libem-v0.0.25

What's Changed

• fix: import clustering dataset to avoid name collision by @daiwaid in #88
• feat: add benchmark results for all gpt models and minor bug fix by @daiwaid in #89
• chore: simplify pr template by @zenodflow in #91
• refactor: strict output for gpt-4o-08-06 and add avg results by @daiwaid in #92

Full Changelog: libem-v0.0.24...libem-v0.0.25

libem-v0.0.24

What's Changed

• refactor benchmark/suite and add batch, block suite by @daiwaid @zenodflow in #76
• feat: configurable deep resolution of callable parameter by @zenodflow in #77
• feat: support simple cluster-based dedupe by @zenodflow in #78
• feat: support record linkage in resolve with link by @zenodflow in #80
• feat: add duckdb and mongodb integration to cluster by @daiwaid in #81
• fix: preserve duplicate records in cluster by @daiwaid in #82
• refactor: simplify cluster id assignment by @daiwaid in #85
• feat: add support for llama 3.1 by @Char15Xu in #71
• feat: allow dynamic value for index by @zenodflow in #87

Full Changelog: libem-v0.0.23...libem-v0.0.24

libem-v0.0.23

What's Changed

• refactor: improve arena and adjust block threshold by @daiwaid in #63
• feat: add cluster tool with pandas integration by @zenodflow in #64
• feat: materialize parameter value at calibrate by @zenodflow in #65
• feat: init resolve and allow cluster to sort output by @zenodflow in #66
• perf: lazy load heavy modules to cut cli init time by @zenodflow in #72
• fix: create a deep copy of args in benchmark by @daiwaid in #73

Full Changelog: libem-v0.0.22...libem-v0.0.23

libem-v0.0.22

What's Changed

• move dataset formatting files to libem-sample-data by @daiwaid in #54
• feat: support indexing for parameter options by @zenodflow in #56
• feat: allow match to report likelihood as output by @zenodflow in #57
• feat: redesigned arena api to operate independently of front-end by @daiwaid in #58
• fix arena frontend bug and add unit test file by @daiwaid in #59
• feat(match): support structured input parsing for dict and pair by @zenodflow in #60
• feat: improve block input parsing and progress report by @zenodflow in #61
• refactor: simplify type hints by @zenodflow in #62

Full Changelog: libem-v0.0.21...libem-v0.0.22

libem-v0.0.21

What's Changed

• Add llama3 support on llama.cpp for non-apple silicon devices and benchmark results by @Char15Xu in #50
• feat: in-context learning with shot selection by @zenodflow in #51
• feat: always pass train set as part of benchmark but hold off materializing train set until first use by @daiwaid in #53

Full Changelog: libem-v0.0.20...libem-v0.0.21

libem-v0.0.20

What's Changed

• feat: add struct module and report confidence in 0-100 range by @zenodflow in #45
• feat: confidence calibration with temperature scaling by @zenodflow in #46
• feat: add support for gpt-4o-mini and implement benchmarks by @zenodflow in #47

Full Changelog: libem-v0.0.19...libem-v0.0.20

libem-v0.0.19

What's Changed

• feat: async model call and match by @daiwaid in #37
• feat: add llama3 on mlx to supported model by @Char15Xu and @zenodflow in #40
• refactor: make model a core module by @zenodflow in #41
• feat: add exec core module by @zenodflow in #42
• feat: separate tool and toolchain level interfaces by @zenodflow in #43
• feat: enable async tool use by @zenodflow in #44

Full Changelog: libem-v0.0.18...libem-v0.0.19

libem-v0.0.18

What's Changed

• Local model support by @Char15Xu in #30
• feat: add stats in model call return by @daiwaid in #33
• refactor: benchmark parses match trace in batch by @zenodflow in #35
• feat: prompt-level batching for match by @daiwaid in #34
• feat: add libem serve container build by @zenodflow in #36

Full Changelog: libem-v0.0.17...libem-v0.0.18

libem-v0.0.17

New Contributors

• @Char15Xu made their first contribution in #29

Full Changelog: libem-v0.0.16...libem-v0.0.17

libem-v0.0.16

Full Changelog: libem-v0.0.15...libem-v0.0.16