arize-phoenix 2.0.0__py3-none-any.whl → 2.2.0rc0__py3-none-any.whl

{arize_phoenix-2.0.0.dist-info → arize_phoenix-2.2.0rc0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: arize-phoenix
-Version: 2.0.0
+Version: 2.2.0rc0
 Summary: ML Observability in your notebook
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -20,6 +20,8 @@ Requires-Dist: ddsketch
 Requires-Dist: hdbscan<1.0.0,>=0.8.33
 Requires-Dist: jinja2
 Requires-Dist: numpy
+Requires-Dist: opentelemetry-proto
+Requires-Dist: opentelemetry-sdk
 Requires-Dist: pandas
 Requires-Dist: protobuf<5.0,>=3.20
 Requires-Dist: psutil
@@ -36,8 +38,10 @@ Requires-Dist: umap-learn
 Requires-Dist: uvicorn
 Requires-Dist: wrapt
 Provides-Extra: dev
+Requires-Dist: anthropic; extra == 'dev'
 Requires-Dist: arize[autoembeddings,llm-evaluation]; extra == 'dev'
 Requires-Dist: gcsfs; extra == 'dev'
+Requires-Dist: google-cloud-aiplatform>=1.3; extra == 'dev'
 Requires-Dist: hatch; extra == 'dev'
 Requires-Dist: jupyter; extra == 'dev'
 Requires-Dist: langchain>=0.0.334; extra == 'dev'

