the exchange of shipment and transport data between multiple systems in real-time.
trigger.dev
November 13, 2025 at 8:47 PM
Read how Pallet uses Trigger.dev to automate global logistics workflows → trigger.dev/customers/p...
They've raised a $27M Series B, lead by General Catalyst, with the goal of giving every logistics team their own AI copilot ("CoPallet"), that understands and acts on real-time freight data across carriers, shipments, and transport modes.
November 13, 2025 at 8:47 PM
They've raised a $27M Series B, lead by General Catalyst, with the goal of giving every logistics team their own AI copilot ("CoPallet"), that understands and acts on real-time freight data across carriers, shipments, and transport modes.
Pallet use Trigger to run long-running, reasoning-heavy AI workflows directly in their TypeScript codebase, with elastic scaling, full observability, and zero extra DevOps. This allows them to move fast while staying in control as they scale.
November 13, 2025 at 8:47 PM
Pallet use Trigger to run long-running, reasoning-heavy AI workflows directly in their TypeScript codebase, with elastic scaling, full observability, and zero extra DevOps. This allows them to move fast while staying in control as they scale.
⭐️ View the code in our examples repo. It's open source and can be used as a starting point for your own projects.
github.com/triggerdotd...
github.com/triggerdotd...
examples/product-image-generator at main · triggerdotdev/examples
A collection of open source full-stack projects using Trigger.dev that anyone can fork and run - triggerdotdev/examples
github.com
November 6, 2025 at 5:45 PM
⭐️ View the code in our examples repo. It's open source and can be used as a starting point for your own projects.
github.com/triggerdotd...
github.com/triggerdotd...
1️⃣ Upload triggers a batch task
2️⃣ 3 images are generated in parallel
4️⃣ Generated images auto-upload to R2 storage
5️⃣ Frontend shows live run progress
2️⃣ 3 images are generated in parallel
4️⃣ Generated images auto-upload to R2 storage
5️⃣ Frontend shows live run progress
November 6, 2025 at 5:45 PM
1️⃣ Upload triggers a batch task
2️⃣ 3 images are generated in parallel
4️⃣ Generated images auto-upload to R2 storage
5️⃣ Frontend shows live run progress
2️⃣ 3 images are generated in parallel
4️⃣ Generated images auto-upload to R2 storage
5️⃣ Frontend shows live run progress
Waitpoints: Pause task execution for @replicate webhooks using wait.forToken(). No polling, no timeouts, just normal async waiting.
November 6, 2025 at 5:45 PM
Waitpoints: Pause task execution for @replicate webhooks using wait.forToken(). No polling, no timeouts, just normal async waiting.
Realtime Updates: Live frontend updates via useRealtimeRun(). Watch generation status change from "Generating..." to "Completed" without any polling.
November 6, 2025 at 5:45 PM
Realtime Updates: Live frontend updates via useRealtimeRun(). Watch generation status change from "Generating..." to "Completed" without any polling.
Batch Triggering: Trigger 3 AI generations in parallel with batchTriggerAndWait(). One line replaces complex coordination logic.
November 6, 2025 at 5:45 PM
Batch Triggering: Trigger 3 AI generations in parallel with batchTriggerAndWait(). One line replaces complex coordination logic.
