Release 1.6.0

Release 1.6.0

🏅 Highlights

• Remote execution across agents & CLI: new skill and --remote flag for automated Ray dispatch (#633, #593).
• Faster metadata access: optimized entity retrieval with targeted SQL queries (#498).
• CLI UX: rich DataFrame output for show/get commands and improved detail views (#584, #586, #571, #578, #577).
• Ray Tune stability: validation fixes, safer parameter handling, NaN serialization, and deadlock prevention (#631, #656, #642, #657).

✨ Features

• Agents: add skill for remote execution; update ADO CLI skill (#633, #632).
• CLI: --remote flag for automated Ray dispatch; rich DataFrame output across handlers; operator descriptions with --details (#593, #584, #586, #571, #578, #577).
• Core: optionally raise exceptions when retrieving multiple resources (#626).
• Utilities: show_index parameter in dataframe_to_rich_table (#531).

🐛 Fixes

• Actuators: correct typing and @classmethod usage (#517).
• CLI: handle virtual properties in show entities; ensure run_experiment results fully printed (#630, #634).
• Core: initialize Ray with raw dictionary instead of RuntimeEnv (#546).
• Custom Experiments: compute measurement status deterministically (#658).
• ordered_pip: skip creating venv if it already exists (#573).
• Ray Tune: validate points_to_evaluate; avoid in-place mutation; serialize NaN; poll to prevent deadlocks (#631, #656, #642, #657).
• Rifferla: correct model default when registering operator (#651).

⚡ Performance

• Core: targeted SQL queries significantly reduce entity retrieval overhead (#498).

🧰 Build

• Bumped dependencies and pre-commit hooks to latest compatible versions (#659, #637, #609, #596, #592, #561, #545, #525, #653, #597, #539).
• Enforce typer >= 0.22.0 (#560).

📝 Docs

• Agents: improved planning guidance for discovery problems; added coding agent skills & rules (#635, #483).
• Contributing: exclude venv folders from checks; updated Prettier instructions (#526, #551).
• Testing: updated test instructions (#530).

🧹 Refactoring

• Core: decorate actuators with ray.remote dynamically (#544).
• CLI: validation message for create --dry-run; refresh ado get actuators (#538, #535).
• Metastore: cleaner field extraction with >>; remove legacy non–tz-aware date handling (#582, #581).
• Modules: simplify docstring extraction logic (#565).
• Examples: refresh CSVSampleStores with old keys (#611).

---

⚙️ Autoconf

• Retrain models due to dependency changes (#549).
• Clarify model versioning rationale (#604).
• Build: require pandas < 3.0.0 (#648).

🧠 SFTTrainer

• Validate dataset path exists and is a file (#574).

📈 vLLM Performance

• Add 50th percentile metric to vLLM bench experiments (#529).
• Make vLLM and GuideLLM optional dependencies (#540).

✂️ TRIM

• Introduce TRIM operator (#329).
• Propagate actuator configuration IDs to sub-operations (#583).
• Correct model default when registering operator (#649).
• Build: rename wheel to ado-trim; switch to uv for pyproject (#566, #542).
• Docs: clarify requirements on target observed property; add PyPI README; include example YAMLs (#602, #621, #552).
• Refactor: adopt Annotated pattern and default_factory in Pydantic models (#548).

---

Complete changes

Features

• (agents) add skill for remote execution (#633) - (730cafb) - Michael Johnston
• (agents) update ado cli skill (#632) - (763ca03) - Michael Johnston
• (cli) add --remote flag for automated remote Ray dispatch of ado CLI commands (#593) - (493ea3f) - Michael Johnston
• (cli) use rich dataframe print for default ado get handlers (#584) - (cd302b9) - Alessandro Pomponio
• (cli) use rich dataframe output in show commands (#586) - (bcaa8df) - Alessandro Pomponio
• (cli) show operator description when using --details (#577) - (0c36592) - Alessandro Pomponio
• (cli) use rich dataframe printing in ado get context (#571) - (c75c33f) - Alessandro Pomponio
• (cli) use rich dataframe printing for ado get operators (#578) - (ffc5000) - Alessandro Pomponio
• (core) optionally raise exceptions when retrieving multiple resources (#626) - (4643be5) - Alessandro Pomponio
• (core) optimize entity retrieval with targeted SQL queries (#498) - (a218a8f) - Michael Johnston
• (trim) introduce TRIM operator (#329) - (b98c115) - Daniele Lotito
• (utilities) add show_index parameter in dataframe_to_rich_table (#531) - (4e3eb9f) - Alessandro Pomponio