{arize_phoenix-2.0.0.dist-info → arize_phoenix-2.2.0rc0.dist-info}/RECORD

@@ -1,16 +1,17 @@
-phoenix/__init__.py,sha256=PMI2WOSvZVHGPpuhM_CPM29ejrA1b-XOHI8jJXHWHeE,1373
+phoenix/__init__.py,sha256=EEh0vZGRQS8686h34GQ64OjQoZ7neKYO_iO5j6Oa9Jw,1402
 phoenix/config.py,sha256=ErvGg22SSiuqPJtIX1WZE5KcM2lt6XOGZ__HwRg3JqA,2390
 phoenix/datetime_utils.py,sha256=D955QLrkgrrSdUM6NyqbCeAu2SMsjhR5rHVQEsVUdng,2773
 phoenix/exceptions.py,sha256=igIWGAg3m8jm5YwQDeCY1p8ml_60A7zaGVXJ1yZhY9s,44
 phoenix/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
 phoenix/services.py,sha256=slL4Uu___QQSKEssgD738-WAld-kzVQnpW92uKLxV4E,4886
+phoenix/version.py,sha256=pG4VqwySwU54SQ_mHFbajVD0oK3-38mb_fLPwWYLdoc,25
 phoenix/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/core/embedding_dimension.py,sha256=zKGbcvwOXgLf-yrJBpQyKtd-LEOPRKHnUToyAU8Owis,87
-phoenix/core/evals.py,sha256=tBHVgEEQW6zsNAAQCYIN-BbVHy5Q-hgF5UpQy1SmkwE,6538
+phoenix/core/evals.py,sha256=OrHeYlh804rpcZIXTA6kan2mzSZMfgpphNNQdPMpNoM,7597
 phoenix/core/model.py,sha256=vQ6RxpUPlncezJvur5u6xBN0Lkrk2gW0cTyb-qqaSqA,4713
 phoenix/core/model_schema.py,sha256=rR9VdhL_oXxbprDTPQJBXs5hw5sMPQmzx__m6Kwsxug,50394
 phoenix/core/model_schema_adapter.py,sha256=3GkyzqUST4fYi-Bgs8qAam5hwMCdQRZTDLjZ9Bnzdm4,8268
-phoenix/core/traces.py,sha256=uNhtCpgXP9L0ahOoP9ZOBmLnubyUTkICiDB4XB-h5GY,14312
+phoenix/core/traces.py,sha256=O01L6qwQfHxHUHNZemKBBsAgqDo1tAIO5-1fK2g0NwE,14618
 phoenix/datasets/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/datasets/dataset.py,sha256=scKVZ7zc6Dpc_ntt-pWhzY-KWqOJEwKePuyNnKSVTGE,30515
 phoenix/datasets/errors.py,sha256=cGp9vxnw4SewFoWBV3ZGMkhE0Kh73lPIv3Ppz_H_RoA,8261
@@ -18,25 +19,27 @@ phoenix/datasets/fixtures.py,sha256=0_PacL3dw49zulKpFpPdhvxJxeGmHTguqIyf2VXkBkk,
 phoenix/datasets/schema.py,sha256=bF1d2Md6NyqQZuC4Ym5A52f2_IcazkyxGFZ11HPqSg0,6668
 phoenix/datasets/validation.py,sha256=dZ9lCFUV0EY7HCkQkQBrs-GLAEIZdpOqUxwD5l4dp88,8294
 phoenix/experimental/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-phoenix/experimental/evals/__init__.py,sha256=90k7hxxghhqDCftoAWsdOiUsJqYwNuJBXjhbpb3Na4I,1499
-phoenix/experimental/evals/evaluators.py,sha256=JEEow9CM3u9GkVey8KEora8e7cbmhJgazd5wh35_LPE,10311
+phoenix/experimental/evals/__init__.py,sha256=I-e_QhT3ezMwc0WeqZuTNbuHkdaLj9WvnFCv6aNAmYM,1595
+phoenix/experimental/evals/evaluators.py,sha256=rLvvXBK2H_cjJyRMBQStTlMYntTJI3RtukwJopeA1jU,13065
 phoenix/experimental/evals/retrievals.py,sha256=o3fqrsYbYZjyGj_jWkN_9VQVyXjLkDKDw5Ws7l8bwdI,3828
 phoenix/experimental/evals/functions/__init__.py,sha256=3FMGrjmgxegXAwgDV_RpaN-73cFVyBiO8YwZvml5P9c,156
-phoenix/experimental/evals/functions/classify.py,sha256=Q0FMOpU6WFr4m3dtBpJlNyrj2ExTYMB-XjR1EreeeXg,18731
-phoenix/experimental/evals/functions/executor.py,sha256=TSw2lVhkl6-VBYcXSlUl3E0U4OuJWnhwhed7NmFgbF8,13376
+phoenix/experimental/evals/functions/classify.py,sha256=uCTZR_ctQorzS0Abcwxzsza0g-4q_91DHiObjJISIXE,18177
+phoenix/experimental/evals/functions/executor.py,sha256=bM7PI2rcPukQQzZ2rWqN_-Kfo_a935YJj0bh1Red8Ps,13406
 phoenix/experimental/evals/functions/generate.py,sha256=sdr6TeXn5JLEKM0NqYtvq01Lq48Q7uatb0fsq5zQgVY,5310
 phoenix/experimental/evals/functions/processing.py,sha256=F4xtLsulLV4a8CkuLldRddsCim75dSTIShEJUYN6I6w,1823
 phoenix/experimental/evals/models/__init__.py,sha256=j1N7DhiOPbcaemtVBONcQ0miNnGQwEXz4u3P3Vwe6-4,320
+phoenix/experimental/evals/models/anthropic.py,sha256=Tcv8R-vTyY8sLAv1wIHeZdMCBtqhyayqMPJXRDc7blI,6267
 phoenix/experimental/evals/models/base.py,sha256=aSE3Al3MsLvzNKuN2e-z6O-RB5mgpisH4UQqwNQcqp0,7734
 phoenix/experimental/evals/models/bedrock.py,sha256=CRPmBuSLc_nRnKKWLHhGMxdWEISIKUJM1tzIlOQ_qWM,7927
 phoenix/experimental/evals/models/litellm.py,sha256=jrRlph22xWxMXMUabUWjIO2e-sHxQzlQwSM-SnAACFQ,4714
 phoenix/experimental/evals/models/openai.py,sha256=Kl2uES3HRcZGFqblfBQZ6D1BpDffuLZDAqVTjhrSXXQ,17101
 phoenix/experimental/evals/models/rate_limiters.py,sha256=5GVN0RQKt36Przg3-9jLgocRmyg-tbeO-cdbuLIx89w,10160
