feat(examples): add multi-agent system demo\n\n- Message passing between agents\n- Task coordination with capability matching\n- Consensus protocol for collective decisions\n- Feedback loop for iterative task improvement\n- 4 comprehensive demos showcasing collaboration patterns

feat(agents): add multi-agent collaboration framework\n\n- MessageBroker for centralized communication\n- CollaborativeAgent with message handlers\n- TaskCoordinator for task distribution\n- ConsensusProtocol for multi-agent decisions\n- Priority-based message queuing and broadcasting

feat(agents): add feedback loop and reinforcement learning\n\n- FeedbackLoop for iterative refinement with learning\n- Pattern extraction from successes and failures\n- AdaptiveFeedback for strategy selection\n- ReinforcementSignal with epsilon-greedy action selection\n- Convergence rate tracking and iteration summaries

feat(examples): add goal-oriented agent demo\n\n- Goal decomposition with hierarchical planning\n- Stub execution via tool registry\n- Self-critique with improvement prompts\n- Memory summary for recorded actions

test(agents): add tests for tool registry and enhanced memory\n\n- Register/execute tools and batch execution\n- Default tools smoke test\n- EnhancedMemory record/recall and working context

feat(agents): add context manager for long-term context\n\n- ContextWindow for timestamped content with importance\n- Automatic compression when approaching token limits\n- Summarization of old low-importance windows\n- Recency-based importance adjustment\n- Context retrieval within token budget\n- Statistics and status reporting

feat(agents): add self-critique system\n\n- AgentCritic for multi-dimensional output evaluation\n- Critique types: correctness, completeness, efficiency, clarity, safety\n- Configurable thresholds and retry logic\n- Improvement prompt generation\n- Critique report formatting

feat(agents): add goal decomposition system\n\n- Task class with status, dependencies, and hierarchy\n- GoalDecomposer for breaking down high-level goals\n- Pattern-based decomposition (research, creation, compound)\n- Topological sort for execution order\n- Task visualization with dependency tracking

feat(examples): add integrated agent demo\n\n- Agent with tools + memory + reasoning chains\n- Tree of thoughts planning with memory tracking\n- Multi-agent collaboration with shared memory\n- Entity tracking and event timeline\n- Full end-to-end agentic workflow demonstration

feat(agents): add enhanced memory system\n\n- SemanticMemory for facts, entities, and knowledge\n- EpisodicMemory for event sequences and timelines\n- WorkingMemory for current context\n- EnhancedMemory unifying all three layers\n- Entity tracking with properties and mention counts

feat(agents): add tool registry system\n\n- Tool class with validation and execution\n- ToolRegistry for registration and discovery\n- Batch execution support\n- Built-in tools: search, calculate, get_time, summarize\n- JSON export for tool descriptions

docs: update README with anomaly detection and reasoning chains quickstarts\n\n- Add agents/anomaly to structure\n- Anomaly CLI and demos usage\n- Reasoning chains overview and test command

test(agents): add comprehensive tests for reasoning chains\n\n- ChainOfThought tests (basic, confidence, format)\n- SelfReflection tests (confidence eval, weak step detection)\n- ReasoningChain tests (success, validation, error handling)\n- TreeOfThoughts tests (branching, navigation, path tracking)\n- 15 tests covering all core functionality

feat(agents): add reasoning chains module\n\n- ChainOfThought for step-by-step reasoning\n- SelfReflection for confidence evaluation\n- ReasoningChain with validation\n- TreeOfThoughts for branching exploration\n- Formatted output for debugging

feat(examples): add advanced anomaly detection demo\n\n- Ensemble detection with multiple detectors\n- Drift detection example\n- Detector sensitivity analysis and comparison\n- Multi-metric evaluation (F1, precision, recall)