enhancement: add APM Stats transform #1095

rayz · 2026-01-09T14:59:25Z

Summary

This PR adds a new transform apm_stats that transforms Trace events into TraceStat events.
The proto file for DDSketch from sketch-go has also been added.

Change Type

Bug fix
New feature
Non-functional (chore, refactoring, docs)
Performance

How did you test this PR?

Claude to copy the unit tests in the datadog-agent.

References

rayz · 2026-01-09T15:00:00Z

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

This stack of pull requests is managed by Graphite. Learn more about stacking.

Copilot

Pull request overview

This PR implements an APM Stats transform that aggregates trace data into time-bucketed statistics and emits TraceStats events. The implementation includes span weight calculation, raw bucket aggregation, and a span concentrator that manages time-bucketed stats with configurable peer tag aggregation and span kind filtering.

Changes:

Added APM Stats transform with time-bucketed aggregation of trace spans
Implemented span weight calculation based on sampling rate
Added support for peer tags aggregation and span kind-based stats computation
Integrated DDSketch protocol buffers for histogram distributions

Reviewed changes

Copilot reviewed 18 out of 19 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
lib/saluki-core/src/data_model/event/trace_stats/mod.rs	Added mutable accessor for grouped stats in ClientStatsBucket
lib/saluki-core/src/data_model/event/mod.rs	Added TraceStats size logging to tests
lib/saluki-components/src/transforms/mod.rs	Registered ApmStatsConfiguration transform
lib/saluki-components/src/transforms/apm_stats/weight.rs	Implemented span weight calculation based on sampling rate
lib/saluki-components/src/transforms/apm_stats/statsraw.rs	Implemented raw bucket aggregation and stats export
lib/saluki-components/src/transforms/apm_stats/span_concentrator.rs	Implemented span concentrator with time-bucketed aggregation
lib/saluki-components/src/transforms/apm_stats/mod.rs	Implemented main APM Stats transform with async flush logic
lib/saluki-components/src/transforms/apm_stats/aggregation.rs	Defined aggregation keys and helper functions for stats computation
lib/saluki-components/src/common/otlp/config.rs	Changed default for enable_otlp_compute_top_level_by_span_kind to true
lib/saluki-components/src/common/datadog/apm.rs	Extended ApmConfig with stats computation and peer tags configuration
lib/saluki-components/Cargo.toml	Added fnv dependency for hashing
lib/protos/datadog/src/serde.rs	Added serialization helpers for protocol buffer bytes
lib/protos/datadog/src/lib.rs	Added sketches module for DDSketch definitions
lib/protos/datadog/proto/sketches-go/ddsketch/pb/ddsketch.proto	Added DDSketch protocol buffer definition
lib/protos/datadog/build.rs	Added TYPE_BYTES handling and sketch proto compilation
lib/protos/datadog/Cargo.toml	Added serde_bytes dependency
lib/ddsketch-agent/src/lib.rs	Implemented to_proto conversion for DDSketch
Cargo.toml	Added serde_bytes workspace dependency

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-09T15:00:19Z

lib/saluki-components/src/transforms/apm_stats/statsraw.rs

+fn round(f: f64) -> u64 {
+    let i = f as u64;
+    let frac = f - (i as f64);
+    if rand::rng().random::<f64>() < frac {


The rand::rng() function uses thread-local RNG which is not Send. This could cause issues in async contexts where the transform may be moved across threads. Consider using a thread-safe RNG or storing an RNG instance in the struct.

lib/protos/datadog/src/lib.rs

pr-commenter · 2026-01-09T15:06:07Z

Binary Size Analysis (Agent Data Plane)

Target: 02eac16 (baseline) vs f6ec3e0 (comparison) diff
Baseline Size: 357.34 MiB
Comparison Size: 359.14 MiB
Size Change: +1.80 MiB (+0.50%)
Pass/Fail Threshold: +5%
Result: PASSED ✅