Bug Fixes

• (actuators) use correct typing and classmethod decorators (#517) - (3748122) - Michael Johnston
• (autoconf) retrain models due to dependency changes (#549) - (f9ea285) - Daniele Lotito
• (cli) correctly handle virtual properties in show entities (#630) - (66ade7e) - Michael Johnston
• (cli) ensure run_experiment result is fully printed (#634) - (4b63dbd) - Michael Johnston
• (core) use raw dictionary instead of RuntimeEnv to initialise ray (#546) - (a9a1470) - Michael Johnston
• (custom_experiments) compute measurement status instead of using if/else logic (#658) - (0518e89) - Michael Johnston
• (ordered_pip) check whether venv directory already exists (#573) - (5c41146) - Vassilis Vassiliadis
• (ray_tune) do not modify search algorithm parameters in-place (#656) - (43aa30e) - Alessandro Pomponio
• (ray_tune) prevent deadlock by polling for critical errors (#657) - (241cfff) - Michael Johnston
• (ray_tune) serialize nan as a string (#642) - (bab609a) - Alessandro Pomponio
• (ray_tune) validate points_to_evaluate option of samplers (#631) - (e03c4d3) - Michael Johnston
• (rifferla) return correct model default when registering operator (#651) - (10b8fda) - Alessandro Pomponio
• (sfttrainer) check whether dataset path exists and it's a file (#574) - (3c1615c) - Vassilis Vassiliadis
• (trim) return correct model default when registering operator (#649) - (c179f8b) - Alessandro Pomponio
• (trim) propagate actuator configuration ids to sub-operations (#583) - (3ed26da) - Daniele Lotito
• (vllm_performance) Added 50th percentile metric to vllm bench experiments (#529) - (3bc30b6) - Christian Pinto

Documentation

• (agents) improve planning for formulating a discovery problem (#635) - (6393abb) - Michael Johnston
• (agents) add coding agent skills and rules (#483) - (b4a4042) - Michael Johnston
• (autoconf) clarify model versioning rationale (#604) - (d772e19) - Daniele Lotito
• (changelog) update changelog (#519) - (7061f8f) - Alessandro Pomponio
• (contributing) exclude venv folders from checks (#526) - (e48a20c) - Alessandro Pomponio
• (test) update test instructions (#530) - (d91ce62) - Alessandro Pomponio
• (trim) clarify requirements on target observed property (#602) - (a3aa3e5) - Daniele Lotito
• (trim) add readme for PyPI (#621) - (2f9a49f) - Daniele Lotito
• (trim) fix the inclusion of TRIM example_yamls files (#552) - ([8982a1b](https...

Release 1.5.0

🚀 Highlights

• New Autoconf recommender for per_device_train_batch_size for smarter defaults.
• CLI gains support for ADO get experiments.
• Major core stability & performance: SQL store initialization/refresh improvements, engine reuse, and race-condition fix.

✨ Features

• autoconf: Introduce recommender for per_device_train_batch_size (#500, 1cead81)
• cli: Add support for ado get experiments (#497, 624536e)
• core: Improve SQL sample store initialization and enable efficient refresh (#481, 167ca37)

🐛 Fixes

• cli: Validate experiments against entity space in create space --dry-run (#513, 8a3c800)
• core: Avoid race condition by copying dictionary in DiscoverySpaceManager.monitorUpdates (#509, b2d1736)
• core: Set ray working_dir=None for local workers to support uv run (#506, fa4f99e)
• modules: Avoid displaying empty spinner panel (#494, 5fa1980)

⚡ Performance

• core: Reuse engine for SQL resource and sample stores (#492, 8e57d38)

🧰 Build

• ray_tune: Require minimum numba version (#504, 9b366d8)
• deps: Update dependencies (#512, e6cbb71; #490, 93ac58a)
• Add detect-secrets to dev dependencies (#484, ce9eb9d)

📝 Docs

• sfttrainer: Example for checking variability of fms-hf-tuning measurements (#473, 174fcb1)
• Add instructions on ensuring locked dependencies (#515, e85405c)

