eval-studio-client 1.0.0__py3-none-any.whl → 1.0.1__py3-none-any.whl

eval_studio_client/leaderboards.py

@@ -1,6 +1,7 @@
 import dataclasses
 import datetime
 import json
+import os
 import time
 from typing import Dict
 from typing import List
@@ -33,6 +34,7 @@ class Leaderboard:
     update_time: Optional[datetime.datetime] = None
     problems: List[p6s.Problem] = dataclasses.field(default_factory=list)
     insights: List[i6s.Insight] = dataclasses.field(default_factory=list)
+    summary: Optional[str] = None
     existing_collection: Optional[str] = None
     _report: Optional[str] = None
     _leaderboard: Optional[str] = None
@@ -86,6 +88,42 @@ class Leaderboard:
         if self._client:
             self._leaderboard_api.leaderboard_service_delete_leaderboard(self.key)
 
+    def download_result(self, dest: str):
+        """Downloads the leaderboard result to a JSON file.
+
+        Args:
+            dest (str): The destination path for the report.
+        """
+        if not os.path.exists(dest):
+            raise ValueError("Destination path does not exist.")
+
+        if os.path.isdir(dest):
+            dest = os.path.join(dest, "results.json")
+
+        if self._client and self.finished:
+            headers: Dict[str, str] = {}
+            url = urljoin(
+                self._client.configuration.host, f"/content/{self.key}/results"
+            )
+            self._client.update_params_for_auth(
+                headers=headers,
+                queries=[],
+                auth_settings=[],
+                resource_path=url,
+                method="GET",
+                body=None,
+            )
+            response = urllib3.request("GET", url, headers=headers)
+
+            if response.status == 200:
+                with open(dest, "wb") as f:
+                    f.write(response.data)
+                    return
+            else:
+                raise RuntimeError("Failed to retrieve leaderboard result.")
+
+        raise ValueError("Cannot download result for unfinished leaderboard.")
+
     def download_report(self, dest: str):
         """Downloads the leaderboard report to a zip file.
 
@@ -114,6 +152,30 @@ class Leaderboard:
 
         raise ValueError("Cannot download report for unfinished leaderboard.")
 
+    def get_result_json(self) -> str:
+        """Retrieves the leaderboard result as a JSON string."""
+        if self._client and self.finished:
+            headers: Dict[str, str] = {}
+            url = urljoin(
+                self._client.configuration.host, f"/content/{self.key}/results"
+            )
+            self._client.update_params_for_auth(
+                headers=headers,
+                queries=[],
+                auth_settings=[],
+                resource_path=url,
+                method="GET",
+                body=None,
+            )
+            response = urllib3.request("GET", url, headers=headers)
+
+            if response.status == 200:
+                return str(response.data)
+            else:
+                raise RuntimeError("Failed to retrieve leaderboard result.")
+
+        raise ValueError("Cannot download result for unfinished leaderboard.")
+
     def get_table(self) -> LeaderboardTable:
         """Retrieves the leaderboard table."""
         if self._client and self.finished:
@@ -170,6 +232,7 @@ class Leaderboard:
         """Refresh the leaderboard with the latest API data."""
         self.key = api_leaderboard.name or ""
         self.update_time = api_leaderboard.update_time
+        self.summary = api_leaderboard.leaderboard_summary
         self._leaderboard = api_leaderboard.leaderboard_table
         self._report = api_leaderboard.leaderboard_report or ""
         self._status = api_leaderboard.status
@@ -192,6 +255,7 @@ class Leaderboard:
             update_time=api_leaderboard.update_time,
             problems=problems,
             insights=insights,
+            summary=api_leaderboard.leaderboard_summary,
             existing_collection=api_leaderboard.h2ogpte_collection or None,
             _evaluator_name=api_leaderboard.evaluator or "",
             _test_names=api_leaderboard.tests or [],
@@ -208,3 +272,62 @@ class Leaderboard:
             models.V1LeaderboardStatus.LEADERBOARD_STATUS_COMPLETED,
             models.V1LeaderboardStatus.LEADERBOARD_STATUS_FAILED,
         ]
+
+    @staticmethod
+    def from_operation(
+        operation: models.V1Operation, client: Optional[api.ApiClient]
+    ) -> Optional["Leaderboard"]:
+        """Retrieves the leaderboard from the operation, which created it.
+
+        Args:
+            operation: The operation that created the dashboard.
+            client: The API client to use for the leaderboard retrieval.
+
+        Returns:
+            Leaderboard: The leaderboard instance created by the operation.
+        """
+        if not client:
+            raise RuntimeError("API Client is not provided")
+
+        if not operation.metadata:
+            raise RuntimeError(
+                "Operation metadata missing, it's not possible to retrieve leaderboard from operation"
+            )
+
+        leaderboard_api = api.LeaderboardServiceApi(client)
+        leadeboard_id = operation.metadata.to_dict().get("leaderboard", "")
+        res = leaderboard_api.leaderboard_service_get_leaderboard(str(leadeboard_id))
+        if res and res.leaderboard:
+            return Leaderboard._from_api_leaderboard(res.leaderboard, client)
+
+        return None
+
+
+class _Leaderboards:
+    def __init__(self, client: api.ApiClient):
+        self._client = client
+        self._api = api.LeaderboardServiceApi(client)
+
+    def get(self, key: str) -> Leaderboard:
+        """Gets an individual leaderboard with a given key from Eval Studio.
+
+        Args:
+            key: The leaderboard resource name to retrieve.
+        """
+        res = self._api.leaderboard_service_get_leaderboard(key)
+        if res and res.leaderboard:
+            return Leaderboard._from_api_leaderboard(res.leaderboard, self._client)
+
+        raise KeyError("Leaderboard not found.")
+
+    def list(self) -> List[Leaderboard]:
+        """Lists all user leaderboards in Eval Studio."""
+        res = self._api.leaderboard_service_list_leaderboards()
+        if res:
+            res_leaderboards = res.leaderboards or []
+            return [
+                Leaderboard._from_api_leaderboard(lb, self._client)
+                for lb in res_leaderboards
+            ]
+
+        return []

