VaultLayer
Run AI training jobs on managed GPU capacity with checkpointing, log streaming, and provider failover.
pip install -U vaultlayer
vl init
vl run train.py
Job submitted
Training on vast_ai
Training output:
...
Training completed successfully.
What It Does
VaultLayer sits between your training script and the cloud. It:
- Checkpoints automatically — syncs model weights + optimizer state to a zero-egress R2 Vault on every save
- Detects interruptions — intercepts AWS/GCP/Azure termination signals before your job dies
- Migrates instantly — provisions a replacement node on the cheapest available provider and resumes from last checkpoint
- Tracks savings — shows real-time cost vs what you would have paid on AWS On-Demand
No changes to your PyTorch or JAX code. No YAML configs. No PhD-level infra knowledge required.
Commands
# Training
vl run train.py
vl ps
vl logs <job-id> --follow
vl stop <job-id>
# Dataset storage (no S3 required)
vl sync ./data --dataset-id my-dataset
vl run --data r2://my-dataset train.py
vl datasets
Supported Providers
| Provider | Type | Status |
|---|---|---|
| Vast.ai | Marketplace | Production-included |
| RunPod | Neocloud | Production-included |
| Lambda Labs | Neocloud | Production-included |
| AWS Spot | Hyperscaler | Production-included (single-node; GPUs A10G + V100 only; multi-node gated by VL_MULTINODE_ENABLED), as of 2026-06-28 |
| AWS On-Demand | Hyperscaler | Pending validation (not dispatchable) |
| Nebius (EU) | Neocloud | Testing (VL_TEST_USER_IDS only) |
| GCP, CoreWeave, Crusoe, Voltage Park, Hyperstack, Azure | Mixed | Pending validation |
Current provider status lives in docs/tech/providers/provider_test_matrix.md and docs/tech/providers/provider_testing_matrix.md.
Model Size Support
| Model Size | Method | Checkpoint Size | Status |
|---|---|---|---|
| 1B | QLoRA | small | Validated smoke path |
| 3B | QLoRA | small | Validated matrix path |
| 7B | QLoRA | medium | Validated matrix path |
| 72B | QLoRA | large | Routed to 96GB+ high-VRAM capacity |
| Full fine-tune / multi-GPU | varies | varies | Future work |
Tech Stack
| Layer | Technology | Cost |
|---|---|---|
| Code + Docs | GitHub (this repo) | Free |
| CI/CD | GitHub Actions | Free (2k min/mo) |
| Vault / Storage | Cloudflare R2 | Free up to 10GB |
| Agent Runtime | Railway | Free $5/mo credit |
| Webhooks | Cloudflare Workers | Free 100k req/day |
| Agent Message Queue | Upstash Redis | Free 10k cmd/day |
Repository Structure
vaultlayer/
├── README.md
├── docs/
│ ├── PRD.md # Full product requirements
│ ├── ARCHITECTURE.md # System design + agent topology
│ └── AGENTS.md # Agent specs + build order
├── dashboard/
│ └── index.html # Savings dashboard prototype
└── src/
├── cli/
│ ├── main.py
│ ├── run.py
│ ├── checkpoint_template.py
│ └── init.py
├── vaultlayer/
│ └── _resume_hook.py
├── agents/
│ ├── orchestration/
│ ├── pricing/
│ ├── watchdog/
│ │ └── signals.py
│ ├── vault/
│ ├── broker/
│ ├── finops/
│ └── namespace/
└── shared/
SLA
VaultLayer tracks job completion, checkpoint persistence, and resume behavior. Public SLA numbers are not committed during beta; see docs/tech/reliability/SLA_SLI.md for definitions.
Dataset Storage (No S3 Required)
VaultLayer's Neutral Zone (Cloudflare R2) is a first-class storage provider. Users with no AWS or cloud storage account can upload training data directly and train from it on any provider.
# Upload from your laptop / on-prem server
vl sync ./training-data --dataset-id my-dataset
# Train — data is unpacked to $VAULTLAYER_LOCAL_DATA (./data beside your script)
vl run --data r2://my-dataset train.py
# See what you're storing and the monthly cost
vl datasets
Pricing:
| Action | Cost |
|---|---|
| Upload (local → R2) | Free |
| Storage | $0.020 / GB / month ($0.0195 — 30% markup over Cloudflare R2 base rate) |
| Read (R2 → training node) | $0.00 (zero egress within Cloudflare network) |
| S3 mirror (one-time) | AWS egress charge (~$0.09/GB, first 100 GB/month free) |
Storage quotas by plan:
| Plan | Storage limit |
|---|---|
| Free | 10 GB |
| Pro | 500 GB |
| Enterprise | Unlimited |
Datasets are soft-deleted with vl datasets --delete <id> — billing stops immediately,
R2 objects are purged within 24 hours.
Getting Started
pip install -U vaultlayer
vl init
vl run train.py
License
Private — © 2026 VaultLayer
Metadata
Release files for vaultlayer 0.1.83
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| vaultlayer-0.1.83.tar.gz | 3.2 MB | With provenance | Details for vaultlayer-0.1.83.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| vaultlayer-0.1.83-py3-none-any.whl | Python 3 | none | any | Details for vaultlayer-0.1.83-py3-none-any.whl |
Total release size: 4.9 MB
Release files / vaultlayer-0.1.83.tar.gz
| Download URL | vaultlayer-0.1.83.tar.gz |
|---|---|
| Size | 3.2 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4312b489ce5be2ff70b4a97e13e8b0fdc4e1d79ea52786d38dd66f338bb0cc23
|
|
BLAKE2b-256 checksum How to use checksums |
b18787acfd4abbe1b043cad90004fac4971d3462d0f3d2c7e89030dc06991fab
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 13, 2026.
Transparency logRelease files / vaultlayer-0.1.83-py3-none-any.whl
| Download URL | vaultlayer-0.1.83-py3-none-any.whl |
|---|---|
| Size | 1.7 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
2ecdee4442776f1f9aea8275538bac92147d3c687b8a2313cee0f804f6823a85
|
|
BLAKE2b-256 checksum How to use checksums |
685a7c075d96367d789e0a567bf9c15cee36f2e8653bdac8418d69d3bffb25f7
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 13, 2026.
Transparency logSecurity for vaultlayer 0.1.83
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for vaultlayer-0.1.83.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 13, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.