aboutsummaryrefslogtreecommitdiffstats
path: root/metrics
diff options
context:
space:
mode:
authorBjørn Christian Seime <bjorncs@yahooinc.com>2023-08-04 13:33:13 +0200
committerBjørn Christian Seime <bjorncs@yahooinc.com>2023-08-04 13:52:27 +0200
commit3ddce3c158941cb8eabdb391d207fe004095c434 (patch)
treedb8121fe016fa7bb5a0ad345af9f1bd58fbbfb6d /metrics
parent0f46015e498ecb622473cd3e2403283c99f9f5d5 (diff)
Add generic metrics for embedders
Diffstat (limited to 'metrics')
-rw-r--r--metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java5
-rw-r--r--metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java4
2 files changed, 7 insertions, 2 deletions
diff --git a/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java b/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
index 4c7e203fee5..e515bbf9275 100644
--- a/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
+++ b/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
@@ -200,9 +200,10 @@ public enum ContainerMetrics implements VespaMetrics {
SERVER_TOTAL_FAILED_RESPONSE_LATENCY("serverTotalFailedResponseLatency", Unit.MILLISECOND, "Total duration for execution of failed responses"),
SERVER_TIME_TO_FIRST_BYTE("serverTimeToFirstByte", Unit.MILLISECOND, "Time from request has been received by the server until the first byte is returned to the client"),
- SERVER_STARTED_MILLIS("serverStartedMillis", Unit.MILLISECOND, "Time since the service was started");
-
+ SERVER_STARTED_MILLIS("serverStartedMillis", Unit.MILLISECOND, "Time since the service was started"),
+ EMBEDDER_LATENCY("embedder.latency", Unit.MILLISECOND, "Time spent creating an embedding"),
+ EMBEDDER_SEQUENCE_LENGTH("embedder.sequence_length", Unit.BYTE, "Size of sequence produced by tokenizer");
private final String name;
private final Unit unit;
diff --git a/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java b/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
index 4ec596f8ce7..b7ed7293d6c 100644
--- a/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
+++ b/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
@@ -108,6 +108,10 @@ public class VespaMetricSet {
// Routing layer metrics
addMetric(metrics, RoutingLayerMetrics.WORKER_CONNECTIONS.max()); // Hosted Vespa only (routing layer)
+ // Embedders
+ addMetric(metrics, ContainerMetrics.EMBEDDER_LATENCY, EnumSet.of(max, sum, count));
+ addMetric(metrics, ContainerMetrics.EMBEDDER_SEQUENCE_LENGTH, EnumSet.of(max, sum, count));
+
return metrics;
}