🧹 Refactoring

• test: Support locked and unlocked runners (#508, 8dd15e0)

---

⚙️ Autoconf

• Introduce recommender for per_device_train_batch_size (#500, 1cead81)
• Update to AutoGluon 1.5 (#470, f793fa6)

🧠 SFTTrainer

• Example for checking variability of fms-hf-tuning measurements (#473, 174fcb1)

📈 vLLM Performance

• Rename image_secret to image_pull_secret_name (#514, b9677d7)

---

Complete Changes

Features

• (autoconf) Introduce a new recommender for per_device_train_batch_size (#500) - (1cead81) - Srikumar Venugopal
• (cli) add support for ado get experiments (#497) - (624536e) - Alessandro Pomponio
• (core) improve sql sample store initialisation and allow for efficient refresh (#481) - (167ca37) - Alessandro Pomponio

Bug Fixes

• (cli) validate experiments against entity space in create space dry run (#513) - (8a3c800) - Michael Johnston
• (core) copy dictionary to avoid race condition in DiscoverySpaceManager monitorUpdates (#509) - (b2d1736) - Michael Johnston
• (core) set ray working_dir=None for local workers to support uv run (#506) - (fa4f99e) - Michael Johnston
• (modules) avoid displaying empty spinner panel (#494) - (5fa1980) - Alessandro Pomponio

Performance Improvements

• (core) reuse engine for sql resource and sample stores (#492) - (8e57d38) - Alessandro Pomponio

Documentation

• (changelog) update changelog (#518) - (db580b9) - Alessandro Pomponio
• (sfttrainer) example for checking the variability of fms-hf-tuning measurements (#473) - (174fcb1) - Vassilis Vassiliadis
• add instructions on ensuring locked dependencies (#515) - (e85405c) - Alessandro Pomponio

Build system

• (autoconf) update lockfile (#493) - (e210cfa) - Alessandro Pomponio
• (autoconf) update to autogluon 1.5 (#470) - (f793fa6) - Srikumar Venugopal
• (deps) update dependencies (#512) - (e6cbb71) - Alessandro Pomponio
• (deps) update dependencies (#490) - (93ac58a) - Alessandro Pomponio
• (ray_tune) require minimum numba version (#504) - (9b366d8) - Alessandro Pomponio
• add detect-secrets to dev dependencies (#484) - (ce9eb9d) - Alessandro Pomponio

Refactoring

• (test) support locked and unlocked runners (#508) - (8dd15e0) - Alessandro Pomponio
• (vllm_performance) rename image_secret to image_pull_secret_name (#514) - (b9677d7) - Alessandro Pomponio

Full Changelog: 1.4.1...1.5.0

Release 1.4.1

Highlights

✨ Features

• schema: add function to generate entity IDs (#471) — [5412418] (Michael Johnston)

🧹 Refactoring

• swap IPython pretty for rich formatting for cleaner output and less dependencies (#474) — [622a242] (Alessandro Pomponio)
• move imports into type-checking blocks for faster startup & cleaner deps (#461) — [ed8d6ab] (Alessandro Pomponio)

📝 Documentation

• website: various fixes and polish (#475) — [e76c968] (Michael Johnston)

---

📈 vllm_performance

• refreshed examples to match latest workflows and improvements (#472) — [7955ce6] (Michael Johnston)

---

Complete changes

Features

• (schema) add function for generating entity id (#471) - (5412418) - Michael Johnston

Documentation

• (changelog) update changelog (#467) - (f3628d7) - Alessandro Pomponio
• (vllm_performance) update examples (#472) - (7955ce6) - Michael Johnston
• (website) various fixes (#475) - (e76c968) - Michael Johnston

Refactoring

• use rich instead of IPython's pretty (#474) - (622a242) - Alessandro Pomponio
• move imports into type checking sections (#461) - (ed8d6ab) - Alessandro Pomponio

Full Changelog: 1.4.0...1.4.1

Release 1.4.0

🚀 Highlights

✨ Features

• Core: CSV sample store now supports specifying an actuator for more flexible data ingestion.
• Ray Tune: Introduced a difference stopper to halt sampling when its probable that the difference in two metrics is above or below a threshold

🐛 Fixes

• Core: Record relationships with nested operations when they are interrupted.
• Ordered Pip: pip_install_options now correctly propagates across all phases of an ordered install
• Modules: Correct type annotations for closures.

