feature: add compute provider to worker deployment by jaypipes · Pull Request #704 · temporalio/api

jaypipes · 2026-01-28T15:30:03Z

feature: introduce compute config

Introduces a new compute.v1 package of proto messages that house the
various instructions and information needed by worker control plane
controllers

temporal.api.compute.v1.Config has two fields, provider and scaler
that store the configuration information for the server-owned worker
setup and how the worker control plane should treat worker lifecycle
events (such as when no pollers exist for a task queue that has received
a task).

The temporal.api.compute.v1.Config.provider field is of type
temporal.api.compute.v1.Provider. This message is a simple wrapper
around a blob of bytes that contains configuration settings for a worker
control plane controller.

The configuration settings for the worker control plane controller are
specific to different compute providers. The first compute provider we
are modeling is AWS Lambda, and the ProviderDetailAWSLambda message
houses the ARN of the AWS Lambda Function to be invoked and an optional
ARN of the IAM Role that should be assumed by a worker control plane
controller when invoking the function.

The temporal.api.compute.v1.Config.scaler field is of type
temporal.api.compute.v1.Scaler. This message contains instructions for
when the worker lifecycle controller should scale up or down the number
of workers in a WorkerDeployment or WorkerDeploymentVersion.

Adds a new field compute_config to WorkerDeploymentInfo and
WorkerDeploymentVersionInfo that contains a
temporal.api.compute.v1.Config message.

Both WorkerDeployment and WorkerDeploymentVersion have this field
in order to handle versioned workers properly. For example, if the user
wants to use PINNED temporal versioning, they will be creating a
WorkerDeploymentVersion that is specific to, say, a qualified Lambda
Function ARN. Therefore we need to place compute provider configuration
on a WorkerDeploymentVersion in addition to WorkerDeployment.

Adds three new gRPC API methods to WorkflowService:

CreateWorkerDeployment creates a new WorkerDeployment containing the
supplied compute configuration.
UpdateWorkerDeploymentComputeConfig updates an existing
WorkerDeployment's compute configuration.
UpdateWorkerDeploymentComputeConfig updates an existing
WorkerDeploymentVersion's compute configuration.

Signed-off-by: Jay Pipes jay.pipes@temporal.io

CLAassistant · 2026-01-28T15:30:10Z

All committers have signed the CLA.

02strich · 2026-01-28T17:28:36Z

temporal/api/deployment/v1/compute_provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message ComputeProvider {


can this have a field like provider_type or similar to identify who should interpret the details?

Mentioned in adjacent comment at https://github.com/temporalio/api/pull/704/changes#r2737966940, but I think the config about which control plane this goes to is missing atm

google.protobuf.Any is made for this sort of situation, but it doesn't handle the encryption and other payload metadata. Maybe we should say the Payload contains an Any, which contains the underlying config plus a type url?

Payloads for proto messages actually do put their type in the metadata just like Any, so they work for this use case, but I am going to be making a comment down there right now concerning issues with using proto here.

02strich · 2026-01-28T17:29:08Z

buf.lock

is this change intentional?

no... it happened when I ran make buf-lint locally :(

cretz

Looking good (and simple which is awesome). This cannot really be merged to main anytime soon until the implementation using it has basically been completely built due to concerns of this being mutated as we work on the impl.

cretz · 2026-01-28T18:30:39Z

temporal/api/deployment/v1/compute_provider.proto

+// Function via its ARN. This will start the Worker that is housed within that
+// AWS Lambda Function and the Worker will begin listening to Tasks on the
+// WorkerDeployment's configured TaskQueue.
+message ComputeProviderAWSLambda {


Suggested change

message ComputeProviderAWSLambda {

message ComputeProviderDetailsAWSLambda {

Because this is not a "compute provider" as we've now defined it since "compute provider" as we've now defined it is "compute provider details" + "scaling config" + "endpoint config".

@cretz I don't think putting scaling config into ComputeProvider is appropriate. The scaling algorithm for how often a StartWorker API is called is orthogonal to configuration for how that StartWorker code is implemented for a particular compute platform.

I've added a different Scaler field to WorkerDeploymentVersionInfo that is used to store settings related to auto-scaling algorithms.

