arize-phoenix 5.10.0__py3-none-any.whl → 5.11.0__py3-none-any.whl

{arize_phoenix-5.10.0.dist-info → arize_phoenix-5.11.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 5.10.0
+Version: 5.11.0
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -53,6 +53,7 @@ Requires-Dist: wrapt
 Provides-Extra: container
 Requires-Dist: anthropic; extra == 'container'
 Requires-Dist: fast-hdbscan>=0.2.0; extra == 'container'
+Requires-Dist: google-generativeai; extra == 'container'
 Requires-Dist: numba>=0.60.0; extra == 'container'
 Requires-Dist: openai>=1.0.0; extra == 'container'
 Requires-Dist: opentelemetry-exporter-otlp; extra == 'container'

{arize_phoenix-5.10.0.dist-info → arize_phoenix-5.11.0.dist-info}/RECORD

@@ -1,12 +1,12 @@
 phoenix/__init__.py,sha256=X3eUEwd2rG8KKWWYVNNDJoqo08ihfjgHhlP29dcdNJE,5481
 phoenix/auth.py,sha256=JpkwJbis2INlIXWcQ-M_Nk5Ln9LBgHMdWNnaAQp0D2w,10940
-phoenix/config.py,sha256=xOM5eupLzXXCfZ4dzCYW-4pKG1xcLDNuu7vxvyiGfoM,25591
+phoenix/config.py,sha256=QLaKuc7N3qZsaAzJnsUVyFuWaROYTVlZsVXTr0Re6d8,26191
 phoenix/datetime_utils.py,sha256=iJzNG6YJ6V7_u8B2iA7P2Z26FyxYbOPtx0dhJ7kNDHA,3398
 phoenix/exceptions.py,sha256=n2L2KKuecrdflB9MsCdAYCiSEvGJptIsfRkXMoJle7A,169
 phoenix/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
 phoenix/services.py,sha256=kpW1WL0kiB8XJsO6XycvZVJ-lBkNoenhQ7atCvBoSe8,5365
 phoenix/settings.py,sha256=ht-0oN-sMV6SPXrk7Tu1EZlngpAYkGNLYPhO8DyrdQI,661
-phoenix/version.py,sha256=PtYE7255x6FNDU5KsxORw9DOsNBQOKOzcRu8NtdUMBk,23
+phoenix/version.py,sha256=6_DlfgAX4MBZnyPuaEthIkiqPKB1Zz3_f4Z2L40k-Ws,23
 phoenix/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/core/embedding_dimension.py,sha256=zKGbcvwOXgLf-yrJBpQyKtd-LEOPRKHnUToyAU8Owis,87
 phoenix/core/model.py,sha256=qBFraOtmwCCnWJltKNP18DDG0mULXigytlFsa6YOz6k,4837
@@ -18,7 +18,7 @@ phoenix/db/alembic.ini,sha256=GIS6HpHaKaJbbuahZg1Rc1D2_QqyCkV9r58wdARGf6w,3262
 phoenix/db/bulk_inserter.py,sha256=faNjuwLqqsw4ky8sa4D0h9u5TvEDTOjrccUW89L008E,12725
 phoenix/db/engines.py,sha256=BYFiHFGABtNuCo7OB8fLc46gMxur4rePK_sJxrMrMjA,6508
 phoenix/db/enums.py,sha256=tt7iovXLhVTLZ3_LbHNGgcI44SnNjXfkKtLAZG57T54,428
-phoenix/db/facilitator.py,sha256=Aulr9-sUvRn20Z7mjpWhW6TFE1amRzL8qSEjscvL-xw,4123
+phoenix/db/facilitator.py,sha256=sAYqzBXYSVBKPTQVYrd7ZmtqMAr1zP9dVJwjfNGW7hc,4207
 phoenix/db/helpers.py,sha256=daKbpY2QhTPo9a_T1xNHKI4WzWHkMmmrGIws7Hw-RZ4,4884
 phoenix/db/migrate.py,sha256=lf6FFGzdyNjwtQ4TFi3mmHNGnnazMtG30PPB2qRF1Xc,2648
 phoenix/db/models.py,sha256=wx2XQgf6qhUO2ZQ4oUENoAguf440qaR0Ea_q1XB1bBU,26484
@@ -94,7 +94,7 @@ phoenix/server/api/exceptions.py,sha256=TA0JuY2YRnj35qGuMSQ8d0ToHum9gWm9W--3fSKH
 phoenix/server/api/interceptor.py,sha256=ykDnoC_apUd-llVli3m1CW18kNSIgjz2qZ6m5JmPDu8,1294
