Move provider to new llama-stack-api module

distribution/run.yaml

@@ -93,5 +93,3 @@ registered_resources:
   tool_groups: []
 server:
   port: 8321
-telemetry:
-  enabled: true

src/llama_stack_provider_ragas/compat.py

@@ -0,0 +1,110 @@
+"""
+Compatibility layer for llama_stack imports.
+
+This module provides backward compatibility by attempting to import from
+the legacy llama_stack package first, then falling back to the newer
+llama_stack_api package structure.
+"""
+
+# Provider datatypes and API definitions
+try:  # Legacy llama_stack layout
+    from llama_stack.apis.datatypes import Api
+    from llama_stack.providers.datatypes import (
+        BenchmarksProtocolPrivate,
+        InlineProviderSpec,
+        ProviderSpec,
+        RemoteProviderSpec,
+    )
+except (ImportError, ModuleNotFoundError):
+    # Newer llama_stack_api layout
+    from llama_stack_api import (
+        Api,
+        BenchmarksProtocolPrivate,
+        InlineProviderSpec,
+        ProviderSpec,
+        RemoteProviderSpec,
+    )
+
+# Benchmarks
+try:
+    from llama_stack.apis.benchmarks import Benchmark
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import Benchmark
+
+# Common job types
+try:
+    from llama_stack.apis.common.job_types import Job, JobStatus
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import Job, JobStatus
+
+# DatasetIO
+try:
+    from llama_stack.apis.datasetio import DatasetIO
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import DatasetIO
+
+# Eval
+try:
+    from llama_stack.apis.eval import BenchmarkConfig, Eval, EvaluateResponse
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import BenchmarkConfig, Eval, EvaluateResponse
+
+# Inference
+try:
+    from llama_stack.apis.inference import (
+        Inference,
+        OpenAICompletionRequestWithExtraBody,
+        OpenAIEmbeddingsRequestWithExtraBody,
+        SamplingParams,
+        TopPSamplingStrategy,
+    )
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import (
+        Inference,
+        OpenAICompletionRequestWithExtraBody,
+        OpenAIEmbeddingsRequestWithExtraBody,
+        SamplingParams,
+        TopPSamplingStrategy,
+    )
+
+# Scoring
+try:
+    from llama_stack.apis.scoring import ScoringResult
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import ScoringResult
+
+# Schema utils
+try:
+    from llama_stack.schema_utils import json_schema_type
+except (ImportError, ModuleNotFoundError):
+    from llama_stack_api import json_schema_type
+
+__all__ = [
+    # API and Provider types
+    "Api",
+    "BenchmarksProtocolPrivate",
+    "InlineProviderSpec",
+    "ProviderSpec",
+    "RemoteProviderSpec",
+    # Benchmarks
+    "Benchmark",
+    # Job types
+    "Job",
+    "JobStatus",
+    # DatasetIO
+    "DatasetIO",
+    # Eval
+    "BenchmarkConfig",
+    "Eval",
+    "EvaluateResponse",
+    # Inference
+    "Inference",
+    "OpenAICompletionRequestWithExtraBody",
+    "OpenAIEmbeddingsRequestWithExtraBody",
+    "SamplingParams",
+    "TopPSamplingStrategy",
+    # Scoring
+    "ScoringResult",
+    # Schema utils
+    "json_schema_type",
+]

src/llama_stack_provider_ragas/config.py

@@ -1,6 +1,7 @@
-from llama_stack.schema_utils import json_schema_type
 from pydantic import BaseModel, Field, SecretStr
 
+from .compat import json_schema_type
+
 
 class RagasConfig(BaseModel):
     """Additional configuration parameters for Ragas evaluation."""

src/llama_stack_provider_ragas/inline/__init__.py

@@ -1,7 +1,6 @@
 from typing import Any
 
-from llama_stack.apis.datatypes import Api
-
+from ..compat import Api
 from ..config import RagasProviderInlineConfig
 from .ragas_inline_eval import RagasEvaluatorInline
 

src/llama_stack_provider_ragas/inline/provider.py

@@ -1,5 +1,4 @@
-from llama_stack.providers.datatypes import Api, InlineProviderSpec, ProviderSpec
-
+from ..compat import Api, InlineProviderSpec, ProviderSpec
 from ..constants import PROVIDER_TYPE
 
 
@@ -15,6 +14,5 @@ def get_provider_spec() -> ProviderSpec:
             Api.files,
             Api.benchmarks,
             Api.datasetio,
-            Api.telemetry,
         ],
     )

src/llama_stack_provider_ragas/inline/ragas_inline_eval.py

@@ -3,14 +3,6 @@
 import logging
 from typing import Any
 
-from llama_stack.apis.benchmarks import Benchmark
-from llama_stack.apis.common.job_types import Job, JobStatus
-from llama_stack.apis.datasetio import DatasetIO
-from llama_stack.apis.eval import BenchmarkConfig, Eval, EvaluateResponse
-from llama_stack.apis.inference import Inference
-from llama_stack.apis.scoring import ScoringResult
-from llama_stack.providers.datatypes import BenchmarksProtocolPrivate
-from llama_stack.schema_utils import json_schema_type
 from ragas import EvaluationDataset
 from ragas import evaluate as ragas_evaluate
 from ragas.metrics import (
@@ -22,6 +14,19 @@
 )
 from ragas.run_config import RunConfig
 