Makes sense, but we are back at the problem of "what do you call this collection of things needed"? Because you are still putting scaling config on a ComputeConfigRequest. Is it part of "compute config"? If so, IMO there should be an all encompassing ComputeConfig (even if that has just a field for scaler, a field for compute_provider, etc). If not, then it shouldn't be in a call about updating compute config. These two fields have no value independently and aren't independent based on the current modelling/RPC.

cretz · 2026-01-28T18:32:20Z

temporal/api/deployment/v1/compute_provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message ComputeProvider {


I think even as placeholders, we should go ahead and add fields that represent scaling algo config and control plane Nexus endpoint config (even if for many they will use a "system" one that we run)

cretz · 2026-01-28T18:32:54Z

temporal/api/deployment/v1/compute_provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message ComputeProvider {


Mentioned in adjacent comment at https://github.com/temporalio/api/pull/704/changes#r2737966940, but I think the config about which control plane this goes to is missing atm

cretz · 2026-01-28T18:34:58Z

temporal/api/workflowservice/v1/request_response.proto

+
+    // Contains the new worker compute provider configuration for the
+    // WorkerDeployment.
+	temporal.api.deployment.v1.ComputeProvider compute_provider = 20;


I assume the absence/null of this represents removing compute provider?

cretz · 2026-01-28T18:36:40Z

temporal/api/workflowservice/v1/service.proto

+    // Creates or replaces the compute provider configuration for a Worker
+    // Deployment.


Needs to be pretty clear/obvious here or somewhere that this completely changes how task queues operate and completely changes the worker paradigm.

I fear the triviality of the API/CLI names and such don't convey the significance of this "switch flip" that completely changes how these task queues will work.

Does it? Can these task queues not be polled like normal? (as a "base load" below burstable compute) I know we're not going to focus on that case at first but will we disable it?

I think for the proper server-side worker lifecycle management and clearer understanding by users, a task queue needs to use server-owned workers or user-owned workers. Having a hybrid can cause unpredictability and shift expected numbers, known workers, etc. If a base load is needed, why can't the server control the lifetimes of those workers too (or at least know about them)? If that means that you just tell the server that these may exist, that can make sense maybe. But to have the server not even know about workers on a task queue where it expects to control workers is going to cause confusion and unpredictability IMO.

I'd go as far as saying from a server POV a worker should not be allowed to poll on a serverless task queue unless it was a worker the server configured. This way you don't have silent processing workers out there the server can't account for in its scaling/versioning logic.

dnr · 2026-01-29T00:21:43Z

temporal/api/deployment/v1/compute_provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message ComputeProvider {


google.protobuf.Any is made for this sort of situation, but it doesn't handle the encryption and other payload metadata. Maybe we should say the Payload contains an Any, which contains the underlying config plus a type url?

dnr · 2026-01-29T00:24:47Z

temporal/api/compute/v1/provider.proto

+message ComputeProviderAWSLambda {
+    // The Qualified or Unqualified ARN of the AWS Lambda Function to be
+    // invoked.
+    string invoke = 1;


maybe:

Suggested change

string invoke = 1;

string invoke_arn = 1;

without arn in the name I would have assumed this is an http url

@dnr I've updated the comment above. It can actually be the function URL too :) Which is why I left the type suffix off the field name.

@jaypipes this needs to be the ARN - no plan for supporting function URLs at this point.

dnr · 2026-01-29T00:26:12Z

temporal/api/workflowservice/v1/service.proto

+    // Creates or replaces the compute provider configuration for a Worker
+    // Deployment.


Does it? Can these task queues not be polled like normal? (as a "base load" below burstable compute) I know we're not going to focus on that case at first but will we disable it?

cretz · 2026-01-29T13:50:52Z

temporal/api/deployment/v1/compute_provider.proto

+    // An individual ComputeProviderAWSLambda,
+    // ComputeProviderGoogleCloudRun, etc will be encrypted and
+    // serialized into compute_provider_config.
+    temporal.api.common.v1.Payload provider_details = 1;


I wanted to add a note here based on internal discussions. This can only really be a payload for custom providers, because people have "encrypt every payload" proxies and cloud side needing to read these prevents people from encrypting this.

But there is a problem using proto at all here IMO, see comment below at https://github.com/temporalio/api/pull/704/changes#r2741776135.

So we should have provider type w/ a oneof I think:

Suggested change

temporal.api.common.v1.Payload provider_details = 1;

// Passed to provider Nexus endpoint to identify the contents of the provider details

string provider_type = 1;

oneof provider_details {

// Non-encoded provider details, used in situations where decoding cannot occur

bytes provider_details_open = 2;

// Encoded provider details

temporal.api.common.v1.Payload provider_details_payload = 3;

}

Note how provider_details_open is bytes and not google.protobuf.Any here

@cretz I've given it another try... please see the updated struct.

