Re: Agentic Communication Protocols and MIME Messages

Dear everybody,
@Lorenzo, thanks for bringing this up, as apparently Adam's message flew 
under my radar.
@Adam, that's very interesting personally, not much for the MIME part 
(that I believe is brilliant but out of my expertise), but for the 
(computational) argumentation part.

I believe that, in a world where "agents" mostly means "LLM-based" 
software (not to say that I agree with this vision...), the natural and 
most appropriate way of "flexibly" coordinating them within a 
multi-agent system is by means of argumentation models and protocols.

     (my earlier message in another thread about FIPA protocols was not 
to say that FIPA is the way forward, but to bring an example of solid 
research that is being mostly ignored in today's "agentic AI" hype)

Incidentally (and I truly, wholeheartedly apologise for the shameless 
plug), before all the hype about LLMs and their derivatives sky-rocketed 
to also breach multi-agent systems research, myself and a few colleagues 
wrote about "argumentation-based coordination" (although applied to the 
IoT domain).

While reading your message, I thought that you could maybe find it useful:
  - https://ieeexplore.ieee.org/document/8114170 (the "vision")
  - https://link.springer.com/chapter/10.1007/978-3-031-21203-1_3 (a 
first simple application)

Just my two cents :)

Il 11/07/2025 15:41, Lorenzo Moriondo ha scritto:
> Thanks. I have just found this message.
>
> I will take a look at this. This is interesting as it is closer to the 
> idea of using existing formats (or formats closer to NL like RFC 
> documents).
> I am running some experiments and it is evident that most popular 
> formats work better for generations as there are obviously more 
> samples for those in the training dataset.
>
> On Mon, 7 Jul 2025 at 02:01, Adam Sobieski <adamsobieski@hotmail.com> 
> wrote:
>
>     Autonomous Agents on the Web (WebAgents) Community Group,
>
>     Hello. With respect to human-to-agent and agent-to-agent
>     communication protocols and models of messages, in addition to
>     those models in use today (e.g., [1][2][3][4][5][6][7][8][9]), I
>     would like to express that MIME could also be of use.
>
>     I explore these topics in greater detail here:
>     https://github.com/AdamSobieski/Argumentation . There, in a
>     section titled: /A Comparison of Message Models/, I indicate some
>     of the benefits and advantages of using the MIME message format,
>     including its support for multipart content (e.g.,
>     "multipart/related" and "multipart/alternative"), for agentic
>     communication scenarios including instant-messaging, chat, and
>     Internet forums.
>
>     Benefits and advantages of using the MIME format would include:
>
>     1.
>         a well-defined hierarchical structure for message parts to
>         simplify processing,
>     2.
>         the capability to provide one or more message attachments,
>         e.g., documents, images, audio, video, and data,
>     3.
>         the capability to provide content alternatives, enabling
>         content negotiation, internationalization and localization,
>     4.
>         interoperability with RFC 2392 URL schemes,
>     5.
>         the capability to express and include complex data involving
>         multiple interrelated parts in multiple formats,
>     6.
>         enabling scenarios beyond agentic instant-messaging and chat,
>         e.g., agentic Internet forums,
>     7.
>         the capability to provide parallel forms of content for
>         simultaneous consumption, e.g., natural-language text and
>         structured-knowledge argumentation.
>
>
>     Thank you. I hope that these ideas are of some interest to the group.
>
>
>     Best regards,
>     Adam Sobieski
>     http://www.phoster.com
>
>     [1] https://docs.anthropic.com/en/api/messages
>     [2]
>     https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/responses?tabs=python-secure
>
>     [3] https://ai.google.dev/gemini-api/docs/text-generation
>     [4]
>     https://huggingface.co/docs/text-generation-inference/main/en/messages_api
>
>     [5] https://ollama.readthedocs.io/en/api/#generate-a-chat-completion
>
>     [6]
>     https://microsoft.github.io/autogen/stable/user-guide/agentchat-user-guide/tutorial/messages.html
>
>     [7] https://python.langchain.com/api_reference/core/messages.html
>
>     [8]
>     https://a2aproject.github.io/A2A/latest/specification/#64-message-object
>
>     [9]
>     https://agentcommunicationprotocol.dev/core-concepts/message-structure
>
>
>
> -- 
> ¤ acM ¤
> Lorenzo
> Moriondo
> @lorenzogotuned
> https://www.linkedin.com/in/lorenzomoriondo
> https://github.com/Mec-iS

-- 
------------------------------
Stefano Mariani, PhD
Tenure-track researcher (RTDb) @ University of Modena and Reggio Emilia
   >https://smarianimore.github.io
------------------------------