v0.12.0

What's Changed

• Disallowing PLR6104 for graph bugs and latest Ruff formatter by @jamesbraza in #123
• Simplify ReAct agent by @albertbou92 in #111
• Promoted StoreTrajectoriesCallback into main ldp by @jamesbraza in #127
• Moving from large-runner to ubuntu-latest for CI by @jamesbraza in #132
• Pulling in latest aviary for core namespace by @jamesbraza in #133
• Lock file maintenance by @renovate in #131
• ReAct Agent working with Anthropic models by @albertbou92 in #128
• tdqm.gather over asyncio.gather in TreeSearchRollout.sample_trees by @jamesbraza in #134
• Small optimizer & rollout improvements by @sidnarayanan in #135

Full Changelog: v0.11.3...v0.12.0

v0.11.3

What's Changed

• Moved Renovate config validation into GitHub Actions by @jamesbraza in #114
• Documented and type hinted _TrainingModeContext by @jamesbraza in #115
• Offline learning tweak; adding gradients for tool_choice by @sidnarayanan in #113
• Properly pinning openai and litellm for LiteLLM bugs by @jamesbraza in #116
• Mark truncated earlier by @sidnarayanan in #118
• Fixing merge of stochastic graph nodes by @jamesbraza in #119
• Lock file maintenance by @renovate in #120
• Exposing hash function for TransitionTree.merge_identical_nodes's observations by @jamesbraza in #121
• Enabling OpResult->Op lookup and simplifying graph traversal by @sidnarayanan in #122

Full Changelog: v0.11.2...v0.11.3

v0.11.2

What's Changed

• Documenting units on AsyncBufferedWorker.max_wait_interval by @jamesbraza in #108
• Lock file maintenance by @renovate in #110
• Lock file maintenance by @renovate in #112
• Added main entry point by @whitead in #109

New Contributors

• @whitead made their first contribution in #109

Full Changelog: v0.11.1...v0.11.2

v0.11.1

What's Changed

• [Fix] Torch grads minor fixes by @albertbou92 in #97
• Pulling in hynek/build-and-inspect-python-package and latest dependencies by @jamesbraza in #98
• Populated ldp.graph.__init__ by @jamesbraza in #99
• Lock file maintenance by @renovate in #100
• Lock file maintenance by @renovate in #104
• Update dependency pre-commit to v4 by @renovate in #102
• Accounting for litellm.caching breaking change by @jamesbraza in #105
• Update all non-major dependencies by @renovate in #101
• Renovate once/month, removing stale LiteLLM disables, removing check-sdist by @jamesbraza in #106
• Updating LiteLLM after litellm.caching import change in litellm>=1.49.3 by @jamesbraza in #107

Full Changelog: v0.11.0...v0.11.1

v0.11.0

What's Changed

• Interactive agent improvements by @sidnarayanan in #95
• AsyncBufferedWorker by @sidnarayanan in #94
• Tracking tool usage as part of metrics by @jamesbraza in #96

Full Changelog: v0.10.0...v0.11.0

v0.10.0

What's Changed

• Interactive agent by @sidnarayanan in #84
• Refactor to expose a Memory factory in MemoryOpt by @jamesbraza in #85
• Refactor to expose a Memory.query factory in MemoryAgent by @jamesbraza in #86
• Silenced and resolving DeprecationWarnings by @jamesbraza in #87
• Handling div0 error in WandBLoggingCallback by @jamesbraza in #88
• Moving to uv publish and adding check-sdist by @jamesbraza in #91
• Dropping defaultdict in callback and over-sampling eval by @sidnarayanan in #90
• Refactor of MemoryOpt/Memory to enable lessons-based optimization by @jamesbraza in #89
• Fixing publish CI by @jamesbraza in #93

Full Changelog: v0.9.1...v0.10.0

v0.9.1

What's Changed

• Added missing TrajectoryFileCallback to ldp.alg import shortcuts by @jamesbraza in #82
• Resolved discounted_returns circular import by @jamesbraza in #83

Full Changelog: v0.9.0...v0.9.1

v0.9.0

What's Changed

• [Feature] Embedding Op by @albertbou92 in #75
• [Feature] MSP Op by @albertbou92 in #76
• [Fix] LLMCallOp can directly return the logprob when temperature if 1.0 or 0.0. by @albertbou92 in #77
• Callback hooks for Agent.init_state and Environment.reset by @jamesbraza in #78
• Adding LoggingCallback by @Ryan-Rhys in #79
• Created __init__.py shortcuts for ldp.alg by @jamesbraza in #81

Full Changelog: v0.8.1...v0.9.0

v0.8.1

What's Changed

• Accounting for LiteLLM LITELLM_LOG in configure_log_levels by @jamesbraza in #67
• More understandable training in OnlineTrainer by @jamesbraza in #68
• Ability to opt into rich progress bar by @jamesbraza in #69
• Fixed missing lower call in LDP_TQDM_USE_RICH by @jamesbraza in #70
• chore(deps): lock file maintenance by @renovate in #71
• chore(deps): lock file maintenance by @renovate in #72
• Dropping build in favor of uv build by @jamesbraza in #73
• Fixing None error in TrajectoryMetricsCallback by @jamesbraza in #74

Full Changelog: v0.8.0...v0.8.1

v0.8.0

What's Changed

• Testing mmkal/runovate for batched Renovate by @jamesbraza in #45
• Moving to runovate that doesn't require package.json by @jamesbraza in #47
• Fixing runovate incorrectly editing renovate config by @jamesbraza in #49
• Runovate CI deprecation and required config by @jamesbraza in #51
• More readable RolloutDebugDumpCallback by @jamesbraza in #46
• Running tests on non-main branches, better explaining pydot pinning by @jamesbraza in #52
• Only running pushon main or renovate/* branches by @jamesbraza in #54
• Allowing runovate to git push by @jamesbraza in #55
• Using actions over contents to allow git push by @jamesbraza in #56
• Giving contents: write permissions by @jamesbraza in #57
• renovate: update 🌲 by @github-actions in #53
• Removed unnecessary permissions from deps CI by @jamesbraza in #58
• Simplify TrajectoryMetricCallback by @sidnarayanan in #59
• Modernizing Renovate config by @jamesbraza in #60
• Resolved FBT003 in OnlineTrainerConfig by @jamesbraza in #63
• Validating OnlineTrainerConfig.update_every by @jamesbraza in #64
• Exposing MemoryOpt.memory_template for configuration by @jamesbraza in #66
• stdout_level arg name to level by @jamesbraza in #61
• Defaulting OptimizerConfig.optimizer_type to auto-select by @jamesbraza in #62
• Exposing OnlineTrainer.evaluate by @jamesbraza in #65

Full Changelog: v0.7.2...v0.8.0