v0.3.1

What's Changed

• Remove uv.lock by @eb8680 in #488
• Update encodable type class to have optional prompt prefix (closes #498) by @kiranandcode in #501
• Adds context parameter to encode and decode calls (closes #499) by @kiranandcode in #500
• Implements Polymorphic Templates by @kiranandcode in #504
• Revert "Implements Polymorphic Templates (#504)" by @kiranandcode in #507
• Drop cacheLLMhandler in favour of LiteLLM's built in caching mechanisms. by @kiranandcode in #510
• Remove LoggingLLMHandler by @kiranandcode in #508
• Support for literal type annotations by @kiranandcode in #512
• Switch type_to_encodable_type interface to lexically-sensitive Encodable design by @kiranandcode in #521
• Clean up internal LLM API by @eb8680 in #484
• Implement internal synthesis API by @kiranandcode in #519
• EvalProvider using RestrictedPython [Depends on #519] by @kiranandcode in #520
• Implement RetryHandler on top of new internal API (closes #495) by @kiranandcode in #522
• Update internal interface to enforce message ids by @kiranandcode in #527
• Abstract message sequence to handlers that can be overriden. by @kiranandcode in #528
• Bump version from 0.3.0 to 0.3.1 by @eb8680 in #530

Full Changelog: v0.3.0...v0.3.1

v0.3.0

What's Changed

• Install prettyprinter for term when library is available by @jfeser in #386
• Fix pypandoc issue in CI by @jfeser in #397
• Convert evaluate to a singledispatch by @jfeser in #398
• add jnp.pi and ArrayTerm.T by @cooijmanstim in #394
• Deprecate defterm by @jfeser in #399
• Rework numpyro distribution handling to enable symbolic distributions and handling of distribution methods by @jfeser in #311
• Box the output of __type_rule__ by @jfeser in #387
• Fix syntactic_eq implementation for jax arrays by @jfeser in #405
• Fix recursion error in sizesof by @jfeser in #406
• Allow _BaseOperation subclasses to have an overrideable apply method by @jfeser in #414
• Try pulling in pyproject.toml from staging-llm to master by @eb8680 in #425
• Generate instance-level Operations for bound methods by @eb8680 in #351
• Fix dataclass @defops and added dataclass metaclass by @kiranandcode in #439
• Type check and lint example code by @jfeser in #449
• Add beam search example using thermometer continuations by @jfeser in #431
• Fix for jax 0.8.2 by @jfeser in #455
• Enabled term construction for all numerical comparison operators by @kiranandcode in #465
• Deprecate raising notimplementederror by @jfeser in #472
• Remove defterm by @jfeser in #473
• Replace remaining uses of dm-tree with pytrees by @eb8680 in #480
• Move LLM interface code from robotl by @jfeser in #358
• Implement basic tool calling by @jfeser in #366
• Fix product_n when used on terms containing dataclasses by @jfeser in #373
• Enable strict tool calling by @jfeser in #375
• Enable structured generation by @jfeser in #376
• Add support for methods in Template.define (#377) by @kiranandcode in #378
• Adding a lower-level event and a logger example by @datvo06 in #382
• Add support for tools returning images by @kiranandcode in #385
• Implement Caching Handler for LLM by @datvo06 in #392
• Implement FirstToKAhead Sampler by @kiranandcode in #412
• Add inheritable class for stateful templates by @jfeser in #416
• Support multiple providers (via litellm) by @kiranandcode in #418
• store source of generated functions in __src__ attribute by @kiranandcode in #403
• Adds type-based encoding and support for legacy APIs by @kiranandcode in #411
• Add LLM Integration tests to the workflows. by @kiranandcode in #420
• Merge master into llm-staging by @jfeser in #423
• Fix staging-llm diff against master by @eb8680 in #426
• Implement a RetryHandler for LLM module by @datvo06 in #428
• Merge master into staging-llm again by @eb8680 in #443
• Implements a unified encodeing/decodeing pipeline for llm by @kiranandcode in #442
• Initial version of Lexical Context Collection - Collecting Tools and Template by @datvo06 in #434
• Update staging-llm from master by @eb8680 in #457
• Convert Template into an operation by @jfeser in #424
• Fail when encoding terms or operations by @jfeser in #474
• Implemented record and replay fixtures for LLM calls by @kiranandcode in #467
• Remove program synthesis code by @jfeser in #475
• Disables direct recursion on templates by default by @kiranandcode in #466
• Remove k-ahead sampler by @jfeser in #479
• Document Template and Tool by @jfeser in #478
• Remove agent.py from handler and add docs on how to implement Agents by @kiranandcode in #464
• Fix RetryHandler not to use Template.replace() by @datvo06 in #483
• Rename handlers.llm.providers to completions by @eb8680 in #485
• Land staging-llm on master by @eb8680 in #486
• Bump version from 0.2.3 to 0.3.0 by @eb8680 in #487

New Contributors

• @datvo06 made their first contribution in #382
• @cooijmanstim made their first contribution in #394

Full Changelog: v0.2.3...v0.3.0

v0.2.3

What's Changed

• Fix product_n when used on terms containing dataclasses by @jfeser in #373
• Release v0.2.3 by @jfeser in #374