⚡ Performance

• Adopted NumPy’s new RNG APIs for faster and more consistent randomness.

🧰 Tooling & CI

• Broad dependency updates, pre-commit hook refresh, and stricter linters (ANN, Bandit).
• Test coverage reporting added.

📝 Docs

• Changelog refresh and HYBRID_SHARD documentation improvements.

---

🧪 sfttrainer

• Added support for fms-hf-tuning v3.1.0 and fixed related issues.
• Packaging streamlined: only the ado-sfttrainer wheel is propagated.
• Documentation updated for HYBRID_SHARD behavior.
• General code hygiene: type annotations & small readability tweaks.

🚄 vllm_performance

• New GuideLLM experiments to expand benchmarking coverage.
• Kubernetes namespace now RFC1123-compliant.
• Bumped requirement to vLLM ≥ 0.12.0 and refreshed dependencies.
• Hardening & cleanup: safer command invocation (no shell=True) and parameter upgrade-path removal.

⚙️ autoconf

• Added comprehensive type annotations for better IDE support and static analysis.

---

Complete changes

Features

• (core) allow specifying actuator in csv sample store (#455) - (bda00ad) - Michael Johnston
• (ray_tune) add difference stopper (#412) - (3d2f46c) - Michael Johnston
• (sfttrainer) add support for fms-hf-tuning==v3.1.0 (#442) - (b060d3d) - Vassilis Vassiliadis
• (vllm_performance) Add GuideLLM experiments (#459) - (46ffac8) - Christian Pinto
• (vllm_performance) ensure namespace is RFC1123-compliant (#358) - (2bf1072) - Alessandro Pomponio

Bug Fixes

• (core) add relationships to interrupted nested operations (#379) - (e6eacdb) - Michael Johnston
• (modules) use correct type annotations for closures (#447) - (335903b) - Alessandro Pomponio
• (ordered_pip) keep fields other than ordered_pip and inject pip_install_options to all phases (#390) - (f8f9def) - Vassilis Vassiliadis
• (sfttrainer) fix the support for fms-hf-tuning==3.1.0 (#448) - (e9bf2b2) - Vassilis Vassiliadis
• (sfttrainer) the only ado wheel to propagate is ado-sfttrainer (#431) - (b7583ec) - Vassilis Vassiliadis

Performance Improvements

• use new numpy RNG apis (#357) - (1c9c7f3) - Alessandro Pomponio

Documentation

• (changelog) update changelog (#377) - (7372902) - Alessandro Pomponio
• (sfttrainer) update documentation for HYBRID_SHARD (#389) - (4364b4e) - Vassilis Vassiliadis

Tests

• (core) add coverage report (#391) - (d7afb9c) - Alessandro Pomponio

Build system

• (deps) update dependencies (#466) - (1a12fa6) - Alessandro Pomponio
• (deps) update dependencies (#451) - (e1dc4a1) - Alessandro Pomponio
• (deps) update dependencies (#430) - (efe881d) - Alessandro Pomponio
• (deps) update dependencies (#361) - (ef91044) - Alessandro Pomponio
• (ruff) enable ANN linter (#440) - (a0c8a63) - Alessandro Pomponio
• (vllm_performance) update dependencies (#439) - (b5162ec) - Alessandro Pomponio
• (vllm_performance) require vllm>=0.12.0 (#426) - (866c067) - Alessandro Pomponio
• update pre-commit hooks (#371) - (b051c50) - Alessandro Pomponio

Refactoring

• (core) mark ADOResource.identifier as Defaultable (#446) - (93b8e34) - Alessandro Pomponio
• (core) change validate_model to instance method (#370) - (2020eb6) - Alessandro Pomponio
• (samplestores) disallow None parameters (#408) - (2b31405) - Alessandro Pomponio
• (sfttrainer) rewrite simple string joins as fstrings (#454) - (d6bf32a) - Alessandro Pomponio
• (tests) clean up tox file (#465) - (994b2f7) - Alessandro Pomponio
• (vllm_performance) remove upgrade path for parameters (#418) - (7746924) - Alessandro Pomponio
• (vllm_performance) avoid invoking bench command with shell=True (#359) - (0314c02) - Alessandro Pomponio
• remove commented-out code (#453) - (46a645f) - Alessandro Pomponio
• use Annotated pattern for pydantic models (#443) - (526af2e) - Alessandro Pomponio
• use Annotated type hint pattern for Pydantic models (#436) - (9a83af2) - Alessandro Pomponio
• auto-add type annotations where possible (#393) - (c7daab5) - Alessandro Pomponio
• enable ruff's Bandit linter (S) (#365) - (5eb333f) - Alessandro Pomponio
• rewrite PropertyDescriptor.eq (#349) - (0ce1db6) - Alessandro Pomponio