Changes by Module

Module	File Size	Symbols
[debug sections]	+1.45 MiB	7
figment	+252.97 KiB	584
otlp_protos::otlp_include::opentelemetry	+32.96 KiB	278
prost	-31.12 KiB	498
anyhow	+26.34 KiB	1394
hyper_util	-22.89 KiB	200
saluki_env::workload::providers	+20.68 KiB	21
saluki_env::workload::collectors	-18.93 KiB	235
alloc	+15.06 KiB	1736
[sections]	+14.50 KiB	8
hyper	+13.89 KiB	603
serde	+13.54 KiB	40
saluki_core::data_model::event	+12.26 KiB	73
serde_json	+11.79 KiB	177
saluki_components::common::datadog	-11.42 KiB	322
anon.c1ad624bdfe951ca673c256fb61fa178.119.llvm.13580561373593502968	+9.79 KiB	1
anon.f30def023657add84aa3a6d7d4197fa6.35.llvm.13640408623153174339	-9.79 KiB	1
tower_layer	+9.15 KiB	10
saluki_common::cache::Cache<K,V,W,H>	-8.80 KiB	37
tokio	-8.61 KiB	4137

Detailed Symbol Changes

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  +0.9%  +885Ki  [ = ]       0    [section .debug_loc]
  +0.3%  +340Ki  [ = ]       0    [section .debug_info]
  +1.6%  +322Ki  +1.7%  +235Ki    [37861 Others]
  [NEW]  +163Ki  [NEW]  +162Ki    agent_data_plane::cli::run::handle_run_command::_{{closure}}::hda70ed1a89fd2d1a
  [NEW]  +104Ki  [NEW]  +104Ki    agent_data_plane::main::_{{closure}}::hdbdeb3f696b83878
  +0.2%  +103Ki  [ = ]       0    [section .debug_str]
  +0.4% +89.1Ki  [ = ]       0    [section .debug_line]
  [NEW] +80.1Ki  [NEW] +80.0Ki    agent_data_plane::cli::debug::handle_debug_command::_{{closure}}::h46cb9932ae2c93e5
  [NEW] +61.5Ki  [NEW] +61.4Ki    saluki_core::topology::built::BuiltTopology::spawn::_{{closure}}::h3eea7ae28e7b9fba
  +0.2% +58.8Ki  [ = ]       0    [section .debug_ranges]
  [NEW] +56.0Ki  [NEW] +55.8Ki    saluki_core::topology::blueprint::TopologyBlueprint::build::_{{closure}}::hd9ac07e3f1b31a34
  [NEW] +46.7Ki  [NEW] +46.5Ki    _<saluki_components::forwarders::otlp::OtlpForwarder as saluki_core::components::forwarders::Forwarder>::run::_{{closure}}::hda328cfd5d7c0653
  [NEW] +46.0Ki  [NEW] +45.8Ki    saluki_config::ConfigurationLoader::with_default_secrets_resolution::_{{closure}}::h001afafded53abf5
  [NEW] +44.2Ki  [NEW] +44.1Ki    saluki_components::sources::otlp::metrics::translator::OtlpMetricsTranslator::map_metrics::h4d518cfa8c148b47
  [DEL] -42.7Ki  [DEL] -42.6Ki    saluki_components::sources::otlp::metrics::translator::OtlpMetricsTranslator::map_metrics::hd0be7a53ec163d20
  [DEL] -45.7Ki  [DEL] -45.5Ki    saluki_config::ConfigurationLoader::with_default_secrets_resolution::_{{closure}}::hf71d34d965f0471b
  [DEL] -56.1Ki  [DEL] -56.0Ki    saluki_core::topology::blueprint::TopologyBlueprint::build::_{{closure}}::h7cd9eeae8f1c3216
  [DEL] -61.7Ki  [DEL] -61.5Ki    saluki_core::topology::built::BuiltTopology::spawn::_{{closure}}::h867a99b8656ff505
  [DEL] -80.1Ki  [DEL] -79.9Ki    agent_data_plane::cli::debug::handle_debug_command::_{{closure}}::hc359c1c67927acca
  [DEL]  -105Ki  [DEL]  -104Ki    agent_data_plane::main::_{{closure}}::h94428d325095caca
  [DEL]  -166Ki  [DEL]  -166Ki    agent_data_plane::cli::run::handle_run_command::_{{closure}}::h3ba317381d86c2d5
  +0.5% +1.80Mi  +0.9%  +279Ki    TOTAL

pr-commenter · 2026-01-09T15:36:59Z

Regression Detector (Agent Data Plane)

