CHANGELOG

2.0.0 - 2025-12-31

Instance-Scoped Supervisors: Jido now uses user-owned, instance-scoped supervisors instead of auto-started global singletons
Jido Supervisor Module: New Jido module acts as a Supervisor managing per-instance Registry, TaskSupervisor, and AgentSupervisor
Agent Lifecycle API: New functions Jido.start_agent/3, Jido.stop_agent/2, Jido.whereis/2, Jido.list_agents/1
Test Isolation: New JidoTest.Case module for automatic test isolation with unique Jido instances
Per-Instance Scheduler: Scheduler can now be scoped to a Jido instance
Multi-Agent Coordination: New Jido.await/3 and Jido.await_child/4 for coordinating agent completion
Discovery Delegates: Discovery functions accessible via main Jido module

Explicit Supervision: Users must now add {Jido, name: MyApp.Jido} to their supervision tree
Agent Start: Use Jido.start_agent(instance, agent, opts) instead of AgentServer.start/1
Instance Option: Pass jido: MyApp.Jido option when starting agents directly

See the Migration section in README.md for upgrade instructions.

Modular API Architecture: Refactored monolithic Jido.ex module into specialized modules:
- Jido.Agent.Lifecycle - Agent lifecycle management (start, stop, restart, clone operations)
- Jido.Agent.Interaction - Agent communication (signals, instructions, requests)
- Jido.Agent.Utilities - Helper functions (via/2, resolve_pid/1, generate_id/0, log_level/2)
Comprehensive Test Helpers: New JidoTest.Support module with common test utilities:
- Registry management helpers (start_registry!/0)
- Agent lifecycle helpers (start_basic_agent!/1, automatic cleanup)
- Signal and state assertion helpers
- Consistent test patterns across all modules
Table-Driven Test Coverage: Implemented data-driven delegation tests for maintainable 100% API coverage
Performance-Optimized Test Suite: Consolidated and streamlined tests with @tag :slow for optional performance tests

None: All public API functions remain fully backward compatible through delegation in main Jido module

Improved Test Performance: ~40% reduction in test execution time through:
- Elimination of redundant test cases
- Reduced UUID generation from thousands to ~110 total
- Parameterized test scenarios instead of copy-paste patterns
- Common setup helpers reducing boilerplate
Enhanced Test Maintainability:
- Centralized test helpers eliminate code duplication
- Table-driven tests make adding new API methods trivial
- Consistent patterns across all test files
Better Resource Management: Automatic test cleanup prevents resource leaks and test isolation issues

Test Suite Optimization: Removed redundant tests while maintaining high coverage (68.4% overall, >80% for key modules)
Code Quality: Addressed Credo warnings in test files (replaced length/1 with Enum.empty?/1)

Automatic Action Registration: Skills can now declare required actions in their configuration using the actions field, which are automatically registered when the skill is mounted
Module Path Refactoring: Moved action modules from Jido.Actions.* to Jido.Tools.* namespace for better alignment with jido_action library
Default Skills: Agent server now uses default skills (Jido.Skills.Basic and Jido.Skills.StateManager) instead of hardcoded actions

Action module paths: All action modules moved from Jido.Actions.* to Jido.Tools.* to align with the jido_action library
Runner system removed: The separate Jido.Runner.Simple and Jido.Runner.Chain modules have been deprecated. Agents now use built-in execution logic that processes one instruction at a time (equivalent to the former Simple Runner behavior)

Skill configuration: Skills can now declare required actions using the actions field for automatic registration
Agent options: Both actions: and skills: options are supported, with skills providing the preferred declarative approach

Task update action now only updates provided fields instead of overwriting with nil values
State manager update action now uses :update operation instead of :set for proper semantic behavior
Improved error handling for action registration failures