Skip to content

Products

GPU CloudNVIDIA H200 by the hour.
Dedicated ClustersPreviewPrivate, interconnected GPU clusters for sustained workloads.
SandboxesPreviewIsolated, short-lived environments for developers and AI agents.
Serverless GPUPreviewGPU containers that scale with requests, down to zero.
Serverless InferenceOpen models behind one OpenAI-compatible endpoint.
Dedicated EndpointsReserved serving capacity for one model, billed per GPU-hour.
PlaygroundSend a prompt to any model your workspace can reach.
Model DeploymentsPreviewImmutable deployments with instant promote and rollback.
Fine-tuningLoRA fine-tunes on the open models in our catalog.
Model TrainingPreviewFoundation-model training runs, managed end to end.
EvaluationsPreviewBenchmark suites and model comparisons you can rerun.
DatasetsManaged and synthetic training data, versioned and ready to use.
Data ManagementPreviewDatasets, connections and storage, with lineage across the platform.
PipelinesPreviewVersioned DAGs that automate data, training and deployment.
Data RoomsPreviewSecure document sharing with guests, NDAs and an audit trail.
NokuvaPreviewThe AI-native design editor.
TavocPreviewA Rust-native code platform: IDE and desktop apps.
FoltracComing soonMulti-cloud deployment with autonomous operations.
W0rkTreeComing soonVersion control for agent fleets.
Identity & AccessMembers, roles, policies, SSO and audit for every workspace.
Billing & CostUsage, invoices, credits and budgets in one place.
Logs & MonitoringRequest logs, latency and error rates across services.
Project CanvasPreviewSee and change a project's resources on one live canvas.

Pages

Home/
Platform/platform
All products/products
Pricing/pricing
GPU pricing/pricing#compute
Inference services/services/inference
Model training services/services/model-training
Protocol design/services/protocol-design
Research partnerships/services/research-partnerships
Protocols/protocols
Datasets/datasets
Documentationdocs.legionedge.ai
Better IAM on GitHubgithub.com/Better-IAM/better-iam
Better Compute on GitHubgithub.com/Better-Compute/better-compute
Better Files on GitHubgithub.com/Better-Files/better-files
W0rkTree on GitHubgithub.com/seanfilimon/worktree
Changelog/changelog
News/news
Research/research
Blog/blog
Articles/articles
Events/events
Startup Program/programs/startup
Ambassador Program/programs/ambassador
Partner Program/programs/partner
Event Program/programs/events
Company/company
Careers/careers
Partners/partners
Contact/contact
Statusstatus.legionedge.ai
Terms of Service/legal/terms
Privacy Policy/legal/privacy

All blog posts

W0rkTree v2: a version control system, not a Git wrapper

The W0rkTree rebuild is in alpha: its own object model, sync protocol and multi-tenant server, with work in progress visible to the team before it lands.

Category
Engineering
Published
Reading time
2 min
Sections
05
Source
Engineering blog
01

Overview

W0rkTree has been rebuilt, and the v2 alpha is on GitHub. It’s an independent version control system with its own object model, sync protocol, multi-tenant server, access control, license enforcement and real-time visibility of work in progress. It speaks Git only to import and export.

This post walks through what changed and what’s still ahead.

02

Three pieces with hard boundaries

The wt command line is a thin layer over a Rust SDK that other programs can embed. An optional per-user daemon, worktree-bg, watches your worktrees, takes snapshots when you go idle and uploads them as staged work. The server, written in Go, is the source of truth for shared history and serves gRPC, REST, server-sent events, metrics and the admin panel from a single port.

The boundaries are strict. The daemon never enforces access control; the server does. The server never touches a working directory or creates snapshots on its own. And objects are byte-identical on disk, on the wire and on the server, so a push or a pull is a byte copy.

03

History you can't rewrite

History is append-only. Snapshots are immutable and content is addressed by BLAKE3. There is no rebase and no force-push; branches are archived instead of deleted, and wt revert adds a new snapshot rather than removing an old one.

Merges are three-way only. Conflicts get the familiar diff3 markers plus a machine-readable file in .wt/conflicts/, so an agent can resolve them without parsing markers out of source code. Every command also takes --json and returns documented exit codes, which helps scripts and agents alike.

04

Built for teams, not just repositories

Staged snapshots make work in progress visible to the team without entering branch history: wt stage uploads it, and wt status --team shows what everyone is working on, with live updates from the server.

The server knows about people, not only files. Accounts, a role matrix from viewer to owner, public, internal and private trees, branch protection and merge requests are built in. Per-path SPDX license rules are enforced on push, secret scanning runs before every snapshot, and nested trees replace submodules with independently versioned sub-projects.

05

What's next

v2 is alpha, and the README is explicit about what’s specified but not yet built: declarative access policies in TOML, chunked storage and lazy loading for large files, QUIC with native TLS and signed snapshots, and live Git mirroring. Each has a written spec in the repository.

If you want to try it, the quickstart builds the CLI, the daemon and the server from source, and wt git import brings an existing repository across.

Share

Want the build notes as they happen?

The blog is written by the engineers shipping Nokuva, Tavoc, and Foltrac — deep dives land here as the work does, and the authors answer questions.