Regression Detector Results

Run ID: dcd599bc-3ec2-4317-a831-02029e32bdd6

Baseline: 02eac16
Comparison: 3f5b868
Diff

❌ Experiments with retried target crashes

This is a critical error. One or more replicates failed with a non-zero exit code. These replicates may have been retried. See Replicate Execution Details for more information.

otlp_ingest_metrics_adp

Optimization Goals: ✅ Improvement(s) detected

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
✅	otlp_ingest_logs_adp	memory utilization	-6.06	[-6.49, -5.62]	1	(metrics) (profiles) (logs)

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	otlp_ingest_metrics_adp	memory utilization	+0.45	[+0.28, +0.61]	1
➖	quality_gates_rss_dsd_heavy	memory utilization	+0.36	[+0.22, +0.50]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_idle	memory utilization	+0.19	[+0.16, +0.23]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_throughput	ingress throughput	+0.02	[-0.18, +0.22]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_throughput	ingress throughput	+0.01	[-0.04, +0.05]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_throughput	ingress throughput	+0.00	[-0.05, +0.06]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_throughput	ingress throughput	+0.00	[-0.02, +0.02]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_ultraheavy	ingress throughput	-0.01	[-0.08, +0.06]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_low	memory utilization	-0.24	[-0.37, -0.11]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_medium	memory utilization	-0.52	[-0.68, -0.35]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_throughput	ingress throughput	-2.22	[-2.35, -2.09]	1	(metrics) (profiles) (logs)
✅	otlp_ingest_logs_adp	memory utilization	-6.06	[-6.49, -5.62]	1	(metrics) (profiles) (logs)

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	links
✅	quality_gates_rss_dsd_heavy	memory_usage	10/10	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_low	memory_usage	10/10	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_medium	memory_usage	10/10	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_ultraheavy	memory_usage	10/10	(metrics) (profiles) (logs)
✅	quality_gates_rss_idle	memory_usage	10/10	(metrics) (profiles) (logs)

Explanation

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

✅ = significantly better comparison variant performance
❌ = significantly worse comparison variant performance
➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.
Its configuration does not mark it "erratic".

Replicate Execution Details

We run multiple replicates for each experiment/variant. However, we allow replicates to be automatically retried if there are any failures, up to 8 times, at which point the replicate is marked dead and we are unable to run analysis for the entire experiment. We call each of these attempts at running replicates a replicate execution. This section lists all replicate executions that failed due to the target crashing or being oom killed.

Note: In the below tables we bucket failures by experiment, variant, and failure type. For each of these buckets we list out the replicate indexes that failed with an annotation signifying how many times said replicate failed with the given failure mode. In the below example the baseline variant of the experiment named experiment_with_failures had two replicates that failed by oom kills. Replicate 0, which failed 8 executions, and replicate 1 which failed 6 executions, all with the same failure mode.

Experiment	Variant	Replicates	Failure	Logs	Debug Dashboard
experiment_with_failures	baseline	0 (x8) 1 (x6)	Oom killed		Debug Dashboard

The debug dashboard links will take you to a debugging dashboard specifically designed to investigate replicate execution failures.

❌ Retried Normal Replicate Execution Failures (non-profiling)

Experiment	Variant	Replicates	Failure	Debug Dashboard
otlp_ingest_metrics_adp	baseline	6, 3	Failed to shutdown when requested	Debug Dashboard

tobz · 2026-01-09T16:17:22Z

lib/protos/datadog/proto/sketches-go/ddsketch/pb/ddsketch.proto

@@ -0,0 +1,54 @@
+// Copyright 2021 Datadog, Inc.


Just saying this out loud so I (hopefully) don't forget: we should really pull this in via update-protos.sh so we have a known way to update. This is non-blocking, though, and we shouldn't do it here.

(I realize the proto definition probably hasn't changed in years, but I prefer that we have clear provenance / ways to bootstrap things if necessary.)