+phoenix/experimental/evals/models/vertex.py,sha256=nwTIjVn4gGFfoKfGqUGwPD9GLJaBM4HLXDnMNs9hSrw,5407
 phoenix/experimental/evals/models/vertexai.py,sha256=NfBpQq0l7XzP-wDEDsK27IRiQBzA1GXEdfwlAf8leX4,5609
-phoenix/experimental/evals/templates/__init__.py,sha256=-KFIVh1xQLPcJP8Mv_tieIndxE02mg_c2QaibAuHJuE,1432
-phoenix/experimental/evals/templates/default_templates.py,sha256=8BetD8Xufc2f62CPZ9mbysyvDofnbFPalwB0u-KckuE,20614
+phoenix/experimental/evals/templates/__init__.py,sha256=GSJSoWJ4jwyoUANniidmWMUtXQhNQYbTJbfFqCvuYuo,1470
+phoenix/experimental/evals/templates/default_templates.py,sha256=_VVxuhPsY8fkasA9XMNNM_fvftltkIfVCfElSdFbsQY,21056
 phoenix/experimental/evals/templates/template.py,sha256=VAX_ZeV3vNWCODipMy7EtaYdQ0c7WA6H2Mx0i1axXf0,6005
-phoenix/experimental/evals/utils/__init__.py,sha256=PnCsNppI9KRFrQpsKWbLN58FNXsnUA3qqjLwNhoe_6Q,3715
+phoenix/experimental/evals/utils/__init__.py,sha256=608EX7sG0f5oDG__II16J8xnFJiNpY9dI9AC8vXwR00,5601
 phoenix/experimental/evals/utils/threads.py,sha256=ksI-egarPnlxit0qKKjtjZ2L82qGLxqxZ6s92O0eBA4,1005
 phoenix/metrics/README.md,sha256=5gekqTU-5gGdMwvcfNp2Wlu8p1ul9kGY_jq0XXQusoI,1964
 phoenix/metrics/__init__.py,sha256=sLp7td1GIt_0Z8dPUyP4L0-_4x9c871yAaGX30oMsvg,2433
@@ -52,16 +55,17 @@ phoenix/pointcloud/pointcloud.py,sha256=ms-h1FLC0xXb3sk256zpSuZQDE2hdOAJzRNBklP0
 phoenix/pointcloud/projectors.py,sha256=zO_RrtDYSv2rqVOfIP2_9Cv11Dc8EmcZR94xhFcBYPU,1057
 phoenix/pointcloud/umap_parameters.py,sha256=lJsEOrbSuSiqI7g4Yt6xj7kgYxEqoep4ZHWLr6VWBqw,1760
 phoenix/server/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-phoenix/server/app.py,sha256=SEKUBgiXCH6tCXelrzbuwuY_EbDFC8CD7J5W7BkZSzM,7125
+phoenix/server/app.py,sha256=ptm47TRYPj36fNiko8Ja2R40J8HQFlfspSRUIQ1by88,7239
 phoenix/server/evaluation_handler.py,sha256=HzaoD8Cv9HbEdd0nYSTZoakKsE8Ic5lVjeuBh0vnhoA,1554
-phoenix/server/main.py,sha256=cBJ_S_6TQBlV0Uj6N6ps3xq86VIv6nYKR49OYDcborg,6775
-phoenix/server/span_handler.py,sha256=TCwzbvtKfAc3CaI2Ao7Q5cfORWtQ56cU7j-43wgbOBc,1263
+phoenix/server/main.py,sha256=1puvebfgsD3GBHiMn5Dx-dURnby7iMGXK5Uce_KaQG4,6804
+phoenix/server/span_handler.py,sha256=reYUDaN5bavSFjEiSfvYyAG_mpJs6S3iB-RNCkZrSUU,1295
 phoenix/server/thread_server.py,sha256=a9Vnzc69ZLqJbI_FUSOY3eeuYCiCq6aprPj2gS_RB-M,2097
+phoenix/server/trace_handler.py,sha256=pXanrp9L21Mh7MnyJbj202NJ-Rn4bCjG0oL4DtdKcls,2074
 phoenix/server/api/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/server/api/context.py,sha256=02vRgyLFpDCmh97QwsjWD5cdNZkoCUtDPPs1YItbdbI,583
 phoenix/server/api/helpers.py,sha256=_V1eVkchZmTkhOfRC4QqR1sUB2xtIxdsMJkDouZq_IE,251
 phoenix/server/api/interceptor.py,sha256=do_J4HjPPQ_C7bMmqe1YpTmt_hoxcwC2I8P3n5sZBo4,1302
-phoenix/server/api/schema.py,sha256=3yUzUU9C3r-lH7ouxxZOO-8kiEVoLWn_dG-DUjKu78w,14413
+phoenix/server/api/schema.py,sha256=b_GiRJKkfnqR_Fy51N4NWN2nh7clao2V6C8G94nTYo4,15303
 phoenix/server/api/input_types/ClusterInput.py,sha256=EL4ftvZxQ8mVdruUPcdhMhByORmSmM8S-X6RPqU6GX0,179
 phoenix/server/api/input_types/Coordinates.py,sha256=meTwbIjwTfqx5DGD2DBlH9wQzdQVNM5a8x9dp1FfIgA,173
 phoenix/server/api/input_types/DataQualityMetricInput.py,sha256=LazvmQCCM5m9SDZTpyxQXO1rYF4cmsc3lsR2S9S65X4,1292
@@ -121,19 +125,20 @@ 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/index.css,sha256=KKGpx4iwF91VGRm0YN-4cn8oC-oIqC6HecoPf0x3ZM8,1885
-phoenix/server/static/index.js,sha256=hxV2PRT6IFyBLfCFykM1Xi537pefpFiNVeJTcU4YZyI,3216221
+phoenix/server/static/index.js,sha256=gP89MKs13uK-k9A_hQFQM6E03GPltzNoEFgaAHaAgBs,3257264
 phoenix/server/static/modernizr.js,sha256=mvK-XtkNqjOral-QvzoqsyOMECXIMu5BQwSVN_wcU9c,2564
 phoenix/server/templates/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-phoenix/server/templates/index.html,sha256=UKAw1N5ysyOYiFgP5Hfd4oaSqc_Y7quEtC9FGvtsIBA,1425
+phoenix/server/templates/index.html,sha256=_ZVGz2JYDgJjCL2kxSbVjN1qY7drKYerVIXNfmcW30k,1765
 phoenix/session/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/session/evaluation.py,sha256=88wSWlUi71RDHZksRE3hG91GHq8rMRYXirkyA3IbM8Y,4681
