Also, I'm about to release [@lumenize/alarms](<https://github.com/lumenize/lumenize/tree/main/packag

Also, I'm about to release @lumenize/alarms package modeled after Actor Alarms, with a few differences including built in support for testing. Docs still need work, but can be found here

LLarry Also, I'm about to release [@lumenize/alarms](<https://github.com/lumenize/lumen...

echoes221•11/11/25, 4:03 PM

Thanks - I was looking for something that would let me force some alarms / time travel, but I ended up adding a trigger locally and updating my DB so that I can setup the right pathways, then force it to work

Eechoes221 Thanks - I was looking for something that would let me force some alarms / time ...

LarryOP•11/11/25, 4:11 PM

That's the right instinct! It's why I added a testing trigger to Lumenize Alarms. TBH, I rarely use my 100x speedup alarm simulation since I added that.

LLarry That's the right instinct! It's why I added a testing trigger to Lumenize Alarms...

echoes221•11/11/25, 5:00 PM

I've got a bit of a complex stripe subscription schedule setup, and needed to start a sub in future so that billing anchors etc are in the right place, so needed to trigger the future event to ensure that the alarm triggers the schedule build correctly

1984 Ford Laser•11/13/25, 1:20 AM

Do DOs still have a 1MiB limit on outgoing WS messages? The limits page states that the new 32MB message limit is only for incoming messages

LarryOP•11/14/25, 1:16 AM

Does anyone have any experience with setting a zero delay alarm? I assume that they won't be skipped but that's one question. However, my real question is what's the delay between when you set the alarm and when it actually fires? Is it 1-2 ms? or can it be upwards of 10ms?

LLarry Does anyone have any experience with setting a zero delay alarm? I assume that t...

lambrospetrou•11/14/25, 7:42 AM

It can be upwards of 10ms for sure. You are not going to get a frequency at the single-double-digit milliseconds.

LarryOP•11/14/25, 1:44 PM

I assume that awaiting a workers rpc call opens input/output gate? But how does not awaiting work? There might not be a fetch or message handler to finish to open them?

mihaaai•11/14/25, 7:11 PM

Is calling ctx.storage.deleteAll()ctx.storage.deleteAll() in a hibernated or non-hibernated Durable Object gonna also disconnect all clients attached to it? Or is just cleaning the storage (kv, sqlite)?

LarryOP•11/14/25, 7:57 PM

Also, what round trip latency should I expect for Workers RPC calls. I'm seeing about 20ms?

Mmihaaai Is calling `ctx.storage.deleteAll()` in a **hibernated or non-hibernated** Durab...

LarryOP•11/14/25, 9:03 PM

If it's hibernated, the only thing it can do is wake up. But it will not disconnect clients. It will just delete storage.

LLarry If it's hibernated, the only thing it can do is wake up. But it will not disconn...

1984 Ford Laser•11/14/25, 10:49 PM

tbh I would have expected it to disconnect all clients, that's interesting

avi•11/15/25, 3:01 AM

has anyone here managed to get https://platform.openai.com/docs/guides/realtime-sip#monitor-call-events working in a worker/do? starting a new session works by either authing through subprotocols or using CloudflareRealtimeTransportLayer however, i can't for the life of me get it to work when hitting the wss://api.openai.com/v1/realtime?call_id={call_id} endpoint to connecting to an inflight call

Aavi has anyone here managed to get https://platform.openai.com/docs/guides/realtime-...

1984 Ford Laser•11/15/25, 3:21 AM

Are you calling the WSS URL in the DO or in Worker code? Are you just fetching, or creating a new WebSocket() object? I've had varied success using either solution with a few different WebSocket servers

11984 Ford Laser Are you calling the WSS URL in the DO or in Worker code? Are you just fetching, ...

avi•11/15/25, 3:22 AM

up to this point just in the DO.

1984 Ford Laser•11/15/25, 3:23 AM

You could try having your DO invoke a method on the Worker to create a new WebSocket connection, then pass the socket back to the DO when done

avi•11/15/25, 3:23 AM

interesting. will give it a shot. thanks

1984 Ford Laser•11/15/25, 3:24 AM

And are you fetch()fetch()ing or WebSocket()WebSocket()ing the URL? Try the other option to whatever you're using now

avi•11/15/25, 3:25 AM

i've tried both. what's strange is that i can create a fresh session with out any issues (the wss://api.openai.com/v1/realtime?model={model} endpoint. i only run into trouble when try to connect to a call originated through a sip between twilio <-> openai

Aavi i've tried both. what's strange is that i can create a fresh session with out an...