 phoenix/server/api/queries.py,sha256=4KJz8TUz3VUTup9MDjr_GoKX0SttWSvHBq2ncWZGxf8,27343
 phoenix/server/api/schema.py,sha256=tHyw2jTbue_-gu0fe9Sw7LUYtzJUCwp9SvccDgOkNPw,1696
-phoenix/server/api/subscriptions.py,sha256=uTGlf6WKrnC6Nogsx_GtIVSMOzgIKMJwyP1oe47QCHg,19688
+phoenix/server/api/subscriptions.py,sha256=HPb5agJPMPBM_v_sHjiwe7hwMAHsBsrKZZMLq5bxhcw,22062
 phoenix/server/api/utils.py,sha256=quCBRcusc6PUq9tJq7M8PgwFZp7nXgVAxtbw8feribY,833
 phoenix/server/api/dataloaders/__init__.py,sha256=jNYvfXjnZzgA2HWTG7AZdqWGla3ZysBUDUei8Zkz6N8,3290
 phoenix/server/api/dataloaders/annotation_summaries.py,sha256=2sHmIDX7n8tuPeBTs9bMKtlMKWn_Ph9awTZqmwn2Owc,5505
@@ -125,9 +125,9 @@ phoenix/server/api/dataloaders/cache/__init__.py,sha256=SYoOM9n8FJaMdQarma5d1blu
 phoenix/server/api/dataloaders/cache/two_tier_cache.py,sha256=cmo8FUT3E91R139IEzh4yCga-6nTamc5KPXAfMrzNDM,2315
 phoenix/server/api/helpers/__init__.py,sha256=m2-xaSPqUiSs91k62JaRDjFNfl-1byxBfY-m_Vxw16U,272
 phoenix/server/api/helpers/dataset_helpers.py,sha256=14mldZp9to3rr9BdvvoFqEwZHHV_k2e7jPm8q9z2OdQ,6896
-phoenix/server/api/helpers/playground_clients.py,sha256=Ah-f8jDr3uFC-MtofG1k9f4WYP_Lpb5eHxYaK154eRA,36471
+phoenix/server/api/helpers/playground_clients.py,sha256=fTWY883qVNCK_IkiYm1yxFx3GaXMWhalMeV0zYzIyUM,37407
 phoenix/server/api/helpers/playground_registry.py,sha256=CPLMziFB2wmr-dfbx7VbzO2f8YIG_k5RftzvGXYGQ1w,2570
-phoenix/server/api/helpers/playground_spans.py,sha256=ecQv7lTBue_vRW_2giIcAyYOEWIwU06C2371Gd92gK4,16253
+phoenix/server/api/helpers/playground_spans.py,sha256=LJjndkVxr6DsfVFerrtMZWlRd5YL5AIbcM2CvFe8ajc,16489
 phoenix/server/api/input_types/AddExamplesToDatasetInput.py,sha256=mIQz0S_z8YdrktKIY6RCvtNJ2yZF9pYvTGgasUsI-54,430
 phoenix/server/api/input_types/AddSpansToDatasetInput.py,sha256=-StSstyMAVrba3tG1U30b-srkKCtu_svflQuSM19iJA,362
 phoenix/server/api/input_types/ChatCompletionInput.py,sha256=g_5ARuwylt-uCVAsGyZPEVtidEQiOhbKakvDQsZumzw,1451
@@ -164,7 +164,7 @@ phoenix/server/api/input_types/UserRoleInput.py,sha256=xxhFe0ITZOgRVEJbVem_W6F1I
 phoenix/server/api/input_types/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/server/api/mutations/__init__.py,sha256=1wClieLNA3_Tin4Ah67rkrQvwSSZAdPU0EPsRiUxyAA,1103
 phoenix/server/api/mutations/api_key_mutations.py,sha256=OmPipsmlQIb6DKvAgO58mZUwkYJihlJB2N4lTyeUlAA,6164
-phoenix/server/api/mutations/chat_mutations.py,sha256=aQ9dBgUmeQOT2fn8dJp6luWc4EiJJeg1QsX8yNNr_ZM,21074
+phoenix/server/api/mutations/chat_mutations.py,sha256=Rky_PwILAhDBjW6K2cxfaIVsT5iOP5pjuViYUpoVHWY,22540
 phoenix/server/api/mutations/dataset_mutations.py,sha256=siwsSmozKRIK8ZhPrfWl-GxKCL4lAmRPms862tG2KXY,27064
