0.24.0

• Added task parameter groups. Task parameters can now be grouped in YAML under parents up to 1 level deep, and groups are rendered in UI
• Reduced clutter in the task properties panel, as well as in submission dialogs
• Added description tooltips for parameters in the task properties panel
• Added effort slider for Claude Agent SDK models in the Beacon optimizer UI
• Improved parameter default value handling by showing them in the UI rather than having blank fields
• Fixed parameter default values when submitting protocol runs
• Fixed auto-rounding when trying to type floating point numbers in the UI
• Added tests for parameter defaults

0.23.0

• Added eos pkg install CLI command for installing user package dependencies more easily
• Added eos update CLI command to pull master, sync dependencies, and run database migrations in one step

0.22.0

• Changed Bayesian optimizer random sampling to use BoFire's RandomStrategy when using constraints, which is more robust
• Improved protocol run execution view in the UI to update task details when task statuses change
• Improved campaign submission form when Beacon optimizer is used to show loading indicator while its loading default parameters
• Fixed error page in the UI to use dark theme when dark theme is enabled
• Fixed failing tasks not being marked as failed in the DB
• Added test for task failure

0.21.0

• Improved and fixed device management code to properly handle device loading, unloading, and reloading
• Fixed Bayesian optimizer random sampling not respecting constraints via rejection sampling
• Improved how error messages are displayed in the web UI
• Added copy button for error messages displayed in the web UI
• Added tests for device loading, unloading, and reloading
• Added tests for task reloading
• Fixed bugs

0.20.0

• Added scheduling simulator in the web UI, allowing users to simulate how protocol tasks would run, estimate parallelism and resource utilization, and more
• Added file change detection in the web UI editor page, warning users if source files on disk were changed outside the web UI before saving
• Added resources view in the web UI, allowing users to view the state of all resources in EOS, such as view metadata, and reset their state to default
• Fixed several bugs

0.19.0

Breaking Release: This version renames experiments to protocols/protocol runs.
Existing user code that references experiments will need to be updated.
A migration script is provided in scripts/migration/migrate_experiment_to_protocol.py.

• Replaced experiments with protocols and protocol runs to describe more general lab workflows beyond experiments
• Refactored code to support the new naming
• Added search for parameters in web UI submission dialogs
• Optimized submission dialog layout to be more vertically compact
• Collapsed Beacon optimizer domain on submission dialogs by default
• Added delayed automatic refresh after making a submission
• Fixed dark theme font color of timestamp columns
• Updated documentation

0.18.0

• Improved entity cloning feature in the web UI to increment a number rather than appending _clone suffix
• Improved task, experiment, and campaign tables to use correct pagination
• Created Docker setup for running the web UI

Thanks @dirkzon and @yordan-ov !

0.17.0

This is the biggest EOS update to date, bringing a brand-new web UI for EOS, a powerful new default optimizer called Beacon combining traditional methods like Bayesian optimization with cutting-edge methods like LLMs, an MCP server for connecting AI agents to EOS with 50 tools, and many improvements to the EOS scheduler and EOS internals.

• New web UI for designing experiments visually, submitting tasks/experiments/campaigns, monitoring optimization, inspecting devices, browsing files, streaming logs, and managing packages.
• New Beacon optimizer that combines Bayesian optimization with LLM reasoning.
• New MCP server allowing AI assistants like Claude to connect to EOS and interact with EOS.
• New device and resource holds in the scheduler to keep allocations locked between tasks so other experiments can't claim them mid-workflow.
• New scheduling simulator (eos sim) to simulate experiment scheduling offline without hardware.
• New REST API endpoints for optimizer state, log streaming, and package management.
• Error messages are now stored for failed tasks, experiments, and campaigns.
• On-demand tasks now go through the scheduler, improving efficiency and stability.
• Optimizer sampling is now async and no longer blocks the orchestrator loop.
• CP-SAT scheduler runs in a thread executor to avoid blocking the event loop.
• Campaign optimizer parameters can be overridden at submission time
• EOS now lazy-loads optimizers, reducing startup time and memory usage.
• Fixed a bug that caused Ray to consume large amounts of disk space.
• Refactored parts of the codebase.
• Improved performance in several parts of the codebase.
• Various bug fixes and stability improvements.
• Updated documentation.
• Updated dependencies.

0.16.0

• Optimized the exclusive allocation of devices and resources. It now scales much better with more concurrent requests.
• Fixed the optimizer to properly handle initial samples with batched sample requests.
• Fixed the interrupt signal handling for EOS getting intercepted by Ray and preventing cleanup logic from firing.

0.15.0

• Replaced MinIO with SeaweedFS
• Made the file DB interface S3-compatible
• Changed the file DB interface to be async
• Made optimizer calls in campaigns non-blocking. This allows the EOS loop to keep going instead of optimizers blocking work
• Made priorities in the CP-SAT scheduler soft to avoid infeasible schedules
• Changed CP-SAT optimization to have 1 phase rather than 2, speeding up solving
• Fixed reference resolution in the CP-SAT scheduler to properly follow reference chains
• Changed the CP-SAT scheduler to use relative time to avoid possible int64 overflow