1984 Ford Laser•11/15/25, 3:28 AM

Does it work with a 'normal' WebSocket client? Something like Postman

1984 Ford Laser•11/15/25, 3:28 AM

I always like to sanity check on other platform

avi•11/15/25, 3:28 AM

yeah. i even threw together a quick elixir implementation to sanity check it and that worked fine

1984 Ford Laser•11/15/25, 3:29 AM

yeah cool

1984 Ford Laser•11/15/25, 3:29 AM

weird then

avi•11/15/25, 3:29 AM

but the key differnce there is with elixir i'm able to pass a standard auth header with the ws.

avi•11/15/25, 3:30 AM

i suspect there maybe an issue on openai's end with handling auth for the call_id endpoint. their docs are abit all over the place due to the beta -> ga transition.

Silvan•11/15/25, 7:31 PM

does anyone know if http streaming with DOs would ever make sense if ws hibernation exists?
My reason being is that cap n web dosent seem to support hibernation. so i thought about just using it with http to send data. and use the websocket to read updates. which seems silly but i guess as good as it gets?

Silvan•11/15/25, 7:46 PM

like yea i could just use the ws bidirectional like a normal human but the dx of rpc is just too :ChefKiss:

SSilvan does anyone know if http streaming with DOs would ever make sense if ws hibernat...

LarryOP•11/15/25, 7:48 PM

Lumenize RPC supports hibernating WebSockets and it's got some other advantages over Cap'n Web. https://lumenize.com/docs/rpc/introduction

Introduction | Lumenize

Understanding Lumenize RPC for Durable Objects

LLarry Lumenize RPC supports hibernating WebSockets and it's got some other advantages ...

Silvan•11/15/25, 7:56 PM

hm interesting i will have a look thanks

Aavi has anyone here managed to get https://platform.openai.com/docs/guides/realtime-...

avi•11/15/25, 8:18 PM

incase anyone slese stumbles across this. the issue is that you need to use the same literal api key that was used to accept the call. since we except the call on our backend and then join it from a DO we had two different keys provisioned. I opened up an issue.

LLarry Lumenize RPC supports hibernating WebSockets and it's got some other advantages ...

Silvan•11/15/25, 8:41 PM

i see you made this! very cool. even tho the testing docs arent my gem. honestly i also kind of miss a super simple getting started.

i wasnt aware id need @lumenize/utils routeDORequest. i mean i technically dont but else it gets a bit rough

SSilvan i see you made this! very cool. even tho the testing docs arent my gem. honestly...

LarryOP•11/15/25, 8:57 PM

Yeah I have stopped using the testing docs format going forward. It's a lot of noise for learning. Someday, I will go back and convert the RPC docs. Huge thanks for the input. Keep it coming even if it is negative.

LLarry Yeah I have stopped using the testing docs format going forward. It's a lot of n...

Silvan•11/15/25, 8:58 PM

i will gather all my thoughts

LarryOP•11/15/25, 8:59 PM

routeDORequest is hono compatible if that helps

SSilvan i see you made this! very cool. even tho the testing docs arent my gem. honestly...

gruntlord6•11/16/25, 2:39 PM

getting started would be a nice addition

Ggruntlord6 getting started would be a nice addition

LarryOP•11/16/25, 6:15 PM

That's two votes. I better do sooner than later. Thanks for the support.

SSilvan does anyone know if http streaming with DOs would ever make sense if ws hibernat...

PhoenixSpace•11/16/25, 11:07 PM

I actually made a Capn Web backend that respects hibernation, and I'll FOSS it soon if I get the time to make unit tests and properly test it. I'll note though, it involves using the sqlite backend a lot to write state information, and requires both sides maintain state correctly.

If you want to implement it yourself, it's not that bad, but the DX of storing RPC function stubs and restoring them is... not great currently is why I haven't FOSS'd it yet. Probably the best approach is to act as if you are calling functions on the client from the server without passing stubs up, but that also doesn't have the best DX.