-phoenix/session/session.py,sha256=35Hf6pmSD0HbXwC7i7eBTnSUfgMIVAJRgi3YwyZ7uXY,17769
+phoenix/session/session.py,sha256=WL0lUGnBp9ZJHOwkLPkf32QvmHOThC5VIERa4SP0o_E,17815
 phoenix/trace/__init__.py,sha256=lnuxATMemAqjURYqOfIo_HyCo5oIWIVTy98XAsiS1d8,215
 phoenix/trace/evaluation_conventions.py,sha256=t8jydM3U0-T5YpiQKRJ3tWdWGlHtzKyttYdw-ddvPOk,1048
-phoenix/trace/exporter.py,sha256=jPZxjCMsil6mfuzFybmZfdCO49Odr8r916zcYUht-Qg,4293
+phoenix/trace/exporter.py,sha256=z3xrGJhIRh7XMy4Q1FkR3KmFZym-GX0XxLTZ6eSnN0Q,4347
 phoenix/trace/fixtures.py,sha256=lFuhPe-di54BmCT-RxS95m3e2-z1YBOo8CxsqYSgDD8,6341
+phoenix/trace/otel.py,sha256=4cJ85O_y_S6C_kkVvYh8d1d1mzKpGk_eQKjg688PI0k,13899
 phoenix/trace/schemas.py,sha256=m1wVlYFT6qL3FovD3TtTYsEgN6OHvv52gNdJkoPCmuY,5400
-phoenix/trace/semantic_conventions.py,sha256=E51V06dT4ER0ZD3efinNDbu1aOHfIotve7fri5gl5bg,4599
+phoenix/trace/semantic_conventions.py,sha256=u6NG85ZhbreriZr8cqJaddldM_jUcew7JilszY7JUk8,4652
 phoenix/trace/span_evaluations.py,sha256=9RTJ8BFhXDJNtqErWRlMj65FG7wJiI41YTgB7vYLqcY,8429
 phoenix/trace/span_json_decoder.py,sha256=Xv-0uCsHgwzQb0dqTa7CuuDeXAPaXjQICyCFK3ZQaSs,3089
 phoenix/trace/span_json_encoder.py,sha256=C5y7rkyOcV08oJC5t8TZqVxsKCZMJKad7bBQzAgLoDs,1763
@@ -147,24 +152,21 @@ phoenix/trace/dsl/missing.py,sha256=BWPOHr2_tBkPDgVeq8GVXXVbNbJiBelu4NtwHBg6mTE,
 phoenix/trace/dsl/query.py,sha256=BAdL5rcliatBtxpcuZ86am5mmSUkhzCMF3PRSH8v3m8,9890
 phoenix/trace/langchain/__init__.py,sha256=vAjrmrreetV7L5IL8VH_9efG9VJunJTgT0iKyWqjFbc,148
 phoenix/trace/langchain/instrumentor.py,sha256=HkNKbFNclTYjRXBM8qU4qvZHdyw06J9bhwgE7JnqbNI,1323
-phoenix/trace/langchain/tracer.py,sha256=W0wssIwqtEHjDQPgobz9Kd9wsr2kpuHl9lXuX1HQ9zI,16388
+phoenix/trace/langchain/tracer.py,sha256=1Oz3orSDpZX1pZKwtZbeM_f9tiAhQb7Of8ARjRlKVQY,16827
 phoenix/trace/llama_index/__init__.py,sha256=wCcQgD9CG5TA8i-1XsSed4ZzwHTUmqZwegQAV_FqEng,178
 phoenix/trace/llama_index/callback.py,sha256=YW3qqzWZUEs9aiDx-2628Eae_rct_Yb-DDzT9dV_xiI,27061
 phoenix/trace/llama_index/debug_callback.py,sha256=SKToD9q_QADSGTJ5lhilqRVKaUnUSRXUvURCzN4by2U,1367
 phoenix/trace/llama_index/streaming.py,sha256=5cTtr8evvcEAB88Xb4ih3WEw0xAF4x5W9PehUX9l5_0,3258
 phoenix/trace/openai/__init__.py,sha256=J3G0uqCxGdksUpaQVHds_Egv2drvh8UEqoLjiQAOveg,79
 phoenix/trace/openai/instrumentor.py,sha256=H1T2_1uqeH2lKCKeMmirEUl6PRtHQlQTXfsLR_hwDFM,24948
-phoenix/trace/v1/__init__.py,sha256=gzPW6vBuE_Z1bymL935cuiFBty1Tl9v2RCgAKY6vfPM,200
+phoenix/trace/v1/__init__.py,sha256=-IbAD0ruESMjvQLvGAg9CTfjBUATFDx1OXseDPis6-0,88
 phoenix/trace/v1/evaluation_pb2.py,sha256=8sXvv2BW_vqD30MOMbmkeE2zpmm7ncik21kl3e-HzeQ,2254
 phoenix/trace/v1/evaluation_pb2.pyi,sha256=cCbbx06gwQmaH14s3J1X25TtaARh-k1abbxQdQCXGm8,4500
-phoenix/trace/v1/trace_pb2.py,sha256=IfvVb4PkUS6W72uJvcZTj7yj8hFvaaRac0RiOYae7Ik,5724
-phoenix/trace/v1/trace_pb2.pyi,sha256=4OErYEvVemBUoCiD2ABG9NSpGDEEzJkHr6x9ALYvE5Y,16497
-phoenix/trace/v1/utils.py,sha256=j7gunL9CuSi7Xif56oWYGx0sc7KjfJhXzWcJia4ZM-8,17815
 phoenix/utilities/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/utilities/error_handling.py,sha256=7b5rpGFj9EWZ8yrZK1IHvxB89suWk3lggDayUQcvZds,1946
 phoenix/utilities/logging.py,sha256=lDXd6EGaamBNcQxL4vP1au9-i_SXe0OraUDiJOcszSw,222
-arize_phoenix-2.0.0.dist-info/METADATA,sha256=aPHvi6tNxDcZhbOnRkeFus1Zw5dvniTbkET_ythER2c,26310
-arize_phoenix-2.0.0.dist-info/WHEEL,sha256=mRYSEL3Ih6g5a_CVMIcwiF__0Ae4_gLYh01YFNwiq1k,87
-arize_phoenix-2.0.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
-arize_phoenix-2.0.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
-arize_phoenix-2.0.0.dist-info/RECORD,,
+arize_phoenix-2.2.0rc0.dist-info/METADATA,sha256=cV0tw8Sq1yms7yncqdyAkIoUOMVBIfDQkS_X9wlmgKY,26482
+arize_phoenix-2.2.0rc0.dist-info/WHEEL,sha256=mRYSEL3Ih6g5a_CVMIcwiF__0Ae4_gLYh01YFNwiq1k,87
+arize_phoenix-2.2.0rc0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
+arize_phoenix-2.2.0rc0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
+arize_phoenix-2.2.0rc0.dist-info/RECORD,,

