* updated dependencies versions
* Updated model validator for HuggingFace-Hub client to catch model and hub url early
* split task registration into discover + register phases, improved LLM client init in tasks and workflow wrappers
* improve decorator to access method attributes
* wrapping workflow decotrators to log, validate, etc., without losing signature/docs
* Improved LLM-based task client and cleaned execution of LLM, agent and python function
* Added an example of multiple models being defined per workflow task after updates
* Updated quickstarts basic agent runs to async
* Added model attribute to huggingface-hub client class
* Fixed random and roundrobin orchestrators TriggerAction schema, trigger action and task to process agent response
* Updated quickstart multi-agent workflows and actor-based agents docs
* added .dapr to gitignore
* Moved Messaging under Workflow and modularized main components.
* Removed WorkflowService and added as_service() to AgenticWorkflow for pub/sub support in headless and service modes.
* Added utility to safely retrieve instance methods decorated with a specific attribute
* Updated Agent and Orchestrator workflows to support the latest messaging system changes.
* Updated AgentActors to integrate with latest messaging system under workflows.
* Made service_port optional to allow random port assignment.
* dapr agents imports and logging
* Updated Multi-agent workflows, deleted JSON logs and tested NVIDIA LLM client notebook
* Integrated new messaging system into multi-agent quickstarts
* Updated agentic workflows docs to reflect changes
* Updated quickstart multi-agent workflows app to run in headless mode
* Fixed Orchestrator as a service and exposed built-in route to start workflow over HTTP
* Updated INFO logging to DEBUG
* updated cookbook with latest changes
* updated quickstart mukti-agent workflows to integrate orchestrator as service change
* Updated docs to trigger agentic workflow over HTTP by default.
* Upgrade Jinja2 to address sandbox breakout vulnerability