Style

• (anomalous_series) add type annotations (#425) - (31f20ae) - Alessandro Pomponio
• (autoconf) add type annotations (#423) - (7a3bd38) - Alessandro Pomponio
• (cli) add type annotations (#400) - (381e645) - Alessandro Pomponio
• (core) add type annotations (#414) - (bb9039e) - Alessandro Pomponio
• (example_actuator) add type annotations (#421) - (9efc5b4) - Alessandro Pomponio
• (examples) add type annotations (#434) - (174d5a9) - Alessandro Pomponio
• (metastore) add type annotations (#402) - (41b57f1) - Alessandro Pomponio
• (modules) add type annotations (#413) - (4adaaf9) - Alessandro Pomponio
• (ray_tune) add type annotations (#437) - (6cde7a1) - Michael Johnston
• (samplestores) add type annotations (#404) - (6d2c683) - Alessandro Pomponio
• (schema) add type annotations (#416) - (a215ec7) - Alessandro Pomponio
• (sfttrainer) add type annotations (#420) - (ca5548f) - Alessandro Pomponio
• (tests) add type annotations (#432) - (44bea4d) - Alessandro Pomponio
• (utilities) add type annotations (#406) - (c4c6a63) - Alessandro Pomponio
• enable PIE and T10 linters (#384) - (3d2f9f5) - Alessandro Pomponi...

Release 1.3.3

✨ Highlights

🚀 Features

• CLI plugins — Load commands via plugins for extensibility (#344).

🐛 Bug Fixes

• CLI details — Corrected calculations in show details operation (#325).

🧱 Build & Containers

• New images — Added support for geo and sft container builds (#341).
• Matrix builds — Build across multiple Python and CUDA versions (#318).
• Core environments — Defined required runtime environments (#323).

⚡ Performance & Quality

• PERF linter enabled for performance best practices (#333).
• Bugbear linter enabled to catch common issues (#330).

🧹 Refactoring

• Faster startup — Delayed expensive imports (#328).
• Cleaner comparisons — Rewrote ProbabilityFunction.__eq__ (#348).

📚 Documentation

• Actuator configs — Improved upgrade guidance (#343).

---

📌 vllm_performance

• Fix: Corrected in_cluster spelling (#326).

---

Complete changes

Features

• (cli) support loading commands via plugins (#344) - (6e61cff) - Michael Johnston

Bug Fixes

• (cli) update calculations in show details operation (#325) - (dfd4884) - Alessandro Pomponio
• (docs) update changelog link in pyproject (#351) - (f1094df) - Alessandro Pomponio
• (docs) update docs for upgrading actuator configurations (#343) - (7a5804d) - Alessandro Pomponio
• enable Bugbear linter (#330) - (f580b34) - Alessandro Pomponio

Performance Improvements

• enable PERF linter (#333) - (179c2b6) - Alessandro Pomponio

Documentation

• (changelog) update changelog (#321) - (66a60f0) - Alessandro Pomponio
• (vllm_performance) fix in_cluster spelling (#326) - (6b9f639) - Christian Pinto
• (website) fix typo in "Target v observed property formats" (#345) - (6da973a) - Daniele Lotito

Build system

• (containers) support geo and sft image (#341) - (dd0c0c7) - Alessandro Pomponio
• (containers) support building on multiple Python and CUDA versions (#318) - (4472e1d) - Alessandro Pomponio
• (core) add required environments (#323) - (5590c16) - Alessandro Pomponio
• (deps) update dependencies (#353) - (b6737ef) - Alessandro Pomponio
• (deps) update dependencies (#336) - (2888b7f) - Alessandro Pomponio
• (deps) update dependencies (#331) - (42c3c0f) - Alessandro Pomponio
• (deps) update dependencies (#320) - (2fc3b91) - Alessandro Pomponio

Refactoring

• (core) delay expensive imports (#328) - (28963fd) - Michael Johnston
• rewrite ProbabilityFunction.eq (#348) - (6282363) - Alessandro Pomponio

Miscellaneous Chores

• update gitignore (#342) - (3e14b3d) - Alessandro Pomponio

Full Changelog: 1.3.2...1.3.3

1.3.2

Highlights

Features

• Ray Tune: Added support for multi-objective optimization using Optuna.
• Core: Improved error handling when loading custom experiments.
• Core: Improved output validation for custom experiments

VLLM Performance

• Added benchmarking support for geospatial models.

