v2.3.0

What's Changed

• 🐛Fix test_draw failure caused by unreliable bmp format support in Graphviz
• 🏗️ Feat add support for Python 3.13

Full Changelog: v2.2.0...v2.3.0

Pipelime v2.2.0

What's Changed

• feat: ✨ Implemented a loguru listener callback, which forwards DAGs progress to loguru in JSON format

Pipelime v2.1.0

What's Changed

• feat: ✨ implement OutputValueInterface, which allows saving output values with primitive types to json file automatically
• fix: ✅ fix TUI toggle descriptions test

Pipelime v2.0.0

What's Changed

• 💻 Several fixes for the TUI, to adapt it to the latest updates of textual.
• ⚡ Multiprocessing is now done with multiprocessing from standard library, insted of billiard which caused random deadlocks. This change prevents spawning nested processes by default, but it is possible to enable it with a flag in the grabber ( ⚠️ this may lead to zombie processes and unreleased resources in case of unhandled crashes).
• 💣 ✨ 🔥 pipelime finally supports pydantic v2! Internally, we still use pydantic v1 by doing import pydantic.v1 as pyd. This can be done starting from pydantic==1.10.17, so that final users can still use pydantic v1, if needed.
• 💥 The data cache is disabled by default for all items! A new option can be passed to the CLI to enable it for all the items or for a specific subset.
• ✨ It is now possible to create a SamplesSequence from a single video (each sample will contain one frame of the video).
• 🆙 Python supported versions have been updated to 3.10, 3.11 and 3.12.
• 🏗️ Tox can now be used to run tests against all supported python versions.

Full Changelog: v1.9.1...v2.0.0

Pipelime v1.9.1

What's Changed

• 91% test coverage
• 🩹 fixed ShellCommand
• 🚸 if LOGURU_LEVEL is not set as env var, then -v sets the logger level from WARNING (no -v) to TRACE (-vvv or more)
• 🗑️ WatcherNodesGraphExecutor has been deprecated (use DirectTrackCallback instead)

Full Changelog: v1.9.0...v1.9.1

Pipelime 1.9.0

What's Changed

🎉 New Features!

• 🚸 last pipelime call (with no --ckpt option) is saved and can be replayed with pipelime resume (❗TUI partial inputs included❗)
  • 👉 actually, last 3 calls are cached: try this pipelime resume +ckpt 2 to replay the second-last
• ✨ cat (both pipe and command) accepts multiple sequences with optional interleave flag
• ✨ batched/unbatched pipes to group/ungroup multiple samples
• ✨ InputDatasetInterface and OutputDatasetInterface accept a custom pipe; now you can:
  • ⚙️ add a custom processing on input/output dataset
  • 🔀 completely replace the data source with your own
  • ✏️ skip the default output underfolder

🪛 Many Fixes And UX Improvements!
*🪄 interactive context filling when, eg, some $vars are missing
*🧵 multiprocessing should now eat almost everything, lambda functions included

• 🧑‍💻 better error reporting in case of malformed @command or failing imports
• 📝 pipelime lc lists DAG commands in a separate section
• 🚸 when pipelime help finds a pydantic Field with expand_help=True, such field will be expanded if the type is a BaseModel
• 🐛 fixed split command with size 1

Full Changelog: v1.8.1...v1.9.0

Pipelime 1.8.1

What's Changed

• stub for SamplesSequence.from_images had an incorrect default
• fixed error when loading extra .py files multiple times
• fixed some use-cases in dag creation and drawing
• fixed missing variable in FileListenerCallback
• fixed pipelime list that was not showing anything

Full Changelog: v1.8.0...v1.8.1

Pipelime 1.8.0

What's Changed

• 🎉 Big news:
  • 🛸 Resume from checkpoint! -> especially when you run a DAG
  • 🧑‍💻 When you import a python symbol, eg, $symbol, CallableDef etc, now you can even use plain python code or lambda -> and it is fully compatible with a multi-process grabber!
  • 🧑‍💻 CLI: no need to repeat a +/@ option when you want to append a value -> see the docs
• 🕸️ DAG news:
  • 🛸 Resume/Replay a DAG
  • ✏️ Advanced node selection -> see the docs
  • 🐍 Python DAGs:
    • 📝 Checkout the docs
    • 🧑‍💻 create_graph can return a list of (lazy) PipelimeCommands: node names will be auto-generated
    • 🗑️ Temp folder is not automatically delete if you use --keep-tmp
• ✨ Commands, stages and pipes:
  • stage-time command to measure the average time spent by each stage of a pipeline
  • resume command to load and run a Pipelime checkpoint
  • map-if command to apply a stage only on a subset of samples
  • repeat now accepts floating point values to repeat a sequence a fractional number of times
• 🚶 General improvements:
  • 🧑‍💻 PipelimeTmp.make_session_dir() now always returns a subfolder of the Pipelime session temp folder
  • 🐛 @command now correctly manages submodel parameters
  • ✨ Input/Output interface now have a default piper port
  • ✨ The pydantic type NewPath can check or set the extension for you

Full Changelog: v1.7.0...v1.8.0

Pipelime 1.7.0

What's Changed

• fix: 🐛 Choixe now supports non-string dictionary keys
• fix: 🐛 Input/Output command ports can now be mappings
• feat: 🧑‍💻 Decoupled DAG command and DAG creation
• feat: ✨ Remove cached data after serialization

Full Changelog: v1.6.2...v1.7.0

Pipelime 1.6.2

What's Changed

• hotfix: 🚑️ Missing imports

Full Changelog: v1.6.1...v1.6.2