phoenix/__init__.py

@@ -5,8 +5,7 @@ from .session.evaluation import log_evaluations
 from .session.session import NotebookEnvironment, Session, active_session, close_app, launch_app
 from .trace.fixtures import load_example_traces
 from .trace.trace_dataset import TraceDataset
-
-__version__ = "2.0.0"
+from .version import __version__
 
 # module level doc-string
 __doc__ = """
@@ -25,6 +24,7 @@ Here are just a few of the things that phoenix does well:
 """
 
 __all__ = [
+    "__version__",
     "Dataset",
     "EmbeddingColumnNames",
     "RetrievalEmbeddingColumnNames",

phoenix/core/evals.py

@@ -1,6 +1,7 @@
 import logging
 import weakref
 from collections import defaultdict
+from datetime import datetime, timezone
 from queue import SimpleQueue
 from threading import RLock, Thread
 from types import MethodType
@@ -46,6 +47,7 @@ class Evals:
         self._document_evaluations_by_name: DefaultDict[
             EvaluationName, DefaultDict[SpanID, Dict[DocumentPosition, pb.Evaluation]]
         ] = defaultdict(lambda: defaultdict(dict))
+        self._last_updated_at: Optional[datetime] = None
         self._start_consumer()
 
     def put(self, evaluation: pb.Evaluation) -> None:
@@ -92,10 +94,16 @@ class Evals:
             )
         else:
             assert_never(subject_id_kind)
+        self._last_updated_at = datetime.now(timezone.utc)
+
+    @property
+    def last_updated_at(self) -> Optional[datetime]:
+        return self._last_updated_at
 
     def get_span_evaluation(self, span_id: SpanID, name: str) -> Optional[pb.Evaluation]:
         with self._lock:
-            return self._evaluations_by_span_id[span_id].get(name)
+            span_evaluations = self._evaluations_by_span_id.get(span_id)
+            return span_evaluations.get(name) if span_evaluations else None
 