eval_studio_client/models.py

@@ -168,7 +168,7 @@ class Model:
             )
 
         if res and res.operation:
-            return self._get_leaderboard_from_operation(res.operation)
+            return l10s.Leaderboard.from_operation(res.operation, self._client)
 
         return None
 
@@ -226,7 +226,7 @@ class Model:
         )
 
         if res and res.operation:
-            return self._get_dashboard_from_operation(res.operation)
+            return d8s.Dashboard.from_operation(res.operation, self._client)
 
         return None
 
@@ -257,7 +257,7 @@ class Model:
         )
         res = self._leaderboard_api.leaderboard_service_import_leaderboard(req)
         if res and res.operation:
-            return self._get_leaderboard_from_operation(res.operation)
+            return l10s.Leaderboard.from_operation(res.operation, self._client)
 
         return None
 
@@ -273,45 +273,6 @@ class Model:
 
         raise RuntimeError("Failed to list base models")
 
-    def _get_leaderboard_from_operation(
-        self, operation: models.V1Operation
-    ) -> Optional[l10s.Leaderboard]:
-        """Retrieves the leaderboard from the operation, which created it.
-
-        Args:
-            operation: The operation that created the leaderboard.
-        """
-        if not operation.metadata:
-            raise RuntimeError("Not possible to retrieve leaderboard from operation")
-
-        leadeboard_id = operation.metadata.to_dict().get("leaderboard")
-        res = self._leaderboard_api.leaderboard_service_get_leaderboard(leadeboard_id)
-        if res and res.leaderboard:
-            return l10s.Leaderboard._from_api_leaderboard(res.leaderboard, self._client)
-
-        return None
-
-    def _get_dashboard_from_operation(
-        self, operation: models.V1Operation
-    ) -> Optional[d8s.Dashboard]:
-        """Retrieves the dashboard from the operation, which created it.
-
-        Args:
-            operation: The operation that created the dashboard.
-        """
-        if not self._client:
-            raise RuntimeError("Client is not set.")
-
-        if not operation.metadata:
-            raise RuntimeError("Not possible to retrieve dashboard from operation")
-
-        dashboard_id = operation.metadata.to_dict().get("dashboard")
-        res = self._dashboard_api.dashboard_service_get_dashboard(dashboard_id)
-        if res and res.dashboard:
-            return d8s.Dashboard._from_api_dashboard(res.dashboard, self._client)
-
-        return None
-
     @staticmethod
     def _from_api_model(api_model: models.V1Model, client: api.ApiClient) -> "Model":
         """Converts the API model to the client model."""

