search

Release Notes

Release Date: February 18, 2026

Welcome to Akamas 3.7.0. This release introduces powerful study editing capabilities, improved study control, and enhanced platform stability. The ability to edit study goals and have all past trial scores automatically recomputed represents a fundamental shift in optimization workflows—allowing you to pivot strategy while preserving all historical data.

hashtag
Study Goal Editing

Edit study goals without losing historical data or re-running past experiments.

Stop a study, update the optimization goal, then resume. Akamas automatically recomputes scores for all previously executed trials against the new goal definition before continuing. No experiments need to be repeated.

Editable goal properties:

  • Formula — the metric expression being optimized

  • Objective — minimize or maximize

  • Constraints — bounds and thresholds applied to the optimization

hashtag
Complete Study Export and Import

Study exports now include execution logs. Previously, log records were dropped during export and were missing after import. Exported archives now carry the full log history, and import restores them correctly alongside all other study data.

hashtag
Enhanced Study Control

Greater flexibility in optimization workflows through experiment termination and trial state management.

New commands enable surgical control over running studies: terminate a specific experiment without stopping the entire study (akamas finish experiment), or update individual trial states to mark them as failed or finished (akamas update trial --fail|--finished). These capabilities enable cost-efficient recovery from transient failures.

Imported studies are now fully resumable — the resume function correctly restores trial context and experiment history. Multi-trial experiments can also be resumed immediately after creation without triggering orchestration errors.

hashtag
3.7.0 Improvements

hashtag
User Interface

  • Study editing dialogs: Redesigned for clarity with real-time validation and preview capability

  • Configuration comparison: NEW - Compare trial configurations side-by-side with filtering and sorting capabilities

  • Error messages: Now include actionable remediation steps instead of error codes

  • Trial visualization: Enhanced charts and metrics for better progress tracking

  • Responsive design: Improved mobile and tablet support for study dashboards

  • Offline optimization: Redesigned with full workflow support, enhanced charting, and better trade-off visualization; large datasets (1000+ trials) load in seconds

hashtag
Platform Stability

  • Study state management: Fixed corruption during long-running optimizations

  • Constraint validation: Better handling of edge cases and conflicting constraints

  • License enforcement: Study operations now validate capacity before execution

  • Error recovery: More robust handling of mid-execution failures

hashtag
Data Management

  • Import/export: Study data preserves all metadata, constraints, and logs without loss

  • Audit logs: Audit log records are correctly exported and restored with study data

hashtag
CLI and Automation

  • Rate limiting: CLI automatically detects API limits and retries with exponential backoff

  • Pagination options: New --no-pagination flag for scripting

  • Batch operations: Endpoints for bulk study updates reduce API call count

  • Updated commands: akamas update study now accepts a YAML file to update study goal and constraints

hashtag
Infrastructure

  • Kubernetes 1.33: Full compatibility with latest Kubernetes features

  • PostgreSQL 16: Database performance improvements and enhanced tooling

  • Container management: Better resource management and custom registry support

  • Telemetry: Upfront configuration validation prevents failed provider installations

  • Custom Metrics Provider: NEW - Built-in provider for ingesting simulated metrics; run and validate studies without a live system (see documentation)

  • Sample study (Ecommerce Cart): New installations include a pre-built study for immediate hands-on exploration

hashtag
Resolved Issues

This release resolves 68 reported issues:

Critical Fixes:

  • Study state corruption during long-running optimizations

  • Data loss when exporting studies containing logs

Bug Fixes (selected):

  • Study editing dialogs missing constraints

  • Experiment resume failing for multi-trial scenarios

  • Trial visualization not updating after resumption

  • Study import failing for constrained studies

  • Study export failing for large datasets

hashtag
Upgrade from 3.6.x

  1. Follow the PostgreSQL 16 migration procedures to migrate your database before upgrading Akamas services.

  2. Update platform images to 3.7.0, following the guide for your deployment type:

  3. Restart services

  4. Database schema is automatically migrated

  5. Sample study is created automatically

hashtag
Documentation

Last updated

Was this helpful?