RPC is great, but for notification/pushed based setups I think just using normal websocket bidirectional will be easier usually

PhoenixSpace•11/16/25, 11:09 PM

(if anyone has creative ideas to make serializing and deserializing function stubs across sqlite simple, please let me know)

PhoenixSpace•11/16/25, 11:13 PM

Tangentially related, but I read the reasoning here https://github.com/cloudflare/capnweb/issues/36 and I find it interesting. Hibernating RPC would be cool, but I don't quite understand how that would work in the reverse direction of notifying clients from the server. Storing function stubs (actually rn I'm storing the entire RPC conversation in sqlite and just re-running it when the DO wakes up) in SQLite and using that to call data to clients honestly seems extremely powerful

PhoenixSpace•11/16/25, 11:14 PM

Also tangential, but huge shoutout to whoever wrote the Capn Web protocol spec because it was one of the most digestible protocol specs I've ever read

1984 Ford Laser•11/16/25, 11:45 PM

Is there a way to have an RpcTargetRpcTarget have access to all DO methods like storage? Not just defined RPC methods

11984 Ford Laser Is there a way to have an `RpcTarget` have access to all DO methods like storage...

lambrospetrou•11/17/25, 10:28 AM

You can pass the DO instance itself to the RpcTarget class you create and then it will be able to call anything public on that DO. Unless I misunderstood what you mean by "access to all DO methods".
Only the code within the DO itself has access to ctx.storagectx.storage unless you expose it yourself.

Llambrospetrou You can pass the DO instance itself to the RpcTarget class you create and then i...

LarryOP•11/17/25, 11:03 AM

@1984 Ford Laser , are you talking about a capability like this?

using client = getLumenizeClient('instance-name');
client.ctx.storage.kv.put('key', 'value');
const result = await client.ctx.storage.kv.get('key');
expect(result).toBe('value');

using client = getLumenizeClient('instance-name');
client.ctx.storage.kv.put('key', 'value');
const result = await client.ctx.storage.kv.get('key');
expect(result).toBe('value');

LLarry @1984 Ford Laser , are you talking about a capability like [this](<https://lumen...

lambrospetrou•11/17/25, 11:06 AM

I would personally be very scared of writing or reviewing this kind of code. It's inevitable someone will think the semantics of "running inside the DO" apply here and get incorrect results in the end.
Apart from quick demos and toy projects, just expose a proper RPC and bundle all your actions inside the DO so that you can reason about what's going on.

Llambrospetrou I would personally be very scared of writing or reviewing this kind of code. It'...

lambrospetrou•11/17/25, 11:09 AM

Side note: you call it "batches" in the Luminize docs, but we call it "RPC promise pipelining" everywhere else in our docs, so it might be confusing to users that go through CF docs as well: https://developers.cloudflare.com/workers/runtime-apis/rpc/#promise-pipelining

Also, a batch might imply some sort of transaction-ish guarantees, which does not exist here.

Llambrospetrou I would personally be very scared of writing or reviewing this kind of code. It'...

LarryOP•11/17/25, 11:09 AM

I actually agree. I turn it off in production. I pretty much only use that capability for prototyping and testing. The testing use case is 80% of that and the prototyping 20%.

Llambrospetrou Side note: you call it "batches" in the Luminize docs, but we call it "RPC promi...

LarryOP•11/17/25, 11:13 AM

It's actually three things for me and batching is the least of those three. It's operation chaining and nesting (OCAN). That's the two most important parts (chaining and nesting). I'm in the process or removing batching for my next round.

I'm not a fan of the phrase, "promise pipelining". The promise part is not even very accurate in Cap'n Web as far as I can tell. It's a thenable Proxy (just like Lumenize RPC), not an actual Promise. More importantly, the phrase describes the implementation. "Operation chaining and nesting" describes what it does, which is why I like it better.

LarryOP•11/17/25, 11:14 AM

Cap'n Web has a special thing for "batching" but I never use my own batching, which is why I'm removing it.

LarryOP•11/17/25, 11:17 AM

Apart from quick demos and toy projects, just expose a proper RPC and bundle all your actions inside the DO so that you can reason about what's going on.

Bottom line. This is exactly what we should be doing!

LarryOP•11/17/25, 11:18 AM

Cap'n Web's .map().map() over RPC is an example that breaks that, IMO.