eval_studio_client/test_labs.py

@@ -7,7 +7,8 @@ from typing import Union
 import uuid
 
 from eval_studio_client import api
-from eval_studio_client import evaluators
+from eval_studio_client import dashboards
+from eval_studio_client import evaluators as e8s
 from eval_studio_client import leaderboards as l10s
 from eval_studio_client.api import models as apiModels
 
@@ -92,11 +93,56 @@ class TestLab:
         self._models.append(_m)
         return _m
 
-    def evaluate(self, evaluator: evaluators.Evaluator) -> Optional[l10s.Leaderboard]:
+    def evaluate(
+        self,
+        evaluators: Union[e8s.Evaluator, List[e8s.Evaluator]],
+        name: Optional[str] = None,
+        description: Optional[str] = None,
+    ) -> Optional[dashboards.Dashboard]:
         """Runs an evaluation for the test lab.
 
+        Args:
+            evaluators (Union[e8s.Evaluator, List[e8s.Evaluator]]): One or many evaluators
+                used to evaluate the test lab.
+            name (str, optional): Optional name for the evaluation.
+            description (str, optional): Optional description for the evaluation.
+
+        Returns:
+           Dashboard: Evaluation dashboard instance. In case launching of evaluation
+              fails, `None` is returned.
+        """
+        _evaluators = (
+            [evaluators] if isinstance(evaluators, e8s.Evaluator) else evaluators
+        )
+        name = name or self.name or "Imported Dashboard"
+        description = description or self.description or ""
+        req = apiModels.V1BatchImportLeaderboardRequest(
+            testLabJson=self.json(),
+            evaluators=[e.key for e in _evaluators],
+            model=None,
+            dashboardDisplayName=name,
+            dashboardDescription=description,
+            testDisplayName=f"{name} - Test",
+            testDescription=f"Test suite for {description}",
+        )
+        res = self._leaderboard_api.leaderboard_service_batch_import_leaderboard(req)
+
+        if res and res.operation:
+            return dashboards.Dashboard.from_operation(res.operation, self._client)
+
+        return None
+
+    def create_leaderboard(
+        self, evaluator: e8s.Evaluator
+    ) -> Optional[l10s.Leaderboard]:
+        """Creates a single leaderboard for the test lab.
+
         Args:
             evaluator: The evaluator to use for the evaluation.
+
+        Returns:
+           Leaderboard: Single evaluation leaderboard instance.
+            In case launching of evaluation fails, `None` is returned.
         """
         req = apiModels.V1ImportLeaderboardRequest(
             testLabJson=self.json(),
@@ -109,7 +155,7 @@ class TestLab:
         )
         res = self._leaderboard_api.leaderboard_service_import_leaderboard(req)
         if res and res.operation:
-            return self._get_leaderboard_from_operation(res.operation)
+            return l10s.Leaderboard.from_operation(res.operation, self._client)
 
         return None
 
@@ -131,24 +177,6 @@ class TestLab:
 
         return json.dumps(lab, indent=4, sort_keys=True)
 
-    def _get_leaderboard_from_operation(
-        self, operation: apiModels.V1Operation
-    ) -> Optional[l10s.Leaderboard]:
-        """Retrieves the leaderboard from the operation, which created it.
-
-        Args:
-            operation: The operation that created the leaderboard.
-        """
-        if not operation.metadata:
-            raise RuntimeError("Not possible to retrieve leaderboard from operation")
-
-        leadeboard_id = operation.metadata.to_dict().get("leaderboard")
-        res = self._leaderboard_api.leaderboard_service_get_leaderboard(leadeboard_id)
-        if res and res.leaderboard:
-            return l10s.Leaderboard._from_api_leaderboard(res.leaderboard, self._client)
-
-        return None
-
     def _llm_model_names(self) -> List[str]:
         return [m.llm_model_name for m in self.models]
 

eval_studio_client/tests.py

@@ -118,6 +118,71 @@ class _TestCaseGenerationHandle:
         )
 