cretz · 2026-01-29T13:54:11Z

temporal/api/deployment/v1/compute_provider.proto

+// Function via its ARN. This will start the Worker that is housed within that
+// AWS Lambda Function and the Worker will begin listening to Tasks on the
+// WorkerDeployment's configured TaskQueue.
+message ComputeProviderAWSLambda {


IMO these should not be protos, these should be JSON schema defined as part of a Nexus IDL (https://github.com/nexus-rpc/nexus-rpc-gen). People should have to use proto to implement a compute provider accepting this (including us).

jaypipes

@cretz @02strich @dnr updated in the following major ways:

Made the compute configuration a field on WorkerDeploymentVersionInfo instead of WorkerDeploymentInfo. This was so we can handle the versioning use cases properly (we will be making the UX for unversioned transparently auto-create a "latest" version when one doesn't already exist, BTW).
Added a Scaler message type with simple min_replicas/max_replicas fields to start off with. Like ComputeProvider, I've placed Scaler on WorkerDeploymentVersionInfo instead of WorkerDeploymentVersion.
Updated the gRPC API call from UpdateWorkerDeploymentComputeProvider to UpdateWorkerDeploymentComputeConfig and added the (orthogonal) compute_provider and scaler fields to UpdateWorkerDeploymentComputeConfigRequest.
Changed the ComputeProvider structure from the single Payload field to using a oneOf with either a google.protobuf.Any (for unencrypted) or temporal.api.common.v1.Payload (for encrypted) provider details.

Please let me know what you think.

jaypipes · 2026-01-29T17:24:38Z

temporal/api/deployment/v1/compute_provider.proto

+// Function via its ARN. This will start the Worker that is housed within that
+// AWS Lambda Function and the Worker will begin listening to Tasks on the
+// WorkerDeployment's configured TaskQueue.
+message ComputeProviderAWSLambda {


@cretz I don't think putting scaling config into ComputeProvider is appropriate. The scaling algorithm for how often a StartWorker API is called is orthogonal to configuration for how that StartWorker code is implemented for a particular compute platform.

I've added a different Scaler field to WorkerDeploymentVersionInfo that is used to store settings related to auto-scaling algorithms.

jaypipes · 2026-01-30T21:46:28Z

temporal/api/compute/v1/provider.proto

+message ComputeProviderAWSLambda {
+    // The Qualified or Unqualified ARN of the AWS Lambda Function to be
+    // invoked.
+    string invoke = 1;


@dnr I've updated the comment above. It can actually be the function URL too :) Which is why I left the type suffix off the field name.

jaypipes · 2026-01-30T21:47:21Z

temporal/api/deployment/v1/compute_provider.proto

+    // An individual ComputeProviderAWSLambda,
+    // ComputeProviderGoogleCloudRun, etc will be encrypted and
+    // serialized into compute_provider_config.
+    temporal.api.common.v1.Payload provider_details = 1;


@cretz I've given it another try... please see the updated struct.

cretz · 2026-01-30T22:34:15Z

temporal/api/deployment/v1/compute_provider.proto

+        google.protobuf.Any object = 2;
+        // will be an encrypted, encoded bytestring containing
+        // provider-specific information. The implementation must understand
+        // how to decrypt the payload.
+        temporal.api.common.v1.Payload payload = 3;


Pedantic, but would qualify these field names (e.g. provider_detail_any and provider_detail_payload), they are not implicitly qualified by their oneof parent in many cases

cretz · 2026-01-30T22:35:52Z

temporal/api/deployment/v1/compute_provider.proto

+// Function via its ARN. This will start the Worker that is housed within that
+// AWS Lambda Function and the Worker will begin listening to Tasks on the
+// WorkerDeployment's configured TaskQueue.
+message ComputeProviderDetailAWSLambda {


I do not believe this should be defined as proto but rather JSON schema or some other form. There is no server API use of this value, it should be blindly passed to the compute provider, and those consuming compute provider details should not have to know anything about our API or have to run protoc to consume this (or future specific provider detail models).

cretz · 2026-01-30T22:36:45Z

temporal/api/workflowservice/v1/request_response.proto

+    // Contains the new worker compute provider configuration for the
+    // WorkerDeploymentVersion.
+	temporal.api.deployment.v1.ComputeProvider compute_provider = 6;
+
+    // Contains the new scaler configuration for the WorkerDeploymentVersion.
+	temporal.api.deployment.v1.Scaler scaler = 7;


Mentioned as reply to other comment, but if we want compute_provider + scaler to be in something called "compute config" (assuming based on request name), IMO we should explicit model such a thing.

cretz · 2026-01-30T22:57:06Z

temporal/api/compute/v1/scaler.proto

+    int32 min_replicas = 1;
+    // The upper limit for the number of Workers (in the WorkerDeployment) to
+    // which the Scaler can scale up. Must be greater than or equal to
+    // minReplicas.
+    int32 max_replicas = 2;


