ghostnexus 0.1.0

SDK Python officiel pour GhostNexus — GPU cloud décentralisé, sécurisé & RGPD

GhostNexus Python SDK

Run Python scripts on GDPR-compliant GPU cloud — billed by the second.

GhostNexus is a decentralized GPU compute network. Submit your Python script, our distributed GPUs run it — securely, in the EU, billed by the second.

• RTX 4090 from $0.50/hr · A100 80GB from $2.20/hr · H100 from $3.50/hr
• GDPR-compliant · EU-hosted · AI Act 2026 ready
• Docker sandbox isolation · Per-second billing · No minimum deposit

---

Install

pip install ghostnexus

Quick Start

import ghostnexus

client = ghostnexus.Client(api_key="gn_live_...")

# Submit a script
job = client.run("train.py")

# Wait for result
result = job.wait()
print(result.output)
print(f"Cost: ${result.cost_credits:.4f}")

Get your API key at ghostnexus.net/dashboard — $15 free credits with code WELCOME15.

Examples

Fine-tune a model

import ghostnexus

client = ghostnexus.Client()  # reads GHOSTNEXUS_API_KEY env var

job = client.run("finetune_mistral.py", task_name="Mistral-7B LoRA fine-tune")
result = job.wait(timeout=7200)  # 2h max

print(result.output)
# Duration: 1h 23m
# Cost: $0.69

Inline script

job = client.run(
    "import torch; print(f'CUDA: {torch.cuda.is_available()}')",
    inline=True,
    task_name="cuda-check"
)
result = job.wait(timeout=30)
print(result.output)  # CUDA: True

Check balance

balance = client.balance()
print(f"Available credits: ${balance:.2f}")

Job history

jobs = client.history(limit=10)
for job in jobs:
    print(f"{job.task_name}: {job.status} — ${job.cost_credits:.4f}")

Error Handling

from ghostnexus.exceptions import InsufficientCreditsError, JobFailedError

try:
    job = client.run("train.py")
    result = job.wait()
except InsufficientCreditsError:
    print("Top up at https://ghostnexus.net/dashboard")
except JobFailedError as e:
    print(f"Job failed. Logs:\n{e.logs}")

Environment Variable

export GHOSTNEXUS_API_KEY="gn_live_..."

# No need to pass api_key= explicitly
client = ghostnexus.Client()

Pricing

GPU	VRAM	Price	Best for
RTX 4070	12 GB	$0.25/hr	Inference, small models
RTX 4090	24 GB	$0.50/hr	Fine-tuning 7B–13B
RTX A6000	48 GB	$0.70/hr	Fine-tuning 30B–70B
A100 80GB	80 GB	$2.20/hr	Large-scale training
H100 80GB	80 GB	$3.50/hr	LLM pre-training

Billed per second — no hourly rounding, no minimum.

Links

• Dashboard — manage jobs and credits
• Pricing — full GPU catalog
• Blog — tutorials and guides
• Contact

License

MIT