Show HN: Durable Actor Session Protocol

(dasp-protocol.github.io)

16 points | by mikehostetler 14 hours ago ago

7 comments

  • mathieu_aithos 7 hours ago

    Do you envision an A2A adapter for Jido, with DASP handling durable execution underneath? I’d be curious how an uncertain outcome would map to an A2A task state.

    • mikehostetler 4 hours ago

      Yes. A2A and ACP adapters would sit above DASP.

      The `uncertain` mapping isn't clean. DASP treats it as terminal: the host can't prove whether effects occurred, so it must not retry. A2A's `TASK_STATE_UNSPECIFIED` is indeterminate, but not terminal. The adapter must preserve the DASP outcome in metadata and map it conservatively, likely to `FAILED`.

  • nkko 14 hours ago

    I’ve been playing a lot with durable agents, can’t say that I know what the final thing should look like but this could be a part of it. For now I have celld in front of sandbox, if I get this right: agent -> dasp -> celld -> sandbox

  • yodon 12 hours ago

    Is this purely polling, or am I missing how to subscribe to events?

    • mikehostetler 11 hours ago

      The protocol is transport agnostic - but I'm primarily building it to work over Web Sockets

      • yodon 8 hours ago

        My question wasn't about transport - it's more do I have to repeatedly poll for events, or can I make a request that effectively subscribes me to a stream of events.

  • HalcyonicStorm 14 hours ago

    Really great stuff! Cant wait to try it out in my harness!