Release 0.2.5

[gubaidulinvadim]

New Features

• ConfigurationManager — explicit multi-step build flow for accelerator configuration; supports REST as a config source
• YellowPages — standardised element name mapping/discovery with query support
• UnboundElement — validation and support for unbound element configuration
• CfgDict — allows plain dicts in config (pydantic-compatible)
• MeasurementTool / TuningTool base classes — unified interface for tuning/measurement
• TuneResponseMatrix / ChromaticityResponseMatrix — new RM tools for tune and chromaticity
• RF orbit correction — RF frequency as a correction variable in orbit tools
• Wildcards & regex in YAML array definitions (~ for exclusion); also Python-based macros via elements_code
• ATuneMonitor diagnostics class
• Shared Tango attribute state in dummy control system
• Automatic versioning (version from build)
• OphydDevice example added

---

Bug Fixes

• LinearSerializedMagnetModel: fixed 4 AttributeError bugs (wrong field names / misplaced attribute access)
• Serialised magnets: fixed sub-magnet list, strength computation (returns per-magnet + sum)
• Private peer access cleaned up to use public property
• Factory error reporting fixed (with line numbers)
• Chromaticity tool fixes
• ignore_external flag fix
• YAML list expansion fix
• UnboundElement config validation
• Discrepancies between tuning and measurement tools
• Averaging bug fix in bc3aa8d

---

Refactoring / Cleanup

• ResponseMatrix renamed to ResponseMatrixData; OrbitResponseMatrixData added
• Renamed input_/output_ → variable_/observable_ (AT naming convention)
• Unified orbit/tune response matrix architecture
• Removed factory custom strategy
• get_peer() renamed to get_peer_name() for clarity
• Peer added to ElementHolder
• Ruff line limit raised to 127; reformatted codebase
• ElementHolder discovery methods are made protected

---

Examples & Docs

• SOLEIL: split config into devices.yaml, arrays.yaml, tuning_tools.yaml; added orbit correction notebook
• BESSY2: updated to new API
• ESRF: added dispersion correction and weight-scan examples
• CONTRIBUTING.md updated for release workflow

[gubaidulinvadim]

Could someone review this? @TeresiaOlsson @kparasch @GamelinAl It's just a changelog to make a new release. JLP create a 0.2.5 tag but I don't think it was actually released. There's no release on GitHub and I quickly checked that pip install 0.1.1 version (which is very wrong)

[TeresiaOlsson]

There is a release of version 0.2.5 on PyPi from 4 May: https://pypi.org/project/accelerator-middle-layer/

As far as I know our Github workflows don't make release notes or releases on Github. They only build and publish the package to PyPi. I think that is an issue and should be fixed. A proper release workflow should make a release on Github first and then publish to PyPi. In issue #210 I had to make the release on Github manually.

So perhaps you can add the missing part in the workflow as part of this PR?

[TeresiaOlsson]

I think it's also not a good solution to have a CHANGELOG.md file. It is better to have Github automatically generate the release notes or alternatively have the workflow make a draft and then add the changelog manually to the release notes. The changelog is then part of the release and not a file that needs to be kept in version control and manually updated every time a release is done.