Releases: stanfordnlp/dspy
2.5.3
What's Changed
- fix(dspy): bug fix for snowflake retriever caused by smart filter update by @sfc-gh-alherrera in #1565
- [1] DatabricksRM: Use databricks-sdk to fetch token / workspace URL + several small improvements by @dbczumar in #1564
Full Changelog: 2.5.2...2.5.3
2.5.2
2.5.1
What's Changed
- Make output processing in typed predictors compatible to dspy.LM (2.5) by @krypticmouse in #1546
- Patche python interpreter to support additional syntax by @arnavsinghvi11 in #1550
- Release workflow updated to also publish to
dspy
as well asdspy-ai
on pypi by @hmoazam in #1557
New Contributors
- @bluewin4 made their first contribution in #1542
- @jannikmaierhoefer made their first contribution in #1554
Full Changelog: 2.5.0...2.5.1
2.5.0
What's Changed
- Adapters type casting & enforcement. Migration notebook. by @okhat in #1532
- Add deprecation warning for old clients on first use (per client) by @okhat in #1533
- Pin cached notebooks to 2.4.17 prior to deprecations by @okhat in #1528
- Top-level file cleanup by @okhat in #1529
- Add warning for deep copy by @chenmoneygithub in #1531
Full Changelog: 2.4.17...2.5.0
2.4.17
What's Changed
- Various improvements for adapters & LMs by @okhat in #1523
- MIPROv2(v2) -- Big Updates to the MIPROv2 Optimizer by @XenonMolecule in #1522
- Improve error logging with a stacktrace by @severi in #1458
- Fix dspy Module's deepcopy by @chenmoneygithub in #1519
- Typo/nit fixes and add tests for BoostrapWithRandomSearch by @chenmoneygithub in #1502
New Contributors
- @siddharthsambharia-portkey made their first contribution in #1496
- @NirantK made their first contribution in #1498
- @PrashantDixit0 made their first contribution in #1444
- @jeffreyrubi made their first contribution in #1484
- @manubhardwaj made their first contribution in #1437
- @arya-krishna-23 made their first contribution in #1432
- @severi made their first contribution in #1458
Full Changelog: 2.4.16...2.4.17
2.4.16
This is a pretty significant release. New dspy.LM
and dspy.Adapter
will be key for DSPy 2.5 onwards. Old clients are unaffected for now.
What's Changed
- Adding new dspy.LM and dspy.Adapter by @okhat in #1486
- Make langchain copy-able in DSPy by @chenmoneygithub in #1488
- feat(dspy): support o1 parameters by @isaacbmiller in #1487
New Contributors
Full Changelog: 2.4.15...2.4.16
2.4.15
What's Changed
- Assertions patches by @arnavsinghvi11 in #1471
- Fix misc assertion errors by @Shangyint in #1085
- Fix assertion bug with new_signature by @Shangyint in #1372
- Fix rendering of evaluation results Pandas DataFrames in console environments by @dbczumar in #1472
- Bug Fix: Multiline init regex parsing (MIPROv2) by @XenonMolecule in #1466
New Contributors
- @karthikscale3 made their first contribution in #1456
- @hacker-4-good made their first contribution in #1469
- @dbczumar made their first contribution in #1472
Full Changelog: 2.4.14...2.4.15
DSPy v2.4.14
Update pyproject.toml
DSPy v2.4.13
- supports configurable backoff time for LM/RM providers
dspy.settings.configure(backoff_time = ...)
- latest PR changes and updates (Langchain Metaclass resolution, addition of Langchain Tool Execution, MIPROv2 parameter updates, LM/RM Provider updates)
Full Changelog: v2.4.12...2.4.13
DSPy v2.4.12
-
Supports treating
dspy.Predict
anddspy.ChainOfThought
directly as Modules, e.g. can compile them without creating a wrappingdspy.Module
object. -
Improves the
experimental=True
Chat LM support from DSPy v2.4.11, which is quoted below for reference:
Initial support for new adapters, with far more smooth support for Chat LMs. The default behavior is unchanged from before. However, dspy.configure(experimental=True) is likely to lead to much more predictable and accurate zero-shot generations with Chat LMs. Tested against several variants of GPT-3.5, GPT-4, Llama3, Mixtral, and DBRX.