8000 GitHub - velo-sh/velo: The Instant Runtime that Python deserves. · GitHub
[go: up one dir, main page]

Skip to content
velo-shPublic

Latest commit

 

History

3,618 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
< F438 /td>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Life is short. Use Python.

Python, Unchained.

Velo Logo

The Instant Runtime that Python deserves.

🚧 Heavy Work In Progress. Not ready for production. Expect breaking changes.


🚀 Why Velo?

The "Python Tax" The Velo Fix
Cold starts take 500ms+ 60x faster with Zygote pre-warming
Model loading is slow Fork pre-loaded models in <20ms
Tests block your flow 40x faster pytest runs
Deploys are complex Single command: velo serve main:app

⚡ 60x Faster. See It To Believe It.

╔════════════════════════════════════════════════════════════╗
║              FastAPI Hello World (Startup)                 ║
╠════════════════════════════════════════════════════════════╣
║  CPython           ██████████████████████████░░░░░  514ms  ║
║  Velo (Instant)    ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░    8.6ms ⚡║
╚════════════════════════════════════════════════════════════╝
                      🚀 59.7x faster than CPython

🎯 Who Is Velo For?

For AI & ML Engineers

"My Lambda function takes 10 seconds to cold start because of PyTorch."

Velo pre-loads your model into a Zygote. Every request forks from that warm state in <20ms. Memory is shared via Copy-on-Write.

→ Read the AI Inference Guide


For Python Developers

"I spend more time waiting for tests than writing code."

Velo's pytest --velo plugin accelerates your test suite by 40x. Combined with Vibe Mode, you get instant feedback on every save.

→ Read the Developer Workflow Guide


🛠️ Quick Start

git clone https://github.com/velo-sh/velo.git && cd velo
cargo build --release
./target/release/velo run examples/hello.py

Commands

# Run any Python script, 60x faster
velo run script.py

# Serve a web app (FastAPI, Django, Flask)
velo serve main:app --workers 4

# Accelerate your tests
pytest --velo

# Vibe Mode: instant hot-reload
velo run --vibe app.py

📊 Compatibility

  • Python: 3.11, 3.12, 3.13+ (single binary)
  • Frameworks: FastAPI, Django, Flask, Celery
  • Packages: 100% of PyPI Top 100 verified ✅

📚 Documentation

Guide Description
AI Inference Deploy AI models with <20ms cold start
Developer Workflow 40x faster tests, instant hot-reload
Compatibility Charter Our commitment to zero breaking changes

🛡️ Enterprise Ready

  • Zero Python Patch: We never modify CPython. Full compatibility guaranteed.
  • SLO-Backed Performance: Fork latency P99 < 100ms, contractually.
  • Security Hardened: Landlock isolation, process sandboxing, supply-chain audited.

🤝 Special Thanks & Integrations

Velo stands on the shoulders of giants. We take pride in our deep integrations with the best tools in the Python ecosystem:

  • uv: Velo's first-class dependency engine. We leverage uv for lightning-fast environment synchronization and reliable lockfile detection.
  • Granian: Our high-performance HTTP server backend. Velo integrates Granian to provide the raw power behind velo serve.

License

Apache-2.0

Releases

Packages

Used by

Contributors

Languages

0