 
+@dataclasses.dataclass
+class _TestCaseLibraryGetHandle(_TestCaseGenerationHandle):
+
+    @staticmethod
+    def _from_operation(
+        res: (
+            models.V1ImportTestCasesFromLibraryResponse | models.V1GetOperationResponse
+        ),
+    ) -> "_TestCaseLibraryGetHandle":
+        """Converts an API operation to prompt library handle."""
+        op: models.V1Operation | None = res.operation
+        if not op:
+            return _TestCaseLibraryGetHandle(name=None)
+
+        # progress
+        if hasattr(op, "metadata") and op.metadata:
+            meta_dict = op.metadata.to_dict() or {}
+        else:
+            meta_dict = {}
+
+        return _TestCaseLibraryGetHandle(
+            name=op.name,
+            progress=meta_dict.get("progress"),
+            progress_message=meta_dict.get("progressMessage"),
+            error=op.error,
+            done=op.done,
+        )
+
+
+@dataclasses.dataclass
+class TestCaseLibraryItem:
+    """Represents a single test case library item - test suite."""
+
+    key: str
+    name: str
+    description: str
+    test_suite_url: str
+    test_count: int
+    test_case_count: int
+    evaluates: List[str]
+    categories: List[str]
+
+    @staticmethod
+    def _from_api_items(
+        api_items: List[models.V1PromptLibraryItem],
+    ) -> List["TestCaseLibraryItem"]:
+        return (
+            [
+                TestCaseLibraryItem(
+                    key=api_item.name or "",
+                    name=api_item.display_name or "",
+                    description=api_item.description or "",
+                    test_suite_url=api_item.test_suite_url or "",
+                    test_count=api_item.test_count or 0,
+                    test_case_count=api_item.test_case_count or 0,
+                    evaluates=list(api_item.evaluates) if api_item.evaluates else [],
+                    categories=list(api_item.categories) if api_item.categories else [],
+                )
+                for api_item in api_items
+            ]
+            if api_items
+            else []
+        )
+
+
 @dataclasses.dataclass
 class TestCase:
     """Represents a single test case, which contains tested prompt, expected answer
@@ -183,6 +248,7 @@ class Test:
     update_time: Optional[datetime.datetime] = None
     _client: Optional[api.ApiClient] = None
     _gen_tc_op_name: Optional[str] = None
+    _lib_tc_op_name: Optional[str] = None
 
     def __post_init__(self):
         if self._client:
@@ -267,7 +333,7 @@ class Test:
 
         Args:
             count (int): Number of test cases to generate (generator may return fewer
-               prompts).
+                prompts).
             model (str): Model to use for generating the prompts.
             base_llm_model (str): Base LLM model to use for generating the prompts.
             generators (List[TestCaseGenerator]): Methods to use for generation.
@@ -342,6 +408,127 @@ class Test:
 
         raise TimeoutError("Waiting timeout has been reached.")
 