Bug Fixes

• Core: Fixed shutdown with nested operations
• Ray Tune: Fixed imports in LHC sampler.
• Run Experiment: Fixed printing of results

Full Changes

Features

• (core) handle errors per custom_experiment (#314)
• (custom_experiments) enforce stricter rules on outputs (#315)
• (ray_tune) support multi-objective optimization with optuna (#307) - (d27f76c)
• (vllm_performance) add support for benchmarking geospatial models (#187)

Bug Fixes

• (ray_tune) update imports in LHC sampler (#310) - (63a1484)
• (run_experiment) print request series with use_markup=False (#319) - (6e9c078)
• (vllm_performance) add missing parameter to execute_random_benchmark and make geospatial experiments beta (#317)

Build system

• (deps) update dependencies (#311)

Refactoring

• (core) separate cleanup logic from signal handling and fix nested-operation shutdown (#281)

Full Changelog: 1.3.1...1.3.2

Release 1.3.1

🚀 Highlights

✅ Bug Fixes

• CLI: Fixed issue where Console.print took too long to print dataframes (#297)

🔧 Build System

• Added support for Python 3.13 (#291)

🛠 Refactoring

• CLI: Improved sizing of live results table (#299)

---

⚙️ Autoconf

• Pinned required AutoGluon version for consistent builds (#304)

---

Full changes

Bug Fixes

• (cli) do not use rich's Console.print with dataframes (#297) - (97c6bea) - Alessandro Pomponio

Documentation

• (changelog) update changelog (#287) - (6b63d40) - Alessandro Pomponio
• (website) update instructions to build python wheels for ado and plugins (#301) - (a62af24) - Vassilis Vassiliadis
• (website) simplify cli examples (#293) - (726aec9) - Michael Johnston

Build system

• (autoconf) pin the required autogluon version (#304) - (d51f324) - Srikumar Venugopal
• (deps) update dependencies (#300) - (a247dfe) - Alessandro Pomponio
• support Python 3.13 (#291) - (0ea5cbb) - Alessandro Pomponio
• update pre-commit hooks (#298) - (3ff6a6e) - Alessandro Pomponio

Refactoring

• (cli) improve sizing of live results table during operations (#299) - (83716ac) - Alessandro Pomponio
• (run_experiment) replace prints with console_prints (#289) - (e728921) - Alessandro Pomponio

Style

• format yaml files with yamlfmt (#286) - (e2eadfd) - Alessandro Pomponio

Full Changelog: 1.3.0...1.3.1

Release 1.3.0

✨ Changelog Highlights

🚀 Features

• CLI: Added support for --with in ado create (#262).
• Core: Enabled custom_experiments to run with or without Ray (#263).

🐛 Bug Fixes

• Custom Experiments: Fixed required_properties parameter to be optional (#278).
• Ordered Pip: Automatically re-create venv if garbage collected (#285).

---

⚙️ Autoconf

• Introduced custom experiments for autoconf (#255).

🧠 SFTTrainer

• Added support for granite-3.3-8b (#276).

📈 vLLM Performance

• Prevent multiple experiments from sharing the same Kubernetes deployment (#268).

---

Complete changelog

Features

• (autoconf) introduce autoconf custom experiments (#255) - (3c1fd87) - Srikumar Venugopal
• (cli) support --with in ado create (#262) - (6de12d6) - Alessandro Pomponio
• (core) allow custom_experiments to execute with or without Ray (#263) - (ea4cab7) - Michael Johnston
• (sfttrainer) support granite-3.3-8b (#276) - (3d1733c) - Vassilis Vassiliadis

Bug Fixes

• (custom_experiments) required_properties parameter of decorator was required instead of optional (#278) - (bec1a19) - Michael Johnston
• (ordered_pip) re-create venv if it has been garbage collected (#285) - (b327f16) - Vassilis Vassiliadis
• (vllm_performance) Avoid multiple experiments using the same kubernetes deployment at the same time (#268) - (34a64af) - Christian Pinto