Full Changelog: v0.2.2...v0.2.4

v0.2.2

What's Changed

• Restrict 3.14 in CI by @jfeser in #369
• Enable test_notebooks on staging by @jfeser in #368
• Use productN to implement defdata by @jfeser in #281
• Release 0.2.2 by @jfeser in #372

Full Changelog: v0.2.1...v0.2.2

v0.2.1

What's Changed

• Bugfix in class canonicalization by @jjcmoon in #352
• Add namedtuple case in evaluate by @eb8680 in #355
• Bump version from 0.2.0 to 0.2.1 by @eb8680 in #357

Full Changelog: v0.2.0...v0.2.1

v0.2.0

Breaking Changes

• In operation definitions, effectful.ops.types.NotHandled should be raised instead of NotImplementedError (#312). Raising NotImplementedError will now produce a warning.

What's Changed

• Fix missing typing of to_tensor by @jfeser in #212
• Add a __str__ implementation for Term and simplify __repr__ by @jfeser in #218
• Enable doctests in ci by @jfeser in #217
• Implement sizesof for named and positional distributions by @jfeser in #214
• Fix bugs in PyroShim by @jfeser in #223
• Drop isort, black, flake8 in favor of ruff by @jfeser in #219
• Remove unused kwargs by @jfeser in #229
• Fix performance bug in vmap call by @jfeser in #230
• Disable code coverage by @jfeser in #231
• Clean up lint and test actions by @jfeser in #232
• Enable notebook lint in CI by @jfeser in #235
• Enable index tensor as index argument by @jfeser in #237
• Support arithmetic operations on TensorTerms by @jfeser in #245
• Switch to tensor-valued index variables by @jfeser in #248
• Support ordering Operations by @jfeser in #250
• Add indexed distribution tests by @jfeser in #255
• Add features needed by counterfactual by @jfeser in #241
• Fixes for distribution embedding by @jfeser in #254
• Make to_tensor an operation by @jfeser in #247
• Fix typeof bug when returning None by @jfeser in #256
• Register _DistributionTerm with defdata by @jfeser in #258
• Remove unused test parameters by @jfeser in #261
• Stop marking passing tests as xfail by @jfeser in #264
• Introduce bind_dims/unbind_dims by @jfeser in #257
• Switch to pyproject.toml and uv by @jfeser in #260
• Add jax implementation of named dimensions & partial evaluation by @jfeser in #259
• Move callable term conversion to a new function by @jfeser in #278
• Implement __iter__ and __len__ for jax arrays by @jfeser in #268
• Pretty printing for terms containing collections by @jfeser in #265
• Disable flaky test by @jfeser in #286
• Extend scoping rule to handle collections of bound variables by @jfeser in #282
• Implement .at for jax array terms by @jfeser in #284
• Add special case handling for singledispatch functions in defop by @jfeser in #283
• Add support for using defop as a method decorator by @eb8680 in #290
• Allow isinstance checks for Interpretations by @jfeser in #291
• Add defdata patterns and operations for Iterables and Iterators by @eb8680 in #295
• Fix evaluation of operations that return reducible terms by @jfeser in #293
• Correctly rename bound variables that appear as dict keys by @jfeser in #297
• Add module for jax.scipy.special by @jfeser in #294
• Remove buggy check in jax tensor indexing by @jfeser in #302
• Bump minimum Python version to 3.12 by @eb8680 in #303
• Remove non-uniform behavior for numeric equality by @jfeser in #306
• Free array iter bugfix by @jjcmoon in #307
• Restrict jax version by @jfeser in #313
• Support terms that contain jax.array, jax.asarray by @jfeser in #317
• Add jax.numpy.linalg by @jfeser in #316
• Add type inference module for embedding polymorphic types by @eb8680 in #299
• Remove dependency of ops on dm-tree by @eb8680 in #323
• Add support for dataclass in evaluate by @eb8680 in #324
• Support jax 0.7 by bumping numpyro by @jfeser in #325
• Add parallel CI builds for Python 3.12 and 3.13 by @eb8680 in #326
• Introduce NotHandled replacement for NotImplementedError by @jfeser in #327
• Remove intp argument from apply by @eb8680 in #333
• Absorb call into CallableTerm by @eb8680 in #334
• Remove handlers.numbers by @eb8680 in #335
• Fix vestigial calls to apply.default_rule by @eb8680 in #336
• Use sphinx_autodoc_typehints to replace manual type annotations in sphinx docs by @jfeser in #338
• Make dm-tree dependency optional by @eb8680 in #339
• Move docs requirements into pyproject toml by @eb8680 in #341
• Fix lambda_example docstring by @eb8680 in #344
• Convert syntactic_eq to a singledispatch and add tensor support by @jfeser in #342
• Make defdata and __fvs_rule__ rely on inspect.BoundArguments by @eb8680 in #343
• Fix bug in fvsof by @eb8680 in #346
• Add jnp methods to _ArrayTerm by @jjcmoon in #340
• Bump project version from 0.1.0 to 0.2.0 by @eb8680 in #350

New Contributors

• @jjcmoon made their first contribution in #307

Full Changelog: v0.1.0...v0.2.0

Initial GitHub release

Initial release to test package deployment