Has it been decided that these are the knobs we want to expose for controlling scaling, or are these just PoC-style placeholders? (trying to confirm whether I should comment on our scaling knobs yet or not)

@cretz POC style placeholders for now :)

Introduces a new `compute.v1` package of proto messages that house the various instructions and information needed by worker control plane controllers `temporal.api.compute.v1.Config` has two fields, `provider` and `scaler` that store the configuration information for the server-owned worker setup and how the worker control plane should treat worker lifecycle events (such as when no pollers exist for a task queue that has received a task). The `temporal.api.compute.v1.Config.provider` field is of type `temporal.api.compute.v1.Provider`. This message is a simple wrapper around a blob of bytes that contains configuration settings for a worker control plane controller. The configuration settings for the worker control plane controller are specific to different compute providers. The first compute provider we are modeling is AWS Lambda, and the `ProviderDetailAWSLambda` message houses the ARN of the AWS Lambda Function to be invoked and an optional ARN of the IAM Role that should be assumed by a worker control plane controller when invoking the function. The `temporal.api.compute.v1.Config.scaler` field is of type `temporal.api.compute.v1.Scaler`. This message contains instructions for when the worker lifecycle controller should scale up or down the number of workers in a WorkerDeployment or WorkerDeploymentVersion. Adds a new field `compute_config` to `WorkerDeploymentInfo` and `WorkerDeploymentVersionInfo` that contains a `temporal.api.compute.v1.Config` message. Both WorkerDeployment and WorkerDeploymentVersion have this field in order to handle versioned workers properly. For example, if the user wants to use PINNED temporal versioning, they will be creating a WorkerDeploymentVersion that is specific to, say, a qualified Lambda Function ARN. Therefore we need to place compute provider configuration on a WorkerDeploymentVersion in addition to WorkerDeployment. Adds three new gRPC API methods to WorkflowService: * `CreateWorkerDeployment` creates a new WorkerDeployment containing the supplied compute configuration. * `UpdateWorkerDeploymentComputeConfig` updates an existing WorkerDeployment's compute configuration. * `UpdateWorkerDeploymentComputeConfig` updates an existing WorkerDeploymentVersion's compute configuration. Signed-off-by: Jay Pipes <jay.pipes@temporal.io>

cretz

Looks great to me, mostly only pedantic things, but structure looks great (can't approve for obvious reasons, we don't want to merge this anytime soon)

cretz · 2026-02-03T16:54:37Z

temporal/api/compute/v1/provider.proto

+// Lambda Function via its ARN. This will start the Worker that is housed
+// within that AWS Lambda Function and the Worker will begin listening to Tasks
+// on the WorkerDeployment's configured TaskQueue.
+message ProviderDetailAWSLambda {


Mentioned in other comment at #704 (comment), but I think this should not be a (orphaned) proto, but rather defined in a more generic way that doesn't require protobuf/protoc to consume. If we don't want to define JSON schema currently, even just a Go struct w/ JSON tags and a document somewhere saying the definition of this JSON object I think would be good enough. But would like to not marry protobuf here.

Also, somewhere we need to define this compute provider service and what it may accept (but it won't be in this repo).

cretz · 2026-02-03T16:55:49Z

temporal/api/compute/v1/scaler.proto

+message Scaler {
+    // The lower limit for the number of Workers (in the WorkerDeployment) to
+    // which the Scaler can scale down.
+    int32 min_replicas = 1;


Pedantic a bit, but IMO this should be called min_instances. I am worried it may get confused with the use of the term "replicas" by some providers where I could see an instance wanting many replicas (but from server POV, it's an "instance" w/ an "instance key").

cretz · 2026-02-03T16:56:19Z

temporal/api/compute/v1/config.proto

Pedantic, but can just put everything in this proto namespace in the message.proto file, not a ton of benefit in the separation IMO

cretz · 2026-02-03T17:01:11Z

temporal/api/compute/v1/provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message Provider {


This is also going to need to be configured with the target Nexus/implementation endpoint. For Lambda use on our built-in implementation, this can be a built-in "system" string or something, but users will need to be able to provide their own implementation endpoint.

IMO a simple nexus_endpoint field would be great. Don't even need nexus_service because it should be a well-known, pre-defined service that users (or us) would implement. Personally I would like this to be non-optional and we can set a certain system endpoint for wherever we host ours. For example, it is no problem for users to host our Lambda compute provider impl themselves on their side with their endpoint and their end-to-end encryption and such.