     def get_span_evaluation_names(self) -> List[EvaluationName]:
         with self._lock:
@@ -108,28 +116,36 @@ class Evals:
         with self._lock:
             if span_id is None:
                 return list(self._document_evaluations_by_name)
-            return list(self._document_evaluations_by_span_id[span_id])
+            document_evaluations = self._document_evaluations_by_span_id.get(span_id)
+            return list(document_evaluations) if document_evaluations else []
 
     def get_span_evaluation_labels(self, name: EvaluationName) -> Tuple[str, ...]:
         with self._lock:
-            return tuple(self._span_evaluation_labels[name])
+            labels = self._span_evaluation_labels.get(name)
+            return tuple(labels) if labels else ()
 
     def get_span_evaluation_span_ids(self, name: EvaluationName) -> Tuple[SpanID, ...]:
         with self._lock:
-            return tuple(self._span_evaluations_by_name[name].keys())
+            span_evaluations = self._span_evaluations_by_name.get(name)
+            return tuple(span_evaluations.keys()) if span_evaluations else ()
 
     def get_evaluations_by_span_id(self, span_id: SpanID) -> List[pb.Evaluation]:
         with self._lock:
-            return list(self._evaluations_by_span_id[span_id].values())
+            evaluations = self._evaluations_by_span_id.get(span_id)
+            return list(evaluations.values()) if evaluations else []
 
     def get_document_evaluation_span_ids(self, name: EvaluationName) -> Tuple[SpanID, ...]:
         with self._lock:
-            return tuple(self._document_evaluations_by_name[name].keys())
+            document_evaluations = self._document_evaluations_by_name.get(name)
+            return tuple(document_evaluations.keys()) if document_evaluations else ()
 
     def get_document_evaluations_by_span_id(self, span_id: SpanID) -> List[pb.Evaluation]:
         all_evaluations: List[pb.Evaluation] = []
         with self._lock:
-            for evaluations in self._document_evaluations_by_span_id[span_id].values():
+            document_evaluations = self._document_evaluations_by_span_id.get(span_id)
+            if not document_evaluations:
+                return all_evaluations
+            for evaluations in document_evaluations.values():
                 all_evaluations.extend(evaluations.values())
         return all_evaluations
 
@@ -144,7 +160,12 @@ class Evals:
         # of one trillion, we would not want to create a result that large.
         scores: List[float] = [np.nan] * num_documents
         with self._lock:
-            evaluations = self._document_evaluations_by_span_id[span_id][evaluation_name]
+            document_evaluations = self._document_evaluations_by_span_id.get(span_id)
+            if not document_evaluations:
+                return scores
+            evaluations = document_evaluations.get(evaluation_name)
+            if not evaluations:
+                return scores
             for document_position, evaluation in evaluations.items():
                 result = evaluation.result
                 if result.HasField("score") and document_position < num_documents:

phoenix/core/traces.py

@@ -13,20 +13,21 @@ from typing import (
     Iterator,
     List,
     Optional,
+    Set,
     SupportsFloat,
     Tuple,
-    Union,
     cast,
 )
 
+import opentelemetry.proto.trace.v1.trace_pb2 as otlp
 from ddsketch import DDSketch
 from sortedcontainers import SortedKeyList
 from typing_extensions import TypeAlias
 from wrapt import ObjectProxy
 
-import phoenix.trace.v1 as pb
 from phoenix.datetime_utils import right_open_time_range
 from phoenix.trace import semantic_conventions
+from phoenix.trace.otel import decode
 from phoenix.trace.schemas import (
     ATTRIBUTE_PREFIX,
     COMPUTED_PREFIX,
@@ -34,9 +35,10 @@ from phoenix.trace.schemas import (
     Span,
     SpanAttributes,
     SpanID,
+    SpanStatusCode,
     TraceID,
 )
-from phoenix.trace.v1.utils import decode, encode
+from phoenix.trace.semantic_conventions import RETRIEVAL_DOCUMENTS
 
 END_OF_QUEUE = None  # sentinel value for queue termination
 
@@ -74,15 +76,15 @@ class ReadableSpan(ObjectProxy):  # type: ignore
     are ingested, and would need to be re-computed on the fly.
     """
 
-    __wrapped__: pb.Span
-
-    def __init__(self, span: pb.Span) -> None:
+    def __init__(self, otlp_span: otlp.Span) -> None:
+        span = decode(otlp_span)
         super().__init__(span)
+        self._self_otlp_span = otlp_span
         self._self_computed_values: Dict[str, SupportsFloat] = {}
 