Oops, i forgot we had that 😅

tobz · 2026-01-09T16:18:44Z

lib/saluki-components/src/common/datadog/apm.rs

+    #[allow(dead_code)]
+    compute_stats_by_span_kind: bool,
+
+    /// Enables aggregation of peer related tags (e.g., `peer.service`, `db.instance`, etc.) in the Agent.
+    ///
+    /// Defaults to `true`.
+    #[serde(default = "default_peer_tags_aggregation")]
+    #[allow(dead_code)]
+    peer_tags_aggregation: bool,
+
+    /// Optional list of supplementary peer tags that go beyond the defaults. The Datadog backend validates all tags
+    /// and will drop ones that are unapproved.
+    ///
+    /// Defaults to an empty list.
+    #[serde(default)]
+    #[allow(dead_code)]


Why are these all marked as potentially being dead code? 🤔

Had them to pass clippy. Removed

tobz · 2026-01-09T16:19:37Z

lib/saluki-components/src/common/datadog/apm.rs

+
+        // Get hostname from global config if not set in apm_config
+        if apm_config.hostname.is_empty() {
+            apm_config.hostname = config.get_typed_or_default::<String>("hostname").into();


We should be pulling this from the environment provider like, say, the Host Enrichment transform does it. However, given that this is just a common configuration type... not sure how you want to structure that, like maybe just having the transform grab the detected hostname and then set it on the APM config with a dedicated method, like set_hostname_if_empty or something.

lib/saluki-components/src/common/datadog/apm.rs

lib/saluki-components/src/transforms/apm_stats/aggregation.rs

tobz · 2026-01-09T16:25:01Z

lib/saluki-components/src/transforms/apm_stats/aggregation.rs

+            let upper = normalized.to_uppercase();
+
+            if let Some(code) = grpc_status_name_to_code(&upper) {
+                return MetaString::from(code);


Use MetaString::from_static here to avoid allocating.

tobz · 2026-01-09T16:26:08Z

lib/saluki-components/src/transforms/apm_stats/aggregation.rs

+                return MetaString::from(code);
+            }
+
+            return MetaString::default();


Is the intent to actually give up here if we do find a value for one of these four attribute keys but it has a non-integer/non-gRPC-status-code value?

Yup, agent code

tobz · 2026-01-09T16:29:20Z

lib/saluki-components/src/transforms/apm_stats/aggregation.rs

+            }
+
+            if let Ok(code) = str_value.parse::<u64>() {
+                return MetaString::from(code.to_string());


We could just clone str_value here: the parse will fail unless the value is an integer without any leading/trailing/interstitial spaces... so we know if it parses, then the original string is in the same format that it will get rendered as by calling to_string.

tobz · 2026-01-09T16:31:48Z

lib/saluki-components/src/transforms/apm_stats/span_concentrator.rs

+    bsize: i64,
+
+    /// Timestamp of oldest allowed bucket (prevents stats for already-flushed buckets)
+    oldest_ts: i64,
+
+    /// Number of buckets to buffer before flushing
+    buffer_len: i64,
+
+    /// Time-bucketed raw stats: bucket_timestamp -> RawBucket
+    buckets: FastHashMap<i64, RawBucket>,


I don't know why the original code does this with i64, but let's make these u64 and convert from i64 to u64 at the boundary when the span is added to the concentrator.

A negative timestamp would be a big red flag anyways, so we should surface those if they're actually making it through.

tobz · 2026-01-09T16:35:00Z

lib/saluki-components/src/transforms/apm_stats/statsraw.rs

+fn ns_timestamp_to_float(ns: i64) -> f64 {
+    let f = ns as f64;
+    let bits = f.to_bits();
+    let truncated_bits = bits & 0xffff_f800_0000_0000;
+    f64::from_bits(truncated_bits)
+}


What's going on here? 🤔

Taken from here 😅 . I'll copy the comments

Copilot

Pull request overview

Copilot reviewed 21 out of 22 changed files in this pull request and generated 4 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

lib/protos/datadog/proto/sketches-go/ddsketch/pb/ddsketch.proto

lib/saluki-components/src/transforms/apm_stats/statsraw.rs

Copilot · 2026-01-09T18:49:18Z

lib/saluki-components/src/transforms/apm_stats/mod.rs

+        builder.minimum().with_single_value::<ApmStats>("component struct");
+        // TODO: Think about everything we need to account for here.


This TODO indicates incomplete memory accounting for the ApmStats component. Memory bounds should be properly specified to ensure accurate resource tracking, especially for production workloads where the concentrator maintains time-bucketed data.