I would be partial to an address field outside the details. That said, no fields in here should be nexus specific (as it is just one of multiple avenues)

cretz · 2026-02-03T17:27:17Z

temporal/api/workflowservice/v1/request_response.proto

 }

+// Creates a new WorkerDeployment.
+message CreateWorkerDeploymentRequest {


Question - can I see the user experience of configuring compute config for workflows only (not activities) on task queue X? Specifically, can worker deployments be independent by task queue type? If they can't, that's probably ok, but we will want to make sure the worker instances the server tells the provider to spin up are by task queue type, and we arguably should be able to provide unique configuration by task queue type (or at least disallow other task queue types from appearing on the task queue at all).

dnr · 2026-02-03T21:08:40Z

we will be making the UX for unversioned transparently auto-create a "latest" version when one doesn't already exist

If we do that, why do we need a config at the deployment level at all? Can't it just be on that implicit "latest" version?

02strich · 2026-02-04T21:15:00Z

temporal/api/compute/v1/provider.proto

+// on a TaskQueue that has no active pollers in the last five minutes, a
+// serverless worker lifecycle controller might need to invoke an AWS Lambda
+// Function that itself ends up calling the SDK's worker.New() function.
+message Provider {


I would be partial to an address field outside the details. That said, no fields in here should be nexus specific (as it is just one of multiple avenues)

02strich · 2026-02-04T21:16:54Z

temporal/api/compute/v1/provider.proto

+message ComputeProviderAWSLambda {
+    // The Qualified or Unqualified ARN of the AWS Lambda Function to be
+    // invoked.
+    string invoke = 1;


@jaypipes this needs to be the ARN - no plan for supporting function URLs at this point.

02strich · 2026-02-04T21:17:38Z

temporal/api/deployment/v1/message.proto

+    // Contains information used by worker control plane controllers to handle
+    // scale events.
+	temporal.api.compute.v1.Config compute_config = 20;


when would this be used?

jaypipes requested review from a team as code owners January 28, 2026 15:30

jaypipes force-pushed the jaypipes/compute-provider branch 3 times, most recently from 2efdaa8 to 5a5db55 Compare January 28, 2026 17:10

02strich reviewed Jan 28, 2026

View reviewed changes

cretz reviewed Jan 28, 2026

View reviewed changes

dnr reviewed Jan 29, 2026

View reviewed changes

cretz reviewed Jan 29, 2026

View reviewed changes

jaypipes force-pushed the jaypipes/compute-provider branch from 5a5db55 to add3118 Compare January 30, 2026 21:41

jaypipes commented Jan 30, 2026

View reviewed changes

jaypipes force-pushed the jaypipes/compute-provider branch from add3118 to 21d4b1d Compare January 30, 2026 22:30

cretz reviewed Jan 30, 2026

View reviewed changes

jaypipes force-pushed the jaypipes/compute-provider branch from 21d4b1d to cb31098 Compare February 3, 2026 16:49

jaypipes force-pushed the jaypipes/compute-provider branch from cb31098 to a09df28 Compare February 3, 2026 16:57

cretz reviewed Feb 3, 2026

View reviewed changes

02strich reviewed Feb 4, 2026

View reviewed changes

	message ComputeProviderAWSLambda {
	message ComputeProviderDetailsAWSLambda {

		// Creates or replaces the compute provider configuration for a Worker
		// Deployment.

-    temporal.api.common.v1.Payload provider_details = 1;
+    // Passed to provider Nexus endpoint to identify the contents of the provider details
+    string provider_type = 1;
+    oneof provider_details {
+        // Non-encoded provider details, used in situations where decoding cannot occur
+        bytes provider_details_open = 2;
+        // Encoded provider details
+        temporal.api.common.v1.Payload provider_details_payload = 3;
+    }

Conversation

jaypipes commented Jan 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

CLAassistant commented Jan 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz Jan 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz Jan 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cretz Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jaypipes left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

jaypipes commented Jan 28, 2026 •

edited

Loading

CLAassistant commented Jan 28, 2026 •

edited

Loading

cretz Jan 29, 2026 •

edited

Loading

cretz left a comment •

edited

Loading

cretz Jan 30, 2026 •

edited

Loading

cretz Jan 28, 2026 •

edited

Loading

cretz Jan 29, 2026 •

edited

Loading

cretz Jan 29, 2026 •

edited

Loading

jaypipes left a comment •

edited

Loading

cretz Feb 3, 2026 •

edited

Loading

cretz Feb 3, 2026 •

edited

Loading

cretz Feb 3, 2026 •

edited

Loading