Documentation

• (changelog) update changelog (#270) - (c768436) - Alessandro Pomponio
• (test) Add --reinstall flag to uv sync command (#274) - (40ce0b0) - Michael Johnston
• (website) more robust custom experiment docs (#284) - (30f0eb3) - Michael Johnston
• (website) clarify wheel build output location (#277) - (07d0061) - Vassilis Vassiliadis

Build system

• (deps) update dependencies (#282) - (4b4d8c2) - Alessandro Pomponio

Full Changelog: 1.2.4...1.3.0

Release 1.2.4

Highlights

🚀 Features

• CLI: Added shorthand support for resources (#245)
• Core: Actuators and operators now use Rich progress indicators (#248)
• Group Samplers: Improved performance for group generator sampler type (#229)

🛠 Bug Fixes

• Core: Entities now correctly show when missing or unmeasured (#254)
• Custom Experiments: Detection added for unexpected properties returned by experiments (#250)

⚡ vLLM Performance

• Fixed starvation of measurement requests (#249)
• Resolved incompatibility between example entity space and measurement space (#240)

---

Complete changes

Features

• (cli) support shorthands for resources (#245) - (39d4931) - Alessandro Pomponio
• (core) enable actuators and operators to use Rich progress indicators (#248) - (d930308) - Michael Johnston
• (group_samplers) improve performance for group generator sampler type (#229) - (7a79a23) - Christian Pinto

Bug Fixes

• (core) show entities missing/unmeasured (#254) - (f14ea2b) - Michael Johnston
• (custom_experiments) detect if custom experiment returns unexpected properties (#250) - (89375e2) - Michael Johnston
• (vllm_performance) Avoid starvation of measurements requests (#249) - (cd99105) - Christian Pinto
• (vllm_performance) example entity space was incompatible with measurement space (#240) - (fe43998) - Michael Johnston

Documentation

• (changelog) update changelog (#267) - (c7e1ee7) - Alessandro Pomponio

Build system

• (deps) update dependencies (#258) - (abe438b) - Alessandro Pomponio

Refactoring

• (cli) remove HiddenSingularChoice (#246) - (6edcc47) - Alessandro Pomponio
• (core) unify operation execution pathways and remove unused logic (#220) - (13f2259) - Michael Johnston
• (core) do not change signature of functions decorated with custom_experiment (#261) - (ece2f7d) - Michael Johnston

Full Changelog: 1.2.3...1.2.4

Release 1.2.3

🚀 Highlights

Core

• Prevented discarding of operation outputs when shutdown is set (#219)
• Split orchestrate into submodules for better modularity (#217)

vLLM Performance

• Fixed issue where k8s resources were not cleaned up on exception (#230)
• Upgraded vLLM to version 0.11.1 (#223)

Build System

• Updated core dependencies (#222)
• Moved scipy dependency to ado-ray-tune (#231)
• Added test dependency group in pyproject (#215)

---

Complete changes

Bug Fixes

• (build) regenerate lockfile (#239) - (427bae4) - @AlessandroPomponio
• (core) do not discard operation outputs when shutdown is set (#219) - (ac7c932) - @michael-johnston
• (docs) use correct indentation in sublists (#227) - (296aed2) - @michael-johnston
• (vllm_performance) k8s resource not marked for cleaning on exception (#230) - (48ece77) - @michael-johnston

Documentation

• (changelog) update changelog (#228) - (b637931) - @AlessandroPomponio
• (website) update remote raycluster execution (#221) - (3120648) - @michael-johnston

Build system

• (core) update dependencies (#222) - (f2bea1f) - @AlessandroPomponio
• (deps) move scipy dependency to ado-ray-tune (#231) - (322a866) - @michael-johnston
• (vllm_performance) bump vLLM version to 0.11.1 (#223) - (73768cc) - @christian-pinto
• create test dependency group in pyproject (#215) - (adb15eb) - @michael-johnston

Refactoring

• (core) split orchestrate into submodules (#217) - (730ea4b) - @michael-johnston
• (linting) use markdownlint-cli2 configuration file (#226) - (e43209e) - @AlessandroPomponio

Full Changelog: 1.2.2...1.2.3