Fix resuming older experiments

pyproject.toml

@@ -22,7 +22,7 @@ classifiers = [
 
 dependencies = [
     "blake3",
-    "lightning-sdk >=2026.03.31",
+    "lightning-sdk >=2026.04.16",
     "lightning-utilities<=0.15.3",
     "protobuf",
     "psutil",

src/litlogger/api/metrics_api.py

@@ -27,6 +27,7 @@
     V1PhaseType,
     V1SystemInfo,
 )
+from lightning_sdk.lightning_cloud.openapi.rest import ApiException
 
 from litlogger.api.client import LitRestClient
 from litlogger.colors import _create_colors
@@ -103,18 +104,12 @@ def get_experiment_metrics_by_name(
         Returns:
             The metrics stream object for the experiment, or None if not found.
         """
-        response = self.client.lit_logger_service_list_metrics_streams(project_id=teamspace_id)
-
-        if not response.metrics_streams:
-            return None
-
-        # Filter by name
-        matching = [ms for ms in response.metrics_streams if ms.name == name]
-
-        if not matching:
-            return None
-
-        return matching[0]
+        try:
+            return self.client.lit_logger_service_get_metrics_stream(project_id=teamspace_id, name=name)
+        except ApiException as ex:
+            if ex.status == 404:
+                return None
+            raise
 
     def get_or_create_experiment_metrics(
         self,

tests/integrations/test_standalone_lifecycle.py

@@ -3,6 +3,7 @@
 Covers error handling, sequential experiments, resume, console output, and API internals.
 """
 
+import os
 import uuid
 from contextlib import redirect_stderr
 from io import StringIO
@@ -254,6 +255,48 @@ def test_resume_experiment():
     )
 
 
+@pytest.mark.cloud()
+@pytest.mark.skipif(bool(os.environ.get("TEST_GUEST_MODE", "")), reason="guests can only create up to 25 experiments")
+def test_resume_old_experiment():
+    """Resume an experiment that has many newer experiments in the same teamspace."""
+    from litlogger.api.metrics_api import MetricsApi
+
+    target_name = f"standalone_resume_old-{uuid.uuid4().hex}"
+    filler_prefix = f"standalone_resume_old_filler-{uuid.uuid4().hex}"
+
+    exp1 = litlogger.init(name=target_name, teamspace="oss-litlogger")
+    litlogger.log_metrics({"loss": 0.5}, step=0)
+    litlogger.finalize()
+
+    project_id = exp1._teamspace.id
+    target_stream_id = exp1._metrics_store.id
+
+    api = MetricsApi()
+    client = LitRestClient()
+
+    filler_ids: list[str] = []
+    try:
+        for i in range(51):
+            stream = api.create_experiment_metrics(
+                teamspace_id=project_id,
+                name=f"{filler_prefix}-{i}",
+            )
+            filler_ids.append(stream.id)
+
+        exp2 = litlogger.init(name=target_name, teamspace="oss-litlogger")
+        try:
+            assert (
+                exp2._metrics_store.id == target_stream_id
+            ), "Expected to resume the original experiment, got a new stream"
+        finally:
+            litlogger.finalize()
+    finally:
+        client.lit_logger_service_delete_metrics_stream(
+            project_id=project_id,
+            body=LitLoggerServiceDeleteMetricsStreamBody(ids=[target_stream_id, *filler_ids]),
+        )
+
+
 @pytest.mark.cloud()
 def test_new_dict_api_resume():
     """Test resuming an experiment with the new dict-like API."""

tests/unittests/api/test_metrics_api.py

@@ -6,11 +6,13 @@
 
 from unittest.mock import MagicMock, patch
 
+import pytest
 from lightning_sdk.lightning_cloud.openapi import (
     V1Metrics,
     V1MetricValue,
     V1PhaseType,
 )
+from lightning_sdk.lightning_cloud.openapi.rest import ApiException
 from litlogger.api.metrics_api import MetricsApi
 from litlogger.types import PhaseType
 
@@ -30,29 +32,25 @@ def test_get_experiment_metrics_by_name(self):
         mock_stream = MagicMock()
         mock_stream.id = "ms-123"
         mock_stream.name = "my-experiment"
-        mock_stream.version_number = 1
-        mock_response = MagicMock()
-        mock_response.metrics_streams = [mock_stream]
-        mock_client.lit_logger_service_list_metrics_streams.return_value = mock_response
+        mock_client.lit_logger_service_get_metrics_stream.return_value = mock_stream
         api = MetricsApi(client=mock_client)
 
         result = api.get_experiment_metrics_by_name(
             teamspace_id="ts-123",
             name="my-experiment",
         )
 
-        mock_client.lit_logger_service_list_metrics_streams.assert_called_once_with(
+        mock_client.lit_logger_service_get_metrics_stream.assert_called_once_with(
             project_id="ts-123",
+            name="my-experiment",
         )
         assert result.id == "ms-123"
         assert result.name == "my-experiment"
 
     def test_get_experiment_metrics_by_name_not_found(self):
         """Test that get_experiment_metrics_by_name returns None when experiment not found."""
         mock_client = MagicMock()
-        mock_response = MagicMock()
-        mock_response.metrics_streams = []
-        mock_client.lit_logger_service_list_metrics_streams.return_value = mock_response
+        mock_client.lit_logger_service_get_metrics_stream.side_effect = ApiException(status=404)
         api = MetricsApi(client=mock_client)
 
         result = api.get_experiment_metrics_by_name(
@@ -62,13 +60,23 @@ def test_get_experiment_metrics_by_name_not_found(self):
 
         assert result is None
 
+    def test_get_experiment_metrics_by_name_raises_on_other_errors(self):
+        """Test that non-404 errors from get_metrics_stream propagate."""
+        mock_client = MagicMock()
+        mock_client.lit_logger_service_get_metrics_stream.side_effect = ApiException(status=500)
+        api = MetricsApi(client=mock_client)
+
+        with pytest.raises(ApiException):
+            api.get_experiment_metrics_by_name(
+                teamspace_id="ts-123",
+                name="my-experiment",
+            )
+
     def test_get_or_create_experiment_metrics_creates_new(self):
         """Test get_or_create_experiment_metrics creates a new experiment when none exists."""
         mock_client = MagicMock()
-        # First call to list returns empty (no existing experiment)
-        mock_list_response = MagicMock()
-        mock_list_response.metrics_streams = []
-        mock_client.lit_logger_service_list_metrics_streams.return_value = mock_list_response
+        # Get returns 404 (no existing experiment)
+        mock_client.lit_logger_service_get_metrics_stream.side_effect = ApiException(status=404)
         # Create returns a new experiment
         mock_created = MagicMock()
         mock_created.id = "ms-new"
@@ -92,13 +100,11 @@ def test_get_or_create_experiment_metrics_creates_new(self):
     def test_get_or_create_experiment_metrics_returns_existing(self):
         """Test get_or_create_experiment_metrics returns existing experiment without creating."""
         mock_client = MagicMock()
-        # List returns an existing experiment
+        # Get returns an existing experiment
         mock_existing = MagicMock()
         mock_existing.id = "ms-existing"
         mock_existing.name = "my-experiment"
-        mock_list_response = MagicMock()
-        mock_list_response.metrics_streams = [mock_existing]
-        mock_client.lit_logger_service_list_metrics_streams.return_value = mock_list_response
+        mock_client.lit_logger_service_get_metrics_stream.return_value = mock_existing
         api = MetricsApi(client=mock_client)
 
         result, created = api.get_or_create_experiment_metrics(