     @property
     def span(self) -> Span:
-        span = decode(self.__wrapped__)
+        span = decode(self._self_otlp_span)
         span.attributes.update(cast(SpanAttributes, self._self_computed_values))
         # TODO: compute latency rank percent (which can change depending on how
         # many spans already ingested).
@@ -96,9 +98,7 @@ class ReadableSpan(ObjectProxy):  # type: ignore
             return getattr(self.__wrapped__.context, suffix_key, None)
         if key.startswith(ATTRIBUTE_PREFIX):
             suffix_key = key[len(ATTRIBUTE_PREFIX) :]
-            if suffix_key not in self.__wrapped__.attributes:
-                return None
-            return self.__wrapped__.attributes[suffix_key]
+            return self.__wrapped__.attributes.get(suffix_key)
         return getattr(self.__wrapped__, key, None)
 
     def __setitem__(self, key: str, value: Any) -> None:
@@ -113,21 +113,21 @@ ChildSpanID: TypeAlias = SpanID
 
 class Traces:
     def __init__(self) -> None:
-        self._queue: "SimpleQueue[Optional[pb.Span]]" = SimpleQueue()
+        self._queue: "SimpleQueue[Optional[otlp.Span]]" = SimpleQueue()
         # Putting `None` as the sentinel value for queue termination.
         weakref.finalize(self, self._queue.put, END_OF_QUEUE)
         self._lock = RLock()
         self._spans: Dict[SpanID, ReadableSpan] = {}
         self._parent_span_ids: Dict[SpanID, ParentSpanID] = {}
-        self._traces: Dict[TraceID, List[SpanID]] = defaultdict(list)
-        self._child_span_ids: DefaultDict[SpanID, List[ChildSpanID]] = defaultdict(list)
-        self._orphan_spans: DefaultDict[ParentSpanID, List[pb.Span]] = defaultdict(list)
+        self._traces: DefaultDict[TraceID, List[SpanID]] = defaultdict(list)
+        self._child_span_ids: DefaultDict[SpanID, Set[ChildSpanID]] = defaultdict(set)
+        self._orphan_spans: DefaultDict[ParentSpanID, List[otlp.Span]] = defaultdict(list)
         self._num_documents: DefaultDict[SpanID, int] = defaultdict(int)
         self._start_time_sorted_span_ids: SortedKeyList[SpanID] = SortedKeyList(
-            key=lambda span_id: self._spans[span_id].start_time.ToDatetime(timezone.utc),
+            key=lambda span_id: self._spans[span_id].start_time,
         )
         self._start_time_sorted_root_span_ids: SortedKeyList[SpanID] = SortedKeyList(
-            key=lambda span_id: self._spans[span_id].start_time.ToDatetime(timezone.utc),
+            key=lambda span_id: self._spans[span_id].start_time,
         )
         self._latency_sorted_root_span_ids: SortedKeyList[SpanID] = SortedKeyList(
             key=lambda span_id: self._spans[span_id][ComputedAttributes.LATENCY_MS.value],
@@ -136,15 +136,18 @@ class Traces:
         self._min_start_time: Optional[datetime] = None
         self._max_start_time: Optional[datetime] = None
         self._token_count_total: int = 0
+        self._last_updated_at: Optional[datetime] = None
         self._start_consumer()
 
-    def put(self, span: Optional[Union[Span, pb.Span]] = None) -> None:
-        self._queue.put(encode(span) if isinstance(span, Span) else span)
+    def put(self, span: Optional[otlp.Span] = None) -> None:
+        self._queue.put(span)
 
     def get_trace(self, trace_id: TraceID) -> Iterator[Span]:
         with self._lock:
             # make a copy because source data can mutate during iteration
-            span_ids = tuple(self._traces[trace_id])
+            if not (trace := self._traces.get(trace_id)):
+                return
+            span_ids = tuple(trace)
         for span_id in span_ids:
             if span := self[span_id]:
                 yield span
@@ -194,7 +197,7 @@ class Traces:
 
     def get_num_documents(self, span_id: SpanID) -> int:
         with self._lock:
-            return self._num_documents[span_id]
+            return self._num_documents.get(span_id) or 0
 
     def latency_rank_percent(self, latency_ms: float) -> Optional[float]:
         """
@@ -221,11 +224,17 @@ class Traces:
     def get_descendant_span_ids(self, span_id: SpanID) -> Iterator[SpanID]:
         with self._lock:
             # make a copy because source data can mutate during iteration
-            span_ids = tuple(self._child_span_ids[span_id])
+            if not (child_span_ids := self._child_span_ids.get(span_id)):
+                return
+            span_ids = tuple(child_span_ids)
         for child_span_id in span_ids:
             yield child_span_id
             yield from self.get_descendant_span_ids(child_span_id)
 
