Roman Leventov

> Version i am on the fence. I would hate to upgrade marimo and lose all my cache. I can see us making breaking changes to this, so internally we...

> Essentially, you want a '[hot tub time machine](https://github.com/kimono-koans/httm)' on a per-cell basis, where each run is snapshotted with exact code + output, designed to scale and be easily shared/cleaned...

@dmadisetti @mscolnick it seems to me we don't have substantial disagreements, do we? I will move forward with making a draft PR for this framework with some format of content...

> It seems like the cached execution should come first with the current naive structure. > [...] > Implementation with the current setup is still a good start. I don't...

> For instance, you could potentially trim parts of the notebook graph that are intermediate compute steps allowing for complex notebooks to be exported, verified, interactive, and WASM embedded with...

> lookup meta should contain hash, hash_mechanism, blob_location, created_at, last_accessed, total_accesses. I think additional metadata can be stored in the blob itself for now (as is the case rn). last_accessed,...

> It doesn't need to know. I've successfully leveraged the same cache across 2 notebooks using an imported cell. I don't think there's anything wrong with that. I've also used...

Interesting, I realised that: - "Module code _only_" (_without_ the code of the dependencies) is _always_ a part of the module hash, regardless of the "cache type": https://github.com/marimo-team/marimo/blob/d75c3505564c7a49507415c4fb0ad6154eedaab6/marimo/_save/hash.py#L436. - "Module...

> Your proposal is added logic is on when and how to invalidate, such that heuristics like LRU and LFU can be replaced with retention policies that confidently flush or...

> But- more on hot tub time machine. In a managed environment, I think using [jujutsu](https://github.com/jj-vcs/jj) (git wrapper that covers these needs) on cell execution and a mirrored closure of...