+    def list_test_suite_library_items(
+        self,
+        filter_by_categories: Optional[List[str]] = None,
+        filter_by_purposes: Optional[List[str]] = None,
+        filter_by_evaluates: Optional[List[str]] = None,
+        filter_by_origin: Optional[str] = None,
+        filter_by_test_case_count: Optional[int] = None,
+        filter_by_test_count: Optional[int] = None,
+        filter_by_fts: Optional[str] = None,
+    ) -> List[TestCaseLibraryItem]:
+        """Retrieves a list of all available items - suites of tests - in the library.
+
+        Args:
+            filter_by_categories (List[str]): List of categories to filter
+                the library items.
+            filter_by_purposes (List[str]): List of purposes to filter
+                the library items.
+            filter_by_evaluates (List[str]): List of evaluates to filter
+                the library items.
+            filter_by_origin (str): Origin to filter the library items.
+            filter_by_test_case_count (int): Test case count to filter
+                the library items.
+            filter_by_test_count (int): Test count to filter the library items.
+            filter_by_fts (str): FTS to filter the library items - phrase to search for.
+
+        Returns:
+            List[TestCaseLibraryItem]: List of library items.
+        """
+        req = models.TestServiceListTestCaseLibraryItemsRequest(
+            filter_by_categories=filter_by_categories,
+            filter_by_purposes=filter_by_purposes,
+            filter_by_evaluates=filter_by_evaluates,
+            filter_by_origin=filter_by_origin,
+            filter_by_test_case_count=filter_by_test_case_count,
+            filter_by_test_count=filter_by_test_count,
+            filter_by_fts=filter_by_fts,
+        )
+
+        res = self._test_api.test_service_list_test_case_library_items(self.key, req)
+        if res and res.prompt_library_items:
+            return TestCaseLibraryItem._from_api_items(res.prompt_library_items)
+
+        return []
+
+    def add_library_test_cases(
+        self, test_suite_url: str, count: int, test_document_urls: Optional[List[str]]
+    ) -> None:
+        """Sample test cases from the test suite library and add them to the test.
+
+        Args:
+            test_suite_url (str): The URL of the library test suite to get TestCases
+                from (sample).
+            count (int): The number of TestCases to get from the library.
+            test_document_urls (List[str]): The list of target Test corpus
+                document URLs to skip when returning library TestCases corpus.
+        """
+        req = models.TestServiceImportTestCasesFromLibraryRequest(
+            test_suite_url=test_suite_url,
+            count=count,
+            test_document_urls=test_document_urls,
+        )
+
+        res = self._test_api.test_service_import_test_cases_from_library(self.key, req)
+
+        op: models.V1Operation | None = res.operation
+        self._lib_tc_op_name = op.name if op else None
+
+    def wait_for_library_test_case_get(
+        self, timeout: Optional[float] = None, verbose: bool = False
+    ) -> None:
+        """Waits for the library test cases(s) sampling  to finish.
+
+        Args:
+            timeout (float): The maximum time to wait in seconds.
+            verbose (bool): If True, prints the status of the handle while waiting.
+        """
+        if not self._lib_tc_op_name:
+            raise ValueError(
+                "There is no ongoing getting of test case(s) from the library - "
+                "the operation name is not set."
+            )
+
+        if verbose:
+            print(
+                f"Waiting for getting library test case(s) operation to finish "
+                f"({self._lib_tc_op_name}):"
+            )
+        if self._client:
+            # exponential backoff
+            wait_time = 1.0
+            wait_coef = 1.6
+            wait_max = 8.0
+            wait_total = 0.0
+            timeout = timeout or float(2 * 24 * 60 * 60)  # 2 days
+            progress_bar = utils.ProgressBar()
+            while wait_total < timeout:
+                handle = _TestCaseLibraryGetHandle._from_operation(
+                    self._operation_api.operation_service_get_operation(
+                        self._lib_tc_op_name
+                    )
+                )
+
+                if verbose:
+                    progress_bar.update(handle.progress or 0, handle.progress_message)
+
+                if handle.done:
+                    if handle.error:
+                        raise RuntimeError(
+                            f"Getting of library test case(s) failed: {handle.error}"
+                        )
+                    return
+
+                wait_time *= wait_coef
+                time.sleep(min(wait_time, wait_max))
+        else:
+            raise ValueError(
+                "Unable to establish a connection to the Eval Studio host."
+            )
+
+        raise TimeoutError("Waiting timeout has been reached.")
+
     def delete(self, force=False):
         """Deletes the test.
 

{eval_studio_client-1.0.0.dist-info → eval_studio_client-1.0.1.dist-info}/METADATA

@@ -1,10 +1,9 @@
 Metadata-Version: 2.3
 Name: eval-studio-client
-Version: 1.0.0
+Version: 1.0.1
 Project-URL: Source, https://github.com/h2oai/eval-studio/tree/main/client-py/src/
 Project-URL: Issues, https://github.com/h2oai/eval-studio/issues
 Author-email: "H2O.ai" <support@h2o.ai>
-License: MIT
 Classifier: Development Status :: 4 - Beta
 Classifier: Programming Language :: Python
 Classifier: Programming Language :: Python :: 3.9