Changelog

0.8.0 (2026-04-17)

Features

Bug Fixes

  • add –live option to flow list log (#630) (b044b6d)
  • add –provenance flag to flow list log (#649) (e0c28bd)
  • add –tag filtering support to flow list log (#626) (908af10)
  • add numbering to flow list log (#673) (ec57fd7)
  • add tags column to task display (#616) (4f4794d)
  • add tags to flow list log (#629) (04f0f70)
  • better handling of SSO refresh errors (#627) (6f6b097)
  • consistent –store param (#636) (158ae48)
  • flow check command improvements (#614) (d337fe5)
  • flow list log default to multiline format (#628) (6cfbab4)
  • improve –set help text (#622) (8859d5d)
  • improve and document filter behavior (#632) (b069e2f)
  • improve handling of KeyboardInterrupt during eval_set (#607) (ebf3789)
  • improve list log performance by reading samples async (#609) (06f4813)
  • improve performance of adding logs to store (#663) (bb91abc)
  • improved flow list log output format (#638) (37b6f3f)
  • including files with just @step function (#623) (922a9d1)
  • installing git+ additional depcencies (#672) (c0fb27a)
  • move duplicate logs before unnexpected logs in flow check output (#665) (d9f9f10)
  • remove unused _StepDecorator (#670) (ae3a829)
  • return results from check API method (#615) (c2f9318)
  • steps write new logs to store (#625) (499ebd6)
  • support factory instantiation in venv mode (#624) (bc90db4)
  • task differentiator when FlowFactory args are used (#647) (90a119c)
  • update display to show summary at bottom of box (#639) (8124c1b)

Documentation

  • add documentation for flow steps and flow check (#667) (2ea1895)
  • clarify list log falls back to default store when PATH not provided (#634) (c60f83a)

0.7.0 (2026-03-24)

Features

  • FlowFactory to provide type checking of factory argumnets (#580) (8403977)
  • list log command (#575) (87e6b28)
  • store_read and store_write flags. Default store_read off. (#585) (d9f75b6)

Bug Fixes

  • add support for tags on FlowTask (#593) (4d070d7)
  • add viewer url column to flow list log (#586) (c877ecb)
  • count of samples in log when results is missing or invalidated (#598) (8bca19e)
  • do not output file:// prefix in list log tree format (#581) (3fe7859)
  • ensure inspect_ai log level is set correctly before calling eval_set (#576) (78f2923)
  • improved output for store usage (#591) (dc38b34)
  • preserve task name when provided (#563) (a02da20)
  • resolve relative paths on import (#597) (e5d3268)
  • set default display to rich (#577) (77ad180)
  • support multiple –store-filter (#596) (617c9be)
  • support multiple log filters (#592) (1dcdf32)
  • update to new embedded_viewer path (#590) (ebfbe8b)

Documentation

0.6.0 (2026-03-16)

Features

Bug Fixes

  • –copy-from support for s3: paths (#551) (e1fe49a)
  • add embed_viewer location to output (#550) (7a87b43)
  • add support for embed_viewer option (#544) (d310ef5)
  • improve output when both bundle and embed_viewer are set (#553) (2b285a9)
  • print paths using console wrapping so they are copyable (#554) (2ea750d)
  • re-set moto env vars before each S3 test (#543) (216fdeb)
  • Update dependencies (includes embed_viewer fix) (#561) (27d4a4b)
  • update inspect and support new fields and limit matrixing (#538) (553e6c1)
  • upgrade inspect-ai and fix absolute path issue (#542) (a47cbd3)

Documentation

  • add documentation for matrix limits, function configs, and YAML support (#540) (7a2ffc0)
  • document embed_viewer option in viewer bundling section (#546) (dd53edd)

0.5.0 (2026-03-06)

Features

  • add –resume flag to reuse previous log dir (#532) (a2d9ea9)

