Releases: langchain-ai/langsmith-sdk
Releases · langchain-ai/langsmith-sdk
v0.1.144
What's Changed
- Lazy internal import by @hinthornw in #1198
- python[patch]: evaluators can return primitives by @baskaryan in #1203
- perf: cut down cpu time of aevaluate by 30% on 1-4MB examples with this one trick by @jakerachleff in #1217
- nit: stop calling importlib on every api req by @jakerachleff in #1219
- python[patch]: evaluate accept Dataset by @baskaryan in #1222
- python[patch]: accept simple evaluators by @baskaryan in #1200
- NotMyPy by @hinthornw in #1221
- orjson optional wip by @efriis in #1223
- release 0.1.144rc1 by @efriis in #1226
- Revert "orjson optional wip (#1223)" by @baskaryan in #1228
- orjson optional take 2 by @efriis in #1230
- skip flaky js integration test by @efriis in #1232
- infra: use benchmark-fast in pybench ci by @efriis in #1160
- timeout_ms docstring by @efriis in #1234
- fix(js): Fix multipart upload with overridden node fetch by @jacoblee93 in #1235
- python[patch]: accept simple summary eval signature by @baskaryan in #1227
- Bump cross-spawn from 7.0.3 to 7.0.6 in /js by @dependabot in #1233
- sdk patch: raise errors on update/create examples when passing mismatching length sequences by @isahers1 in #1238
- Support comments in evaluate_comparative by @hinthornw in #1211
- [python] Ensure timezone by @hinthornw in #1239
Full Changelog: v0.1.143...v0.1.144
v0.1.144rc1
What's Changed
- Lazy internal import by @hinthornw in #1198
- python[patch]: evaluators can return primitives by @baskaryan in #1203
- perf: cut down cpu time of aevaluate by 30% on 1-4MB examples with this one trick by @jakerachleff in #1217
- nit: stop calling importlib on every api req by @jakerachleff in #1219
- python[patch]: evaluate accept Dataset by @baskaryan in #1222
- python[patch]: accept simple evaluators by @baskaryan in #1200
- NotMyPy by @hinthornw in #1221
- orjson optional wip by @efriis in #1223
- release 0.1.144rc1 by @efriis in #1226
Full Changelog: v0.1.143...v0.1.144rc1
v0.1.143
What's Changed
- [Python] push_prompt update by @hinthornw in #1202
- python[patch]: pass Runnable to evaluate by @baskaryan in #1204
Full Changelog: v0.1.142...v0.1.143
v0.1.142
What's Changed
- fix(js): Use batch endpoint instead of individual run endpoint in case of server info fetch failure by @jacoblee93 in #1189
- Fix urllib patch path by @hinthornw in #1197
Full Changelog: v0.1.141...v0.1.142
v0.1.141
What's Changed
- js feat: Support attachments in traceable by @hinthornw in #1169
- Improve error msg for empty data by @hinthornw in #1178
- chore: update fastapi dev dependency to fix vulnerability report by @agola11 in #1182
- [Python]: Fix concurrency issues in caching for urllib3<2.0 by @hinthornw in #1183
- Remove initial timeout, the batch timeout is now always 250ms by @nhuang-lc in #1184
- fix(js): Export more evaluator types by @bracesproul in #1190
- fix(js): Release 0.2.5 by @bracesproul in #1191
- [Python] Update prompt schema by @hinthornw in #1193
- Python 0.1.141 by @hinthornw in #1194
Full Changelog: v0.1.140...v0.1.141
v0.1.140
What's Changed
- track evaluator errors from sdk by @isahers1 in #1079
- feat(vercel): add debug flag by @dqbd in #1166
- fix(vercel): add more debug logs, fix tracing detection for vercel, bump to 0.2.4 by @dqbd in #1168
- Fix clone dataset from other server by @hinthornw in #1175
- Cleaner error message by @hinthornw in #1177
- [Python] to_pandas() method for ExperimentResults by @baskaryan in #1170
- Pre-read smallish multipart requests by @hinthornw in #1176
New Contributors
Full Changelog: v0.1.139...v0.1.140
v0.1.139rc2
v0.1.139
What's Changed
- Skip model_dump for classes by @hinthornw in #1149
- Minimize Repr by @hinthornw in #1155
- serialize instead of deepcopy for performance gains by @efriis in #1136
- Annotation Support by @hinthornw in #1146
- release 0.1.139 by @efriis in #1161
Full Changelog: v0.1.138...v0.1.139
v0.1.139rc1
What's Changed
- Skip model_dump for classes by @hinthornw in #1149
- Minimize Repr by @hinthornw in #1155
- serialize instead of deepcopy for performance gains by @efriis in #1136
Full Changelog: v0.1.138...v0.1.139rc1
v0.1.138
What's Changed
- feat(js): Use new multipart endpoint when available by @jacoblee93 in #1106
- chore(js): Release 0.1.67 by @jacoblee93 in #1119
- fix(js): Limit queue batch concurrency, bump timeout, add maximum wait period for serverless envs by @jacoblee93 in #1124
- feat(vercel): add OTEL based LangSmith trace exporter by @dqbd in #1104
- factor out background and serde by @efriis in #1130
- integration tests multipart by @efriis in #1118
- properly setup env for tests by @efriis in #1131
- fix(js): Further reduce blocking behavior, try naive stringification instead of circular detection by @jacoblee93 in #1133
- integration tests against staging by @efriis in #1143
- fix(js): Remove focused test, fix tests by @jacoblee93 in #1138
- fix: add required basic auth environment variable by @bvs-langchain in #1128
- feat(py): Add multipart feedback ingestion by @akira in #1129
- use staging for integration tests - js by @efriis in #1145
- use staging for integration tests by @efriis in #1140
- fix(vercel): pass run name if the root run is within a span by @dqbd in #1150
- Add version check for feedback multipart by @akira in #1151
- [python] Bump to 0.1.138 by @akira in #1156
New Contributors
Full Changelog: v0.1.137...v0.1.138