Suggested change

builder.minimum().with_single_value::<ApmStats>("component struct");

// TODO: Think about everything we need to account for here.

builder

.minimum()

.with_single_value::<ApmStats>("component struct")

.with_single_value::<SpanConcentrator>("span concentrator")

.with_single_value::<MetaString>("agent env")

.with_single_value::<MetaString>("agent hostname");

lib/saluki-components/src/common/datadog/apm.rs

Copilot

Pull request overview

Copilot reviewed 24 out of 25 changed files in this pull request and generated 3 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-09T20:24:38Z

lib/saluki-components/src/transforms/apm_stats/statsraw.rs

+fn round(f: f64) -> u64 {
+    let i = f as u64;
+    let frac = f - (i as f64);
+    if rand::rng().random::<f64>() < frac {
+        i + 1
+    } else {
+        i
+    }
+}


The function name round is misleading as it implements stochastic/probabilistic rounding rather than standard mathematical rounding. Consider renaming to stochastic_round or probabilistic_round to clearly indicate its behavior.

lib/protos/datadog/build.rs

lib/saluki-components/src/common/datadog/apm.rs

Copilot

Pull request overview

Copilot reviewed 25 out of 26 changed files in this pull request and generated no new comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot AI review requested due to automatic review settings January 9, 2026 14:59

github-actions bot added area/core Core functionality, event model, etc. area/components Sources, transforms, and destinations. labels Jan 9, 2026

rayz mentioned this pull request Jan 9, 2026

chore: add TraceStats event type #1092

Open

4 tasks

Copilot AI reviewed Jan 9, 2026

View reviewed changes

rayz mentioned this pull request Jan 9, 2026

APM Stats Encode #1096

Draft

4 tasks

rayz changed the title ~~APM Stats transform~~ enhancement: add APM Stats transform Jan 9, 2026

APM Stats transform

b94e04f

rayz force-pushed the rayz/apm-stats-event branch from 53b3e3c to e9accbf Compare January 9, 2026 16:11

rayz force-pushed the rayz/apm-stats-transform branch from 473b8c1 to b94e04f Compare January 9, 2026 16:11

tobz requested changes Jan 9, 2026

View reviewed changes

Update make update-protos

7fedfa9

Copilot AI review requested due to automatic review settings January 9, 2026 18:48

rayz mentioned this pull request Jan 9, 2026

enhancement(correctness): add structural diffing for APM stats in trace analysis mode #1097

Merged

4 tasks

Copilot AI reviewed Jan 9, 2026

View reviewed changes

rayz added 2 commits January 9, 2026 15:19

Feedback, add env + workload provider

31a4ae7

Copilot feedback

f59bd77

Copilot AI review requested due to automatic review settings January 9, 2026 20:23

Copilot AI reviewed Jan 9, 2026

View reviewed changes

rayz added 2 commits January 9, 2026 15:27

Doc

3f5b868

License

f6ec3e0

Copilot AI review requested due to automatic review settings January 9, 2026 20:31

Copilot AI reviewed Jan 9, 2026

View reviewed changes

		builder.minimum().with_single_value::<ApmStats>("component struct");
		// TODO: Think about everything we need to account for here.

enhancement: add APM Stats transform #1095

Are you sure you want to change the base?

enhancement: add APM Stats transform #1095

Uh oh!

Conversation

rayz commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Change Type

How did you test this PR?

References

Uh oh!

rayz commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

pr-commenter bot commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Binary Size Analysis (Agent Data Plane)

Changes by Module

Detailed Symbol Changes

Uh oh!

pr-commenter bot commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Regression Detector (Agent Data Plane)

Regression Detector Results

❌ Experiments with retried target crashes

Optimization Goals: ✅ Improvement(s) detected

Fine details of change detection per experiment

Bounds Checks: ✅ Passed

Explanation

Replicate Execution Details

❌ Retried Normal Replicate Execution Failures (non-profiling)

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!

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!

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!

rayz Jan 9, 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!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

rayz commented Jan 9, 2026 •

edited

Loading

rayz commented Jan 9, 2026 •

edited

Loading

pr-commenter bot commented Jan 9, 2026 •

edited

Loading

pr-commenter bot commented Jan 9, 2026 •

edited

Loading

rayz Jan 9, 2026 •

edited

Loading