Adding support for multiple tags

docs/configuration.md

@@ -459,7 +459,7 @@ storage:
 | CSV column name | Description |
 |-----------------|-------------|
 | conversation_group_id | Conversation group id |
-| tag | Tag for grouping eval conversations |
+| tag | Tag(s) for grouping and filtering conversations (string or list of strings; stored as JSON array) |
 | turn_id | Turn id |
 | metric_identifier | Metric name |
 | result | Result -- PASS/FAIL/ERROR/SKIPPED |

src/lightspeed_evaluation/core/models/data.py

@@ -12,6 +12,21 @@
 logger = logging.getLogger(__name__)
 
 
+def _normalize_tag(v: str | list) -> list[str]:
+    """Normalize tag to list[str] for backward compatibility with single string."""
+    if isinstance(v, str):
+        items: list = [v]
+    else:
+        items = v
+    invalid = [item for item in items if not isinstance(item, str)]
+    if invalid:
+        raise ValueError(f"tag items must be strings, got: {invalid}")
+    result = [item for item in items if item.strip()]
+    if not result:
+        raise ValueError("tag must contain at least one non-empty string")
+    return result
+
+
 class ConversationMetadata(BaseModel):
     """Optional user-defined metadata for a conversation group."""
 
@@ -520,10 +535,10 @@ class EvaluationData(BaseModel):
         min_length=1,
         description="Optional description of the conversation group",
     )
-    tag: str = Field(
-        default="eval",
+    tag: list[str] = Field(
+        default=["eval"],
         min_length=1,
-        description="Tag for grouping and filtering conversations",
+        description="Tag(s) for grouping and filtering conversations",
     )
     skip: bool = Field(
         default=False,
@@ -586,6 +601,12 @@ def is_metric_invalid(self, metric: str) -> bool:
         """Returns True if the metric didn't pass the validation."""
         return metric in self._invalid_metrics
 
+    @field_validator("tag", mode="before")
+    @classmethod
+    def _validate_tag(cls, v: str | list) -> list[str]:
+        """Normalize tag to list[str] for backward compatibility with single string."""
+        return _normalize_tag(v)
+
     @field_validator("conversation_metrics")
     @classmethod
     def validate_conversation_metrics(
@@ -659,10 +680,10 @@ class EvaluationResult(MetricResult, StreamingMetricsMixin):
     conversation_group_id: str = Field(
         ..., min_length=1, description="Conversation group identifier"
     )
-    tag: str = Field(
-        default="eval",
+    tag: list[str] = Field(
+        default=["eval"],
         min_length=1,
-        description="Tag for grouping and filtering results",
+        description="Tag(s) for grouping and filtering results",
     )
     turn_id: Optional[str] = Field(
         default=None, description="Turn ID if turn-level evaluation"
@@ -714,6 +735,12 @@ class EvaluationResult(MetricResult, StreamingMetricsMixin):
         default=None, description="Expected tool calls formatted as string"
     )
 
+    @field_validator("tag", mode="before")
+    @classmethod
+    def _validate_tag(cls, v: str | list) -> list[str]:
+        """Normalize tag to list[str] for backward compatibility with single string."""
+        return _normalize_tag(v)
+
 
 class EvaluationScope(BaseModel):
     """Scope and parameters for metric evaluation."""

src/lightspeed_evaluation/core/output/statistics.py

@@ -222,7 +222,8 @@ def compute_tag_stats(
 
     grouped: dict[str, list[EvaluationResult]] = {}
     for r in results:
-        grouped.setdefault(r.tag, []).append(r)
+        for t in set(r.tag):
+            grouped.setdefault(t, []).append(r)
 
     tag_stats: dict[str, TagStats] = {}
     for tag in sorted(grouped):

src/lightspeed_evaluation/core/storage/langfuse_storage.py

@@ -22,7 +22,7 @@
 
 from __future__ import annotations
 
-import importlib
+import importlib.util
 import logging
 from typing import Any, Optional
 

