What is this?
objgit is an experiment: a Git server that keeps repositories in object storage (Tigris) instead of on a local disk. The author calls it a proof of concept that you probably don't need yet.
objgit is an experiment: a Git server that keeps repositories in object storage (Tigris) instead of on a local disk. The author calls it a proof of concept that you probably don't need yet.Is this a new version control system, or a server that hosts normal Git repos?It's a server, not a new VCS — repos stay ordinary Git, stored in bare repo format inside the bucket and served over ssh, http, and the git protocol.
It's a server, not a new VCS — repos stay ordinary Git, stored in bare repo format inside the bucket and served over ssh, http, and the git protocol.How do I get it running?Get a Tigris account, bucket, and access keypair, put the config in a .env file (envvars take precedence over .env, which beats flags), then run the program and push a repo.
Get a Tigris account, bucket, and access keypair, put the config in a .env file (envvars take precedence over .env, which beats flags), then run the program and push a repo.Can I point it at any S3-compatible bucket, or does it need Tigris specifically?Tigris specifically: the core flow relies on Tigris' RenameObject call, which is not part of the standard S3 API. You also need a system that runs Go, TLS certificate authorities on disk, and an IPv4/IPv6 network stack.
Tigris specifically: the core flow relies on Tigris' RenameObject call, which is not part of the standard S3 API. You also need a system that runs Go, TLS certificate authorities on disk, and an IPv4/IPv6 network stack.So what's missing before I could actually rely on it?Quite a lot: there is no authentication whatsoever, reads are fully buffered in memory (theoretical 2Gi file size limit), object storage is more latent than a local disk, and the author says it is very new and barely tested.
Quite a lot: there is no authentication whatsoever, reads are fully buffered in memory (theoretical 2Gi file size limit), object storage is more latent than a local disk, and the author says it is very new and barely tested.Why did the author build this instead of just running a normal Git server?The README calls it heavily vibe coded — an exploration of the design space, with the claude plans folder documenting what was being explored and why. If the experiment proves shippable, a better implementation is meant to replace the current code.
The README calls it heavily vibe coded — an exploration of the design space, with the claude plans folder documenting what was being explored and why. If the experiment proves shippable, a better implementation is meant to replace the current code.What's on the roadmap?Future goals include Tekton integration for CI pipelines, a web UI built on the Xe Design System, deployment to Kubernetes, HTTPS support with TLS termination, and a blogpost.
Future goals include Tekton integration for CI pipelines, a web UI built on the Xe Design System, deployment to Kubernetes, HTTPS support with TLS termination, and a blogpost.Does it expose any monitoring metrics?Yes, basic Prometheus metrics are included out of the box.
Yes, basic Prometheus metrics are included out of the box.Which topics is it in?experiments, prometheus, s3, version-control, storage, git, monitoring, backend, go, free, tool, server
experiments, prometheus, s3, version-control, storage, git, monitoring, backend, go, free, tool, server