Will be following this closely! I've been working on something similar called `dagger agent`. The idea is to build on Dagger's sandboxing and (ab)use its existing infrastructure of "reproducible function recipes flowing through OpenTelemetry" - by replaying those recipes from their trace in your local engine or from Dagger Cloud.
The durability has been a lifesaver when I'm dogfooding and the TUI crashes or the session spawns 5 ambitious subagents and OOMs my laptop. Quite a few times I've migrated a session to Cloud's engines and kept going from there. Effectively the entire state of multiple sandboxes is put through storage as humble OTLP data and revitalized on whatever hardware you bring it back on, like thawing Walt Disney (and giving him a stimpack I guess).
My end goal is to have long-running sessions that hold curated context via tool state so it's durable to compaction, and to be able to keep a multi-agent month-long workstream going from my phone by messaging its leader through Cloud. It's been a passion project for over a year and it's taken a lot of world-building along the way (new Dagger core APIs, and Dagger 1.0 being the top priority independent from all this agent stuff). I'm finally at the point where I'm using it productively but I don't have a quick-start yet (edit: here's a quick and dirty one - <a href="https://gist.github.com/vito/bed465ed43a09a433b85a360311c7d35" rel="nofollow">https://gist.github.com/vito/bed465ed43a09a433b85a360311c7d3...)
Find me in Dagger.io's Discord if you're interested :)
I'm still using Concourse every day... for fun (and GPU mutex) — I guess I'll switch to Dagger? If it's light enough to run agents, I'm very curious.
Happy to hear from a Concourser! The transition to Dagger isn't fully defined yet, but we're getting there. It's taken a lot of time to find the right primitives and build a solid engine for maximal caching, but we're excited about what we're launching soon. It covers the middle parts of a Concourse pipeline really well (build + test, CI checks) but we still need to solve the beginning/end (monitoring inputs, shipping/releasing).
vito · · focus · HN ↗
The durability has been a lifesaver when I'm dogfooding and the TUI crashes or the session spawns 5 ambitious subagents and OOMs my laptop. Quite a few times I've migrated a session to Cloud's engines and kept going from there. Effectively the entire state of multiple sandboxes is put through storage as humble OTLP data and revitalized on whatever hardware you bring it back on, like thawing Walt Disney (and giving him a stimpack I guess).
My end goal is to have long-running sessions that hold curated context via tool state so it's durable to compaction, and to be able to keep a multi-agent month-long workstream going from my phone by messaging its leader through Cloud. It's been a passion project for over a year and it's taken a lot of world-building along the way (new Dagger core APIs, and Dagger 1.0 being the top priority independent from all this agent stuff). I'm finally at the point where I'm using it productively but I don't have a quick-start yet (edit: here's a quick and dirty one - <a href="https://gist.github.com/vito/bed465ed43a09a433b85a360311c7d35" rel="nofollow">https://gist.github.com/vito/bed465ed43a09a433b85a360311c7d3...)
Find me in Dagger.io's Discord if you're interested :)
gavmor · · focus · HN ↗
I'm still using Concourse every day... for fun (and GPU mutex) — I guess I'll switch to Dagger? If it's light enough to run agents, I'm very curious.
vito · · focus · HN ↗