v 0.0.14
* Based on a2a-acl v0.0.14.
* Attached files in incoming tell-messages are added as the parameter of the content of the message.
    That content is not parsed by the firewall or by the AgentSpeak parser.
    Replaces former usage of upload performative.
* Illocution UPLOAD removed (see above).
* Illocution FAILURE removed. Use A2A task status updates instead.
* Action .send_str renamed into .send_bytes. With .send_bytes the content of a message is not parsed.
* Rename performative tell into inform and achieve into request to align with FIPA-ACL.

v 0.0.13
* Based on a2a-acl v0.0.13.
* Langfuse client included in some examples of agents that use an LLM.
* Actions to notify updates on A2A tasks and to attach an artifact.
* Standard achievement targets to react to task updates.
* Incoming messages with "failure" illocution handled as tell/failure messages.
* New example with a "reference agent" that provides interaction with the user.
* The AgentSpeak agent is single thread / single task and notification for new messages or task updates from the A2A layer are handled in a queue.
* Each achievement resulting from an incoming message creates a new A2A task with the status submitted. The id of that task is available as a tag.
* Message sending can return a message_id.

v 0.0.12
* Based on a2a-acl v0.0.12.
* Streaming task updates for sender when receiving an achieve message.
  Thus the client is aware of failures in agent behavior.
  See the new action .send_str_cb .
* New illocution 'upload' to allow transfer of data without parsing by the firewall.
  See the llm_refactoring example.


v 0.0.11
* Tested with a2a-acl v0.0.11.

v 0.0.10
* Interfaces (in *.asi files) use illocutions (tell, ask, achieve, propose) to declare skills instead of old keywords (input, belief, action).
* New internal action .is_alive to check is an A2A agent is online. See test in directory tests/is_alive.
* Two internal actions to send strings : .send_str (without callback) and .send_str_cb (with a callback).