 phoenix/server/api/mutations/experiment_mutations.py,sha256=p3CoLAa8nFPa3D759Y2A7De_PVJNGOL98mA3HoZBrRQ,3188
 phoenix/server/api/mutations/export_events_mutations.py,sha256=xoDnVWC7eA_8wNQP0-oyiHojyUZ0EhVVSrsAnztetC0,3993
@@ -273,15 +273,15 @@ phoenix/server/static/apple-touch-icon-76x76.png,sha256=CT_xT12I0u2i0WU8JzBZBuOQ
 phoenix/server/static/apple-touch-icon.png,sha256=fOfpjqGpWYbJ0eAurKsyoZP1EAs6ZVooBJ_SGk2ZkDs,3801
 phoenix/server/static/favicon.ico,sha256=bY0vvCKRftemZfPShwZtE93DiiQdaYaozkPGwNFr6H8,34494
 phoenix/server/static/modernizr.js,sha256=mvK-XtkNqjOral-QvzoqsyOMECXIMu5BQwSVN_wcU9c,2564
-phoenix/server/static/.vite/manifest.json,sha256=uCPxIoF7qBxFkaV0H9ur1eYr2sV8go_LqjcguNA6CbQ,1929
-phoenix/server/static/assets/components-BXIz9ZO8.js,sha256=clW5yRW2_5BRWWn3FLFSPhCAVqTsGXJmfUTJpHflHf0,306000
-phoenix/server/static/assets/index-DTut7g1y.js,sha256=iM98orZiM2-8YVnpvZgBgVkteOcvDGZgJo3tWQCxkqo,7290
-phoenix/server/static/assets/pages-B8FpJuXu.js,sha256=y7Q9_wYtuaTanNRZk-OvN1ffiyN_6wCsERnHuqOHHUY,629525
-phoenix/server/static/assets/vendor-BX8_Znqy.js,sha256=M5d1J040pAmV0PhGLJLz3-QQ1m3bogn54ekym9iaAPg,10898641
+phoenix/server/static/.vite/manifest.json,sha256=QVqqb_JM1hcyubZgGH4qh08LUHWKE_FR43j_YdVWk0M,1929
+phoenix/server/static/assets/components-C_HASv83.js,sha256=7xZ3sMiHZkpbHJjpzujdMII4znFs-h5DkFCvY8oX1Aw,306769
+phoenix/server/static/assets/index-D7UiCRtr.js,sha256=i6WbtADGyqEQix4x9_4yIbl1U7dc0NpIRoI5ckFGUZ8,7290
+phoenix/server/static/assets/pages-DYHcAdjT.js,sha256=aUloiI7AopSeUGfKhYmalMk1luUrQpMJYHnjZwLV9vs,632400
+phoenix/server/static/assets/vendor-BCxsh5i3.js,sha256=fPeA9hYHGj1n2Bg8JOlpsddXw7Vkulbvra3z4RCy0Ss,10899238
 phoenix/server/static/assets/vendor-DxkFTwjz.css,sha256=nZrkr0u6NNElFGvpWHk9GTHeGoibCXCli1bE7mXZGZg,1816
-phoenix/server/static/assets/vendor-arizeai-CtHir-Ua.js,sha256=mb8CF7PUkSisMt5HjNC2_HZd48HST1LLCOrUnxZ63iI,307000
-phoenix/server/static/assets/vendor-codemirror-DLlGiguX.js,sha256=0bqHkVJOI4zakkaJLSppKz87jnbPhu3c7my6qqYFycE,392709
-phoenix/server/static/assets/vendor-recharts-CJRple0d.js,sha256=YiimiK6M9fdba3tItYH1sl1-Ye61dM67qfAdWpnBjq8,282859
+phoenix/server/static/assets/vendor-arizeai-C2CDZgMz.js,sha256=Isrlf-QaErzG0UZGd2VeCf1l-d6eghpvLr4v4K5A2Xk,308483
+phoenix/server/static/assets/vendor-codemirror-DYbtnCTn.js,sha256=-xGEPOe20e89kdfMNfTrthj084muiJ3gMKPdUtzdjmA,392709
+phoenix/server/static/assets/vendor-recharts-P6W8G0Mb.js,sha256=0B8yyQbqMwwsPvLkd-tZzntAKMNZpZGJ8OlMfXKvbu4,282859
 phoenix/server/static/assets/vendor-three-DwGkEfCM.js,sha256=0D12ZgKzfKCTSdSTKJBFR2RZO_xxeMXrqDp0AszZqHY,620972
 phoenix/server/templates/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/server/templates/index.html,sha256=ram6sfy2obf_F053ay35V30v-mnRWZ86rK-PstXLy1c,4457