+from ..compat import (
+    Benchmark,
+    BenchmarkConfig,
+    BenchmarksProtocolPrivate,
+    DatasetIO,
+    Eval,
+    EvaluateResponse,
+    Inference,
+    Job,
+    JobStatus,
+    ScoringResult,
+    json_schema_type,
+)
 from ..config import RagasProviderInlineConfig
 from ..constants import METRIC_MAPPING
 from ..errors import RagasEvaluationError

src/llama_stack_provider_ragas/inline/wrappers_inline.py

@@ -3,15 +3,16 @@
 
 from langchain_core.language_models.llms import Generation, LLMResult
 from langchain_core.prompt_values import PromptValue
-from llama_stack.apis.inference import (
+from ragas.embeddings.base import BaseRagasEmbeddings
+from ragas.llms.base import BaseRagasLLM
+from ragas.run_config import RunConfig
+
+from ..compat import (
     OpenAICompletionRequestWithExtraBody,
     OpenAIEmbeddingsRequestWithExtraBody,
     SamplingParams,
     TopPSamplingStrategy,
 )
-from ragas.embeddings.base import BaseRagasEmbeddings
-from ragas.llms.base import BaseRagasLLM
-from ragas.run_config import RunConfig
 
 logger = logging.getLogger(__name__)
 

src/llama_stack_provider_ragas/remote/__init__.py

@@ -1,7 +1,6 @@
 from typing import Any
 
-from llama_stack.apis.datatypes import Api
-
+from ..compat import Api
 from ..config import RagasProviderRemoteConfig
 from .ragas_remote_eval import RagasEvaluatorRemote
 

src/llama_stack_provider_ragas/remote/kubeflow/components.py

@@ -85,11 +85,11 @@ def run_ragas_evaluation(
     import logging
 
     import pandas as pd
-    from llama_stack.apis.inference import SamplingParams
     from ragas import EvaluationDataset, evaluate
     from ragas.dataset_schema import EvaluationResult
     from ragas.run_config import RunConfig
 
+    from llama_stack_provider_ragas.compat import SamplingParams
     from llama_stack_provider_ragas.constants import METRIC_MAPPING
     from llama_stack_provider_ragas.logging_utils import render_dataframe_as_table
     from llama_stack_provider_ragas.remote.wrappers_remote import (

src/llama_stack_provider_ragas/remote/provider.py

@@ -1,9 +1,4 @@
-from llama_stack.providers.datatypes import (
-    Api,
-    ProviderSpec,
-    RemoteProviderSpec,
-)
-
+from ..compat import Api, ProviderSpec, RemoteProviderSpec
 from ..constants import PROVIDER_TYPE
 
 
@@ -26,6 +21,5 @@ def get_provider_spec() -> ProviderSpec:
             Api.files,
             Api.benchmarks,
             Api.datasetio,
-            Api.telemetry,
         ],
     )

src/llama_stack_provider_ragas/remote/ragas_remote_eval.py

@@ -4,14 +4,19 @@
 
 import pandas as pd
 import requests
-from llama_stack.apis.benchmarks import Benchmark
-from llama_stack.apis.common.job_types import Job, JobStatus
-from llama_stack.apis.eval import BenchmarkConfig, Eval, EvaluateResponse
-from llama_stack.apis.scoring import ScoringResult
-from llama_stack.providers.datatypes import BenchmarksProtocolPrivate
-from llama_stack.schema_utils import json_schema_type
 from pydantic import BaseModel
 
+from ..compat import (
+    Benchmark,
+    BenchmarkConfig,
+    BenchmarksProtocolPrivate,
+    Eval,
+    EvaluateResponse,
+    Job,
+    JobStatus,
+    ScoringResult,
+    json_schema_type,
+)
 from ..config import (
     KubeflowConfig,
     RagasConfig,

src/llama_stack_provider_ragas/remote/wrappers_remote.py

@@ -2,14 +2,15 @@
 
 from langchain_core.language_models.llms import Generation, LLMResult
 from langchain_core.prompt_values import PromptValue
-from llama_stack.apis.inference import SamplingParams, TopPSamplingStrategy
 from llama_stack_client import AsyncLlamaStackClient, LlamaStackClient, omit
 from llama_stack_client.types.completion_create_response import CompletionCreateResponse
 from llama_stack_client.types.create_embeddings_response import CreateEmbeddingsResponse
 from ragas.embeddings.base import BaseRagasEmbeddings
 from ragas.llms.base import BaseRagasLLM
 from ragas.run_config import RunConfig
 
+from ..compat import SamplingParams, TopPSamplingStrategy
+
 logger = logging.getLogger(__name__)
 
 

tests/conftest.py

@@ -3,10 +3,10 @@
 
 import pytest
 from dotenv import load_dotenv
-from llama_stack.apis.inference import SamplingParams, TopPSamplingStrategy
 from llama_stack_client import LlamaStackClient
 from ragas import EvaluationDataset
 
+from llama_stack_provider_ragas.compat import SamplingParams, TopPSamplingStrategy
 from llama_stack_provider_ragas.config import (
     KubeflowConfig,
     RagasProviderInlineConfig,