more scoring function for rag

llama_stack/providers/inline/scoring/braintrust/braintrust.py

@@ -19,6 +19,10 @@ from autoevals.ragas import (
     AnswerCorrectness,
     AnswerRelevancy,
     AnswerSimilarity,
+    ContextEntityRecall,
+    ContextPrecision,
+    ContextRecall,
+    ContextRelevancy,
     Faithfulness,
 )
 from pydantic import BaseModel
@@ -34,6 +38,10 @@ from .config import BraintrustScoringConfig
 from .scoring_fn.fn_defs.answer_correctness import answer_correctness_fn_def
 from .scoring_fn.fn_defs.answer_relevancy import answer_relevancy_fn_def
 from .scoring_fn.fn_defs.answer_similarity import answer_similarity_fn_def
+from .scoring_fn.fn_defs.context_entity_recall import context_entity_recall_fn_def
+from .scoring_fn.fn_defs.context_precision import context_precision_fn_def
+from .scoring_fn.fn_defs.context_recall import context_recall_fn_def
+from .scoring_fn.fn_defs.context_relevancy import context_relevancy_fn_def
 from .scoring_fn.fn_defs.factuality import factuality_fn_def
 from .scoring_fn.fn_defs.faithfulness import faithfulness_fn_def
 
@@ -70,6 +78,26 @@ SUPPORTED_BRAINTRUST_SCORING_FN_ENTRY = [
         evaluator=Faithfulness(),
         fn_def=faithfulness_fn_def,
     ),
+    BraintrustScoringFnEntry(
+        identifier="braintrust::context-entity-recall",
+        evaluator=ContextEntityRecall(),
+        fn_def=context_entity_recall_fn_def,
+    ),
+    BraintrustScoringFnEntry(
+        identifier="braintrust::context-precision",
+        evaluator=ContextPrecision(),
+        fn_def=context_precision_fn_def,
+    ),
+    BraintrustScoringFnEntry(
+        identifier="braintrust::context-recall",
+        evaluator=ContextRecall(),
+        fn_def=context_recall_fn_def,
+    ),
+    BraintrustScoringFnEntry(
+        identifier="braintrust::context-relevancy",
+        evaluator=ContextRelevancy(),
+        fn_def=context_relevancy_fn_def,
+    ),
 ]
 
 

llama_stack/providers/inline/scoring/braintrust/scoring_fn/fn_defs/context_entity_recall.py

@@ -0,0 +1,26 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from llama_stack.apis.common.type_system import NumberType
+from llama_stack.apis.scoring_functions import (
+    AggregationFunctionType,
+    BasicScoringFnParams,
+    ScoringFn,
+)
+
+context_entity_recall_fn_def = ScoringFn(
+    identifier="braintrust::context-entity-recall",
+    description=(
+        "Evaluates how well the context captures the named entities present in the "
+        "reference answer. See: github.com/braintrustdata/autoevals"
+    ),
+    provider_id="braintrust",
+    provider_resource_id="context-entity-recall",
+    return_type=NumberType(),
+    params=BasicScoringFnParams(
+        aggregation_functions=[AggregationFunctionType.average]
+    ),
+)

llama_stack/providers/inline/scoring/braintrust/scoring_fn/fn_defs/context_precision.py

@@ -0,0 +1,26 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from llama_stack.apis.common.type_system import NumberType
+from llama_stack.apis.scoring_functions import (
+    AggregationFunctionType,
+    BasicScoringFnParams,
+    ScoringFn,
+)
+
+context_precision_fn_def = ScoringFn(
+    identifier="braintrust::context-precision",
+    description=(
+        "Measures how much of the provided context is actually relevant to answering the "
+        "question. See: github.com/braintrustdata/autoevals"
+    ),
+    provider_id="braintrust",
+    provider_resource_id="context-precision",
+    return_type=NumberType(),
+    params=BasicScoringFnParams(
+        aggregation_functions=[AggregationFunctionType.average]
+    ),
+)

llama_stack/providers/inline/scoring/braintrust/scoring_fn/fn_defs/context_recall.py

@@ -0,0 +1,26 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from llama_stack.apis.common.type_system import NumberType
+from llama_stack.apis.scoring_functions import (
+    AggregationFunctionType,
+    BasicScoringFnParams,
+    ScoringFn,
+)
+
+context_recall_fn_def = ScoringFn(
+    identifier="braintrust::context-recall",
+    description=(
+        "Evaluates how well the context covers the information needed to answer the "
+        "question. See: github.com/braintrustdata/autoevals"
+    ),
+    provider_id="braintrust",
+    provider_resource_id="context-recall",
+    return_type=NumberType(),
+    params=BasicScoringFnParams(
+        aggregation_functions=[AggregationFunctionType.average]
+    ),
+)

llama_stack/providers/inline/scoring/braintrust/scoring_fn/fn_defs/context_relevancy.py

@@ -0,0 +1,26 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from llama_stack.apis.common.type_system import NumberType
+from llama_stack.apis.scoring_functions import (
+    AggregationFunctionType,
+    BasicScoringFnParams,
+    ScoringFn,
+)
+
+context_relevancy_fn_def = ScoringFn(
+    identifier="braintrust::context-relevancy",
+    description=(
+        "Assesses how relevant the provided context is to the given question. "
+        "See: github.com/braintrustdata/autoevals"
+    ),
+    provider_id="braintrust",
+    provider_resource_id="context-relevancy",
+    return_type=NumberType(),
+    params=BasicScoringFnParams(
+        aggregation_functions=[AggregationFunctionType.average]
+    ),
+)