@@ -299,7 +299,7 @@ phoenix/trace/fixtures.py,sha256=bwDD0rgIRdAi_9zwLLipke50oYuKJvZLvhiGsuvnGbs,191
 phoenix/trace/otel.py,sha256=FZGzo0NPJqFBLkL_VmQeBne2z-LdHNPTz3oIZOuPPzQ,9997
 phoenix/trace/projects.py,sha256=9dKv1aiKL4IYMFsg2xnC6EOIRO0YHtkR5o9ALHbMK9g,2178
 phoenix/trace/schemas.py,sha256=0tghsJDaLKvSuxDDQGpUEQ0HkMsKmFWPO3lwJrJ9t84,5972
-phoenix/trace/span_evaluations.py,sha256=9Z9i7KE7qCIwDb9nHmZp7UvDy_EP5rZmP4VP5h2GRNQ,13107
+phoenix/trace/span_evaluations.py,sha256=x3nye9r2SQk1mrR3N05YbuWsgUKpMWwTRBtJTDBSj3Y,13156
 phoenix/trace/span_json_decoder.py,sha256=J1_oDViuUoC4vxPg61U4EOZC1uEMcCzoj-kVjOFEE8k,3224
 phoenix/trace/span_json_encoder.py,sha256=Wa7RvQTZSsFG4pdKSSFSO_4q4DjAOQuG8wLNJvYzsfM,2004
 phoenix/trace/trace_dataset.py,sha256=7XyCe_7WemBmAiHw0kigmwkC1gPSLeVBqtrRe9FpM6s,14490
@@ -316,15 +316,15 @@ phoenix/utilities/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU
 phoenix/utilities/client.py,sha256=kr95EgpTJtCbb7CLoxC-KiyB2Wco03DZGIlcEmozHFs,5605
 phoenix/utilities/deprecation.py,sha256=LL1LWfVqnYulceMQw2CggoIv8Uogrj7fAD_85devnoo,1067
 phoenix/utilities/error_handling.py,sha256=5_ggfwmd0xZPoI3qi1fwm0x1F0i7g5j-gr0ny1TFzPU,1967
-phoenix/utilities/json.py,sha256=F5UbCaplDxrcSiuMMPIwPqcxEmYcbc24Jqomqky8YbA,3565
+phoenix/utilities/json.py,sha256=UE8WHZU060NY_8ZKJ97gmtEC2mp60d8tJFXyRqIz51Y,3743
 phoenix/utilities/logging.py,sha256=NKvX43jWITgSTtHGCi6l5xp8jTjtQ8FQ_irEi9ybOxg,502
 phoenix/utilities/project.py,sha256=auVpARXkDb-JgeX5f2aStyFIkeKvGwN9l7qrFeJMVxI,445
 phoenix/utilities/re.py,sha256=x8Xbk-Wa6qDMAtUd_7JtZvKtrYEuMY-bchB0n163_5c,2006
 phoenix/utilities/span_store.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/utilities/template_formatters.py,sha256=JuOyvukMPLDHa1uVNw0kCFBUnIxy02dwAWNZimdIZU4,2423
-arize_phoenix-5.10.0.dist-info/METADATA,sha256=MYCUDgCZbLC0v04dTkj7SxbQ_hbJZ2FHJS42vzg-Kv4,22614
-arize_phoenix-5.10.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
-arize_phoenix-5.10.0.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
-arize_phoenix-5.10.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
-arize_phoenix-5.10.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
-arize_phoenix-5.10.0.dist-info/RECORD,,
+arize_phoenix-5.11.0.dist-info/METADATA,sha256=6ALW-ZaTA2y2Temcz899qgQfCOjqJVGx9A2jgWPkkOI,22671
+arize_phoenix-5.11.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
+arize_phoenix-5.11.0.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
+arize_phoenix-5.11.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
+arize_phoenix-5.11.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
+arize_phoenix-5.11.0.dist-info/RECORD,,