src/lightspeed_evaluation/core/storage/sql_storage.py

@@ -47,7 +47,7 @@ class EvaluationResultDB(Base):  # pylint: disable=too-few-public-methods
     timestamp = Column(DateTime, nullable=False, index=True)
 
     conversation_group_id = Column(String(255), nullable=False, index=True)
-    tag = Column(String(100), nullable=True)
+    tag = Column(Text, nullable=True)
     turn_id = Column(String(100), nullable=True)
     metric_identifier = Column(String(255), nullable=False, index=True)
     metric_metadata = Column(Text, nullable=True)
@@ -309,7 +309,7 @@ def _result_to_db_record(self, result: EvaluationResult) -> EvaluationResultDB:
             run_id=self._run_info.run_id,
             timestamp=datetime.now(UTC),
             conversation_group_id=result.conversation_group_id,
-            tag=result.tag,
+            tag=json.dumps(result.tag),
             turn_id=result.turn_id,
             metric_identifier=result.metric_identifier,
             metric_metadata=result.metric_metadata,

src/lightspeed_evaluation/core/system/validator.py

@@ -408,7 +408,7 @@ def _filter_by_scope(
         filtered = [
             conv_data
             for conv_data in evaluation_data
-            if conv_data.tag in tag_set
+            if any(t in tag_set for t in conv_data.tag)
             or conv_data.conversation_group_id in conv_id_set
         ]
 

src/lightspeed_evaluation/pipeline/evaluation/errors.py

@@ -22,7 +22,7 @@ def _create_result(  # pylint: disable=too-many-arguments,too-many-positional-ar
         reason: str,
         result_status: str,
         *,
-        tag: str = "eval",
+        tag: list[str],
         turn_id: Optional[str] = None,
         query: str = "",
     ) -> EvaluationResult:
@@ -33,7 +33,7 @@ def _create_result(  # pylint: disable=too-many-arguments,too-many-positional-ar
             metric_id: Metric identifier
             reason: Reason for the result
             result_status: Result status (ERROR, SKIPPED, etc.)
-            tag: Tag for grouping and filtering results
+            tag: Tag(s) for grouping and filtering results
             turn_id: Turn ID (None for conversation-level)
             query: Query text
         """
@@ -53,7 +53,7 @@ def create_error_result(  # pylint: disable=too-many-arguments,too-many-position
         metric_id: str,
         reason: str,
         *,
-        tag: str = "eval",
+        tag: list[str],
         turn_id: Optional[str] = None,
         query: str = "",
     ) -> EvaluationResult:
@@ -63,7 +63,7 @@ def create_error_result(  # pylint: disable=too-many-arguments,too-many-position
             conv_id: Conversation group ID
             metric_id: Metric identifier
             reason: Error reason
-            tag: Tag for grouping and filtering results
+            tag: Tag(s) for grouping and filtering results
             turn_id: Turn ID (None for conversation-level)
             query: Query text
         """
@@ -77,7 +77,7 @@ def create_skipped_result(  # pylint: disable=too-many-arguments,too-many-positi
         metric_id: str,
         reason: str,
         *,
-        tag: str = "eval",
+        tag: list[str],
         turn_id: Optional[str] = None,
         query: str = "",
     ) -> EvaluationResult:
@@ -87,7 +87,7 @@ def create_skipped_result(  # pylint: disable=too-many-arguments,too-many-positi
             conv_id: Conversation group ID
             metric_id: Metric identifier
             reason: Skip reason
-            tag: Tag for grouping and filtering results
+            tag: Tag(s) for grouping and filtering results
             turn_id: Turn ID (None for conversation-level)
             query: Query text
         """

