Docs menu
Documentation

Practical Hostnot GPU guides for GPU cloud compute, SSH access, APIs, serverless inference, billing, security, and troubleshooting.

Quickstart

Create an account, fund a wallet, add an SSH key, and launch compatible capacity.

API Reference

Authentication, production base URLs, pagination, errors, idempotency, and customer API examples.

API Examples

Equivalent cURL, PowerShell, Python, and JavaScript request patterns using placeholder credentials.

GPU Cloud

Understand Hostnot GPU cloud instances, live capacity, GPU selection, wallet authorization, SSH access, and lifecycle operations.

Public GPU Availability API

Poll current Hostnot GPU availability and customer-facing hourly prices without an API key.

SSH Connections and Fingerprints

Connect safely to a Hostnot GPU instance from Linux, macOS, or Windows.

Serverless API

Model discovery, inference requests, usage, pricing, errors, and safe client patterns.

Models

Browse current Hostnot GPU serverless models, modalities, limits, request patterns, and how to verify live model capabilities.

Billing and Wallet

Deposits, reservations, GPU usage, serverless usage, and transactions.

Developer Security

Secure Hostnot GPU API and serverless integrations with scoped keys, safe logs, rate limits, validation, and credential hygiene.

API Keys

Create scoped keys, store them safely, and revoke them after suspected exposure.

Deploying and Managing GPU Instances

Choose a compatible offering, authorize deployment, connect with SSH, and terminate when work is complete.

Organizations and Workspaces

Create a shared workspace, invite members, manage roles, and understand workspace boundaries.

Support and Human Handoff

Use AI Support for guidance and request a human handoff when an operator needs to take ownership.

Account Security and Two-Factor Authentication

Protect sessions, passwords, two-factor authentication, connected accounts, and notification preferences.

Usage, Transactions, and Budgets

Read GPU and Serverless usage, understand settlement, and set account budget safeguards.

Serverless Quickstart

Make a first authenticated Serverless request using a live model slug.

Serverless Errors and Rate Limits

Interpret structured Serverless errors and retry without duplicating work.

Serverless Playground

Use the authenticated Playground to test available chat and image capabilities.

GPU Marketplace

Compare live GPU offerings, topology, regions, environments, and customer rates.

GPU Cloud Troubleshooting

Diagnose Hostnot GPU SSH, GPU driver, Docker, CUDA, instance, serverless, and billing issues with safe evidence-first steps.

Webhooks

Configure Hostnot GPU webhooks, verify signed deliveries, handle retries, and inspect customer-safe event payloads.

Wallet Deposits

Understand Hostnot GPU wallet deposits, payment confirmation, reservations, credits, and safe billing verification steps.

Billing History

Read Hostnot GPU billing history, wallet-period summaries, usage categories, and account-scoped transaction records.

PowerShell API Examples

Use Invoke-RestMethod from PowerShell without exposing API keys in command history.

hostnotgpu / docs

Serverless Quickstart

Make a first authenticated Serverless request using a live model slug.

By Hostnot Network Solutions FZE LLC

Prepare the account and request boundary

Use an active, email-verified account with available wallet funds. Create a backend API key with pricing:read for discovery and inference:write for generation. Check expiry and IP restrictions. Keep the token in your application's secret store; never embed it in public code or a support message. See API-key setup.

Discover a public model ID

curl --fail-with-body --max-time