phoenix/config.py

@@ -97,6 +97,13 @@ ENV_PHOENIX_SERVER_INSTRUMENTATION_OTLP_TRACE_COLLECTOR_GRPC_ENDPOINT = (
 ENV_PHOENIX_ENABLE_AUTH = "PHOENIX_ENABLE_AUTH"
 ENV_PHOENIX_DISABLE_RATE_LIMIT = "PHOENIX_DISABLE_RATE_LIMIT"
 ENV_PHOENIX_SECRET = "PHOENIX_SECRET"
+ENV_PHOENIX_DEFAULT_ADMIN_INITIAL_PASSWORD = "PHOENIX_DEFAULT_ADMIN_INITIAL_PASSWORD"
+"""
+The initial password for the default admin account, which defaults to ‘admin’ if not
+explicitly set. Note that changing this value will have no effect if the default admin
+record already exists in the database. In such cases, the default admin password must
+be updated manually in the application.
+"""
 ENV_PHOENIX_API_KEY = "PHOENIX_API_KEY"
 ENV_PHOENIX_USE_SECURE_COOKIES = "PHOENIX_USE_SECURE_COOKIES"
 ENV_PHOENIX_ACCESS_TOKEN_EXPIRY_MINUTES = "PHOENIX_ACCESS_TOKEN_EXPIRY_MINUTES"
@@ -274,6 +281,12 @@ def get_env_phoenix_secret() -> Optional[str]:
     return phoenix_secret
 
 
+def get_env_default_admin_initial_password() -> str:
+    from phoenix.auth import DEFAULT_ADMIN_PASSWORD
+
+    return os.environ.get(ENV_PHOENIX_DEFAULT_ADMIN_INITIAL_PASSWORD) or DEFAULT_ADMIN_PASSWORD
+
+
 def get_env_phoenix_use_secure_cookies() -> bool:
     return _bool_val(ENV_PHOENIX_USE_SECURE_COOKIES, False)
 

phoenix/db/facilitator.py

@@ -13,13 +13,13 @@ from sqlalchemy.ext.asyncio import AsyncSession
 
 from phoenix.auth import (
     DEFAULT_ADMIN_EMAIL,
-    DEFAULT_ADMIN_PASSWORD,
     DEFAULT_ADMIN_USERNAME,
     DEFAULT_SECRET_LENGTH,
     DEFAULT_SYSTEM_EMAIL,
     DEFAULT_SYSTEM_USERNAME,
     compute_password_hash,
 )
+from phoenix.config import get_env_default_admin_initial_password
 from phoenix.db import models
 from phoenix.db.enums import COLUMN_ENUMS, UserRole
 from phoenix.server.types import DbSessionFactory
@@ -97,7 +97,8 @@ async def _ensure_user_roles(session: AsyncSession) -> None:
         admin_role_id := role_ids.get(admin_role)
     ) is not None:
         salt = secrets.token_bytes(DEFAULT_SECRET_LENGTH)
-        compute = partial(compute_password_hash, password=DEFAULT_ADMIN_PASSWORD, salt=salt)
+        password = get_env_default_admin_initial_password()
+        compute = partial(compute_password_hash, password=password, salt=salt)
         loop = asyncio.get_running_loop()
         hash_ = await loop.run_in_executor(None, compute)
         admin_user = models.User(