Bug Fixes

  • –log-dir-create-unique creates subdir with the current timestamp (#526) (392e34e)
  • add DATETIME substitution (#527) (2ad8104)
  • add task name as context to display on instantiation errors (#536) (ad14540)
  • ensure bundle url ends with a / (#515) (f04a04a)
  • s3 store paths with trailing slash (#535) (8c6568a)
  • support reading default model from env var (#524) (e4eafe9)
  • upgrade packages and import TASK_IDENTIFIER_VERSION from inspect (#513) (5a2f846)
  • use absolute URLs for README images so they render on PyPI (#530) (f6f8f35)

Documentation

  • add llms.txt and llms-full.txt generation (#517) (959cf28)
  • improve API reference docs (#523) (f1fc94f)
  • update docs for –resume, {DATETIME}, and –log-dir-create-unique (#529) (f32fa5c)

0.4.1 (2026-02-20)

Bug Fixes

  • add docstring to DisplayType for quarto docs build (#504) (dfdf661)
  • compile with conflicting git urls (#508) (b0f3324)
  • default to have eval_set use the flow log_level if not explicitly set in FlowOptions (#507) (8bf12d5)
  • tests to handle inspect_ai installed from git (#506) (26b5925)

0.4.0 (2026-02-19)

Features

  • flow store to enable log reuse across directories (#340) (bc5d993)

Bug Fixes

0.3.0 (2026-02-03)

Features

Bug Fixes

  • _flow.py should not include itself when loaded directly (#467) (3e26333), closes #465
  • log-dir-create-unique corrected handling of paths ending in / (#451) (2f5b3ab)
  • test bundle_dir location and stricter type checking (#463) (11f2328)
  • use rich display of exception traceback (#459) (70d0ff7)

0.2.2 (2026-01-26)

Bug Fixes

  • remove duplicated inspect_ai code (#444) (c6c141c)
  • use inspect-ai load_tasks to ensure task path is captured correctly (#441) (b34e303)

0.2.1 (2026-01-23)

Bug Fixes

  • apply substitutions, defaults, and call @after_load functions after overrides (#440) (0b507c9)
  • enable relative imports for local file (#436) (50d8a43)
  • RuntimeError when calling run while loading a spec file (#434) (411aa08)

Documentation

  • document extra_args for per-task customization (#439) (8210b3a)

0.2.0 (2026-01-22)

Features

  • default to running inproc and support for inspect-ai types (#421) (0ad7ff1)
  • support specifying tools for a FlowTask (#401) (3a5903a)

Bug Fixes

Documentation

0.1.4 (2026-01-14)

Bug Fixes

  • add hashes to flow-requirements.txt (#381) (e870d79)
  • correctly write flow-requirements.txt to s3 (#375) (abaecd4)
  • improved str representation of flow_types (#382) (eb04076)

0.1.3 (2026-01-06)

Bug Fixes

  • ensure logging works in run process in venv (#343) (530eb8e)
  • improved errors on duplicate tasks (#353) (21fd03c)
  • remove unneeded generated types (and package upgrade) (#349) (844ab79)
  • support s3 paths on CLI (#339) (b63f483)
  • use python version specified in uv.lock file (#352) (291a89b)

0.1.2 (2025-12-15)

Bug Fixes

  • add logging of include files (#330) (29f6c8c)
  • add uv_sync_args to dependency configuration (#328) (89dfd51)
  • Do not write flow-requirements.txt if dry_run (#335) (fd1d42a)
  • maintain package version for installed packages in venv (#332) (a7564db)
  • warning when multiple dependency files found (#337) (3fa1239)

0.1.1 (2025-12-11)

Bug Fixes

  • Rename FlowJob to FlowSpec
  • Fix log paths in Inspect output
  • Rename bundle_url_map to bundle_url_mappings
  • Apply defaults when loading FlowSpec

Documentation

0.1.0 (2025-12-05)

Features

  • Flow support for configuration repeatability and sharing