src/lightspeed_evaluation/pipeline/evaluation/processor.py

@@ -283,6 +283,7 @@ def _evaluate_turn(
                         conv_data.conversation_group_id,
                         metric_identifier,
                         error_reason,
+                        tag=conv_data.tag,
                         turn_id=turn_data.turn_id,
                         query=turn_data.query or "",
                     )
@@ -311,7 +312,10 @@ def _evaluate_conversation(
                 logger.error(error_reason)
                 results.append(
                     self.components.error_handler.create_error_result(
-                        conv_data.conversation_group_id, metric_identifier, error_reason
+                        conv_data.conversation_group_id,
+                        metric_identifier,
+                        error_reason,
+                        tag=conv_data.tag,
                     )
                 )
                 continue

tests/unit/core/models/test_data.py

@@ -436,12 +436,12 @@ def test_valid_creation(self) -> None:
             conversation_group_id="conv1",
             turns=turns,
             description="Test conversation",
-            tag="test_tag",
+            tag=["test_tag"],
             conversation_metrics=["deepeval:conversation_completeness"],
         )
 
         assert eval_data.conversation_group_id == "conv1"
-        assert eval_data.tag == "test_tag"
+        assert eval_data.tag == ["test_tag"]
         assert len(eval_data.turns) == 2
         assert eval_data.description == "Test conversation"
         assert eval_data.conversation_metrics is not None
@@ -452,14 +452,50 @@ def test_default_tag_value(self) -> None:
         turn = TurnData(turn_id="turn1", query="Query")
         eval_data = EvaluationData(conversation_group_id="conv1", turns=[turn])
 
-        assert eval_data.tag == "eval"
+        assert eval_data.tag == ["eval"]
 
-    def test_empty_tag_rejected(self) -> None:
-        """Test that empty tag is rejected."""
+    def test_single_string_tag_normalized_to_list(self) -> None:
+        """Test that a single string tag is normalized to a list."""
+        turn = TurnData(turn_id="turn1", query="Query")
+        eval_data = EvaluationData(
+            conversation_group_id="conv1", turns=[turn], tag="basic"  # type: ignore[arg-type]
+        )
+
+        assert eval_data.tag == ["basic"]
+
+    def test_list_tag_accepted(self) -> None:
+        """Test that a list of tags is accepted."""
+        turn = TurnData(turn_id="turn1", query="Query")
+        eval_data = EvaluationData(
+            conversation_group_id="conv1", turns=[turn], tag=["basic", "advanced"]
+        )
+
+        assert eval_data.tag == ["basic", "advanced"]
+
+    def test_empty_tag_list_rejected(self) -> None:
+        """Test that empty list tag is rejected."""
         turn = TurnData(turn_id="turn1", query="Query")
 
         with pytest.raises(ValidationError):
-            EvaluationData(conversation_group_id="conv1", turns=[turn], tag="")
+            EvaluationData(conversation_group_id="conv1", turns=[turn], tag=[])
+
+    def test_whitespace_only_tag_rejected(self) -> None:
+        """Test that a list of only whitespace strings is rejected."""
+        turn = TurnData(turn_id="turn1", query="Query")
+
+        with pytest.raises(ValidationError):
+            EvaluationData(conversation_group_id="conv1", turns=[turn], tag=["  "])
+
+    def test_non_string_tag_items_rejected(self) -> None:
+        """Test that non-string items in tag list are rejected."""
+        turn = TurnData(turn_id="turn1", query="Query")
+
+        with pytest.raises(ValidationError):
+            EvaluationData(
+                conversation_group_id="conv1",
+                turns=[turn],
+                tag=[1, "prod"],  # type: ignore[list-item]
+            )
 
     def test_empty_conversation_id_rejected(self) -> None:
         """Test that empty conversation_group_id is rejected."""
@@ -500,7 +536,7 @@ def test_default_values(self) -> None:
         )
 