phoenix/server/api/helpers/playground_clients.py

@@ -2,6 +2,7 @@ import asyncio
 import importlib.util
 import inspect
 import json
+import os
 import time
 from abc import ABC, abstractmethod
 from collections.abc import AsyncIterator, Callable, Iterator
@@ -25,6 +26,7 @@ from phoenix.evals.models.rate_limiters import (
     RateLimiter,
     RateLimitError,
 )
+from phoenix.server.api.exceptions import BadRequest
 from phoenix.server.api.helpers.playground_registry import PROVIDER_DEFAULT, register_llm_client
 from phoenix.server.api.input_types.GenerativeModelInput import GenerativeModelInput
 from phoenix.server.api.input_types.InvocationParameters import (
@@ -99,9 +101,9 @@ class PlaygroundRateLimiter(RateLimiter, KeyedSingleton):
         super().__init__(
             rate_limit_error=rate_limit_error,
             max_rate_limit_retries=3,
-            initial_per_second_request_rate=2.0,
-            maximum_per_second_request_rate=10.0,
-            enforcement_window_minutes=1,
+            initial_per_second_request_rate=1.0,
+            maximum_per_second_request_rate=3.0,
+            enforcement_window_minutes=0.05,
             rate_reduction_factor=0.5,
             rate_increase_factor=0.01,
             cooldown_seconds=5,
@@ -128,10 +130,11 @@ class PlaygroundRateLimiter(RateLimiter, KeyedSingleton):
                     self._rate_limit_handling.set()  # Set the event as a failsafe
                 await self._throttler.async_wait_until_ready()
                 request_start_time = time.time()
-                if inspect.iscoroutinefunction(fn):
-                    return await fn(*args, **kwargs)  # type: ignore
+                maybe_coroutine = fn(*args, **kwargs)
+                if inspect.iscoroutine(maybe_coroutine):
+                    return await maybe_coroutine  # type: ignore
                 else:
-                    return fn(*args, **kwargs)
+                    return maybe_coroutine
             except self._rate_limit_error:
                 async with self._rate_limit_handling_lock:
                     self._rate_limit_handling.clear()  # prevent new requests from starting
@@ -258,6 +261,10 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
         from openai import AsyncOpenAI
         from openai import RateLimitError as OpenAIRateLimitError
 
+        # todo: check if custom base url is set before raising error to allow
+        # for custom endpoints that don't require an API key
+        if not (api_key := api_key or os.environ.get("OPENAI_API_KEY")):
+            raise BadRequest("An API key is required for OpenAI models")
         super().__init__(model=model, api_key=api_key)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.OPENAI.value
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
@@ -341,7 +348,7 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
         openai_messages = [self.to_openai_chat_completion_param(*message) for message in messages]
         tool_call_ids: dict[int, str] = {}
         token_usage: Optional["CompletionUsage"] = None
-        throttled_create = self.rate_limiter.alimit(self.client.chat.completions.create)
+        throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
         async for chunk in await throttled_create(
             messages=openai_messages,
             model=self.model_name,
@@ -510,7 +517,7 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
 
         tool_call_ids: dict[int, str] = {}
 
-        throttled_create = self.rate_limiter.alimit(self.client.chat.completions.create)
+        throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
         response = await throttled_create(
             messages=openai_messages,
             model=self.model_name,
@@ -618,12 +625,14 @@ class AzureOpenAIStreamingClient(OpenAIStreamingClient):
         super().__init__(model=model, api_key=api_key)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.AZURE.value
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
-        if model.endpoint is None or model.api_version is None:
-            raise ValueError("endpoint and api_version are required for Azure OpenAI models")
+        if not (endpoint := model.endpoint or os.environ.get("AZURE_OPENAI_ENDPOINT")):
+            raise BadRequest("An Azure endpoint is required for Azure OpenAI models")
+        if not (api_version := model.api_version or os.environ.get("OPENAI_API_VERSION")):
+            raise BadRequest("An OpenAI API version is required for Azure OpenAI models")
         self.client = AsyncAzureOpenAI(
             api_key=api_key,
-            azure_endpoint=model.endpoint,
-            api_version=model.api_version,
+            azure_endpoint=endpoint,
+            api_version=api_version,
         )
 
 
@@ -648,6 +657,8 @@ class AnthropicStreamingClient(PlaygroundStreamingClient):
         super().__init__(model=model, api_key=api_key)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.ANTHROPIC.value
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.ANTHROPIC.value
+        if not (api_key := api_key or os.environ.get("ANTHROPIC_API_KEY")):
+            raise BadRequest("An API key is required for Anthropic models")
         self.client = anthropic.AsyncAnthropic(api_key=api_key)
         self.model_name = model.name
         self.rate_limiter = PlaygroundRateLimiter(model.provider_key, anthropic.RateLimitError)
@@ -707,7 +718,6 @@ class AnthropicStreamingClient(PlaygroundStreamingClient):
             "messages": anthropic_messages,
             "model": self.model_name,
             "system": system_prompt,
-            "max_tokens": 1024,
             "tools": tools,
             **invocation_parameters,
         }
@@ -820,6 +830,12 @@ class GeminiStreamingClient(PlaygroundStreamingClient):
         super().__init__(model=model, api_key=api_key)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.GOOGLE.value
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.VERTEXAI.value
+        if not (
+            api_key := api_key
+            or os.environ.get("GEMINI_API_KEY")
+            or os.environ.get("GOOGLE_API_KEY")
+        ):
+            raise BadRequest("An API key is required for Gemini models")
         google_genai.configure(api_key=api_key)
         self.model_name = model.name
 