+    @property
+    def last_updated_at(self) -> Optional[datetime]:
+        return self._last_updated_at
+
     @property
     def span_count(self) -> int:
         """Total number of spans (excluding orphan spans if any)"""
@@ -259,24 +268,24 @@ class Traces:
             with self._lock:
                 self._process_span(item)
 
-    def _process_span(self, span: pb.Span) -> None:
-        span_id = SpanID(span.context.span_id)
+    def _process_span(self, span: otlp.Span) -> None:
+        new_span = ReadableSpan(span)
+        span_id = new_span.context.span_id
         existing_span = self._spans.get(span_id)
-        if existing_span and existing_span.HasField("end_time"):
+        if existing_span and existing_span.end_time:
             # Reject updates if span has ended.
             return
-        is_root_span = not span.HasField("parent_span_id")
+        is_root_span = not new_span.parent_id
         if not is_root_span:
-            parent_span_id = SpanID(span.parent_span_id.value)
+            parent_span_id = new_span.parent_id
             if parent_span_id not in self._spans:
                 # Span can't be processed before its parent.
                 self._orphan_spans[parent_span_id].append(span)
                 return
-            self._child_span_ids[parent_span_id].append(span_id)
+            self._child_span_ids[parent_span_id].add(span_id)
             self._parent_span_ids[span_id] = parent_span_id
-        new_span = ReadableSpan(span)
-        start_time = span.start_time.ToDatetime(timezone.utc)
-        end_time = span.end_time.ToDatetime(timezone.utc) if span.HasField("end_time") else None
+        start_time = new_span.start_time
+        end_time = new_span.end_time
         if end_time:
             new_span[ComputedAttributes.LATENCY_MS.value] = latency = (
                 end_time - start_time
@@ -287,7 +296,7 @@ class Traces:
         if is_root_span and end_time:
             self._latency_sorted_root_span_ids.add(span_id)
         if not existing_span:
-            trace_id = TraceID(span.context.trace_id)
+            trace_id = new_span.context.trace_id
             self._traces[trace_id].append(span_id)
             if is_root_span:
                 self._start_time_sorted_root_span_ids.add(span_id)
@@ -303,7 +312,7 @@ class Traces:
                 else max(self._max_start_time, start_time)
             )
         new_span[ComputedAttributes.ERROR_COUNT.value] = int(
-            span.status.code is pb.Span.Status.Code.ERROR
+            new_span.status_code is SpanStatusCode.ERROR
         )
         # Update cumulative values for span's ancestors.
         for attribute_name, cumulative_attribute_name in (
@@ -336,14 +345,16 @@ class Traces:
             self._token_count_total -= existing_span[LLM_TOKEN_COUNT_TOTAL] or 0
         self._token_count_total += new_span[LLM_TOKEN_COUNT_TOTAL] or 0
         # Update number of documents
-        num_documents_update = len(span.retrieval.documents)
+        num_documents_update = len(new_span.attributes.get(RETRIEVAL_DOCUMENTS) or ())
         if existing_span:
-            num_documents_update -= len(existing_span.retrieval.documents)
+            num_documents_update -= len(existing_span.attributes.get(RETRIEVAL_DOCUMENTS) or ())
         if num_documents_update:
             self._num_documents[span_id] += num_documents_update
         # Process previously orphaned spans, if any.
         for orphan_span in self._orphan_spans.pop(span_id, ()):
             self._process_span(orphan_span)
+        # Update last updated timestamp
+        self._last_updated_at = datetime.now(timezone.utc)
 
     def _add_value_to_span_ancestors(
         self,

phoenix/experimental/evals/__init__.py

@@ -1,4 +1,4 @@
-from .evaluators import LLMEvaluator
+from .evaluators import InvalidEvalCriteriaError, LLMEvaluator
 from .functions import llm_classify, llm_generate, run_relevance_eval
 from .models import BedrockModel, LiteLLMModel, OpenAIModel, VertexAIModel
 from .retrievals import compute_precisions_at_k
@@ -16,11 +16,13 @@ from .templates import (
     TOXICITY_PROMPT_RAILS_MAP,
     TOXICITY_PROMPT_TEMPLATE,
     ClassificationTemplate,
+    EvalCriteria,
     PromptTemplate,
 )
 from .utils import NOT_PARSABLE, download_benchmark_dataset
 
 __all__ = [
+    "EvalCriteria",
     "compute_precisions_at_k",
     "download_benchmark_dataset",
     "llm_classify",
@@ -46,4 +48,5 @@ __all__ = [
     "QA_PROMPT_TEMPLATE",
     "NOT_PARSABLE",
     "run_relevance_eval",
+    "InvalidEvalCriteriaError",
 ]