         # Test meaningful defaults
-        assert result.tag == "eval"
+        assert result.tag == ["eval"]
         assert result.score is None
         assert result.reason == ""
         assert result.evaluation_latency == 0
@@ -509,21 +545,45 @@ def test_explicit_tag_value(self) -> None:
         """Test EvaluationResult with explicit tag value."""
         result = EvaluationResult(
             conversation_group_id="conv1",
-            tag="custom_tag",
+            tag=["custom_tag"],
             turn_id="turn1",
             metric_identifier="metric1",
             result="PASS",
             threshold=0.7,
         )
 
-        assert result.tag == "custom_tag"
+        assert result.tag == ["custom_tag"]
+
+    def test_empty_tag_list_rejected(self) -> None:
+        """Test that empty tag list is rejected."""
+        with pytest.raises(ValidationError):
+            EvaluationResult(
+                conversation_group_id="conv1",
+                tag=[],
+                turn_id="turn1",
+                metric_identifier="metric1",
+                result="PASS",
+                threshold=0.7,
+            )
+
+    def test_whitespace_only_tag_rejected(self) -> None:
+        """Test that a list of only whitespace strings is rejected."""
+        with pytest.raises(ValidationError):
+            EvaluationResult(
+                conversation_group_id="conv1",
+                tag=["  "],
+                turn_id="turn1",
+                metric_identifier="metric1",
+                result="PASS",
+                threshold=0.7,
+            )
 
-    def test_empty_tag_rejected(self) -> None:
-        """Test that empty tag is rejected."""
+    def test_non_string_tag_items_rejected(self) -> None:
+        """Test that non-string items in tag list are rejected."""
         with pytest.raises(ValidationError):
             EvaluationResult(
                 conversation_group_id="conv1",
-                tag="",
+                tag=[1, "prod"],  # type: ignore[list-item]
                 turn_id="turn1",
                 metric_identifier="metric1",
                 result="PASS",

tests/unit/core/output/test_statistics_detailed.py

@@ -319,7 +319,7 @@ def test_compute_detailed_stats_by_tag(self) -> None:
         results = [
             EvaluationResult(
                 conversation_group_id="conv1",
-                tag="production",
+                tag=["production"],
                 turn_id="turn1",
                 metric_identifier="metric1",
                 result="PASS",
@@ -329,7 +329,7 @@ def test_compute_detailed_stats_by_tag(self) -> None:
             ),
             EvaluationResult(
                 conversation_group_id="conv2",
-                tag="production",
+                tag=["production"],
                 turn_id="turn1",
                 metric_identifier="metric1",
                 result="PASS",
@@ -339,7 +339,7 @@ def test_compute_detailed_stats_by_tag(self) -> None:
             ),
             EvaluationResult(
                 conversation_group_id="conv3",
-                tag="staging",
+                tag=["staging"],
                 turn_id="turn1",
                 metric_identifier="metric1",
                 result="FAIL",
@@ -372,6 +372,29 @@ def test_compute_detailed_stats_by_tag(self) -> None:
         assert staging_stats["fail_rate"] == 100.0
         assert "score_statistics" in staging_stats
 
+    def test_compute_detailed_stats_multi_tag_result_counted_in_each_bucket(
+        self,
+    ) -> None:
+        """Test a single result with multiple tags appears in each tag's bucket."""
+        result = EvaluationResult(
+            conversation_group_id="conv1",
+            tag=["production", "staging"],
+            turn_id="turn1",
+            metric_identifier="metric1",
+            result="PASS",
+            score=0.9,
+            threshold=0.7,
+        )
+
+        stats = compute_detailed_stats([result]).model_dump()
+
+        assert "production" in stats["by_tag"]
+        assert "staging" in stats["by_tag"]
+        assert stats["by_tag"]["production"]["passed"] == 1
+        assert stats["by_tag"]["production"]["failed"] == 0
+        assert stats["by_tag"]["staging"]["passed"] == 1
+        assert stats["by_tag"]["staging"]["failed"] == 0
+
     def test_compute_detailed_stats_default_tag(self) -> None:
         """Test compute_detailed_stats with default 'eval' tag."""
         results = [