phoenix/server/api/helpers/playground_spans.py

@@ -379,6 +379,11 @@ def _llm_output_messages(
     if content := "".join(chunk.content for chunk in text_chunks):
         yield f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_CONTENT}", content
     for tool_call_index, (_tool_call_id, tool_call_chunks_) in enumerate(tool_call_chunks.items()):
+        if _tool_call_id:
+            yield (
+                f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_TOOL_CALLS}.{tool_call_index}.{TOOL_CALL_ID}",
+                _tool_call_id,
+            )
         if tool_call_chunks_ and (name := tool_call_chunks_[0].function.name):
             yield (
                 f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_TOOL_CALLS}.{tool_call_index}.{TOOL_CALL_FUNCTION_NAME}",
@@ -441,6 +446,7 @@ MESSAGE_CONTENT = MessageAttributes.MESSAGE_CONTENT
 MESSAGE_ROLE = MessageAttributes.MESSAGE_ROLE
 MESSAGE_TOOL_CALLS = MessageAttributes.MESSAGE_TOOL_CALLS
 
+TOOL_CALL_ID = ToolCallAttributes.TOOL_CALL_ID
 TOOL_CALL_FUNCTION_NAME = ToolCallAttributes.TOOL_CALL_FUNCTION_NAME
 TOOL_CALL_FUNCTION_ARGUMENTS_JSON = ToolCallAttributes.TOOL_CALL_FUNCTION_ARGUMENTS_JSON
 TOOL_CALL_ID = ToolCallAttributes.TOOL_CALL_ID

phoenix/server/api/mutations/chat_mutations.py

@@ -1,9 +1,9 @@
 import asyncio
 from dataclasses import asdict, field
 from datetime import datetime, timezone
-from itertools import chain
+from itertools import chain, islice
 from traceback import format_exc
-from typing import Any, Iterable, Iterator, List, Optional, Union
+from typing import Any, Iterable, Iterator, List, Optional, TypeVar, Union
 
 import strawberry
 from openinference.instrumentation import safe_json_dumps
@@ -26,7 +26,7 @@ from phoenix.datetime_utils import local_now, normalize_datetime
 from phoenix.db import models
 from phoenix.db.helpers import get_dataset_example_revisions
 from phoenix.server.api.context import Context
-from phoenix.server.api.exceptions import BadRequest, NotFound
+from phoenix.server.api.exceptions import BadRequest, CustomGraphQLError, NotFound
 from phoenix.server.api.helpers.playground_clients import (
     PlaygroundStreamingClient,
     initialize_playground_clients,
@@ -127,11 +127,19 @@ class ChatCompletionMutationMixin:
         provider_key = input.model.provider_key
         llm_client_class = PLAYGROUND_CLIENT_REGISTRY.get_client(provider_key, input.model.name)
         if llm_client_class is None:
-            raise BadRequest(f"No LLM client registered for provider '{provider_key}'")
-        llm_client = llm_client_class(
-            model=input.model,
-            api_key=input.api_key,
-        )
+            raise BadRequest(f"Unknown LLM provider: '{provider_key.value}'")
+        try:
+            llm_client = llm_client_class(
+                model=input.model,
+                api_key=input.api_key,
+            )
+        except CustomGraphQLError:
+            raise
+        except Exception as error:
+            raise BadRequest(
+                f"Failed to connect to LLM API for {provider_key.value} {input.model.name}: "
+                f"{str(error)}"
+            )
         dataset_id = from_global_id_with_expected_type(input.dataset_id, Dataset.__name__)
         dataset_version_id = (
             from_global_id_with_expected_type(
@@ -158,7 +166,9 @@ class ChatCompletionMutationMixin:
             revisions = [
                 revision
                 async for revision in await session.stream_scalars(
-                    get_dataset_example_revisions(resolved_version_id)
+                    get_dataset_example_revisions(resolved_version_id).order_by(
+                        models.DatasetExampleRevision.id
+                    )
                 )
             ]
             if not revisions:
@@ -181,28 +191,32 @@ class ChatCompletionMutationMixin:
             session.add(experiment)
             await session.flush()
 
+        results = []
+        batch_size = 3
         start_time = datetime.now(timezone.utc)
-        results = await asyncio.gather(
-            *(
-                cls._chat_completion(
-                    info,
-                    llm_client,
-                    ChatCompletionInput(
-                        model=input.model,
-                        api_key=input.api_key,
-                        messages=input.messages,
-                        tools=input.tools,
-                        invocation_parameters=input.invocation_parameters,
-                        template=TemplateOptions(
-                            language=input.template_language,
-                            variables=revision.input,
+        for batch in _get_batches(revisions, batch_size):
+            batch_results = await asyncio.gather(
+                *(
+                    cls._chat_completion(
+                        info,
+                        llm_client,
+                        ChatCompletionInput(
+                            model=input.model,
+                            api_key=input.api_key,
+                            messages=input.messages,
+                            tools=input.tools,
+                            invocation_parameters=input.invocation_parameters,
+                            template=TemplateOptions(
+                                language=input.template_language,
+                                variables=revision.input,
+                            ),
                         ),
-                    ),
-                )
-                for revision in revisions
-            ),
-            return_exceptions=True,
-        )
+                    )
+                    for revision in batch
+                ),
+                return_exceptions=True,
+            )
+            results.extend(batch_results)
 
         payload = ChatCompletionOverDatasetMutationPayload(
             dataset_id=GlobalID(models.Dataset.__name__, str(dataset.id)),
@@ -266,11 +280,19 @@ class ChatCompletionMutationMixin:
         provider_key = input.model.provider_key
         llm_client_class = PLAYGROUND_CLIENT_REGISTRY.get_client(provider_key, input.model.name)
         if llm_client_class is None:
-            raise BadRequest(f"No LLM client registered for provider '{provider_key}'")
-        llm_client = llm_client_class(
-            model=input.model,
-            api_key=input.api_key,
-        )
+            raise BadRequest(f"Unknown LLM provider: '{provider_key.value}'")
+        try:
+            llm_client = llm_client_class(
+                model=input.model,
+                api_key=input.api_key,
+            )
+        except CustomGraphQLError:
+            raise
+        except Exception as error:
+            raise BadRequest(
+                f"Failed to connect to LLM API for {provider_key.value} {input.model.name}: "
+                f"{str(error)}"
+            )
         return await cls._chat_completion(info, llm_client, input)
 
     @classmethod
@@ -486,6 +508,11 @@ def _llm_output_messages(
     if text_content:
         yield f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_CONTENT}", text_content
     for tool_call_index, tool_call in enumerate(tool_calls.values()):
+        if tool_call_id := tool_call.id:
+            yield (
+                f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_TOOL_CALLS}.{tool_call_index}.{TOOL_CALL_ID}",
+                tool_call_id,
+            )
         yield (
             f"{LLM_OUTPUT_MESSAGES}.0.{MESSAGE_TOOL_CALLS}.{tool_call_index}.{TOOL_CALL_FUNCTION_NAME}",
             tool_call.function.name,
@@ -513,6 +540,19 @@ def _serialize_event(event: SpanException) -> dict[str, Any]:
     return {k: (v.isoformat() if isinstance(v, datetime) else v) for k, v in asdict(event).items()}
 
 
+_AnyT = TypeVar("_AnyT")
+
+
+def _get_batches(
+    iterable: Iterable[_AnyT],
+    batch_size: int,
+) -> Iterator[list[_AnyT]]:
+    """Splits an iterable into batches not exceeding a specified size."""
+    iterator = iter(iterable)
+    while batch := list(islice(iterator, batch_size)):
+        yield batch
+
+
 JSON = OpenInferenceMimeTypeValues.JSON.value
 TEXT = OpenInferenceMimeTypeValues.TEXT.value
 LLM = OpenInferenceSpanKindValues.LLM.value
@@ -534,6 +574,7 @@ MESSAGE_CONTENT = MessageAttributes.MESSAGE_CONTENT
 MESSAGE_ROLE = MessageAttributes.MESSAGE_ROLE
 MESSAGE_TOOL_CALLS = MessageAttributes.MESSAGE_TOOL_CALLS
 
+TOOL_CALL_ID = ToolCallAttributes.TOOL_CALL_ID
 TOOL_CALL_FUNCTION_NAME = ToolCallAttributes.TOOL_CALL_FUNCTION_NAME
 TOOL_CALL_FUNCTION_ARGUMENTS_JSON = ToolCallAttributes.TOOL_CALL_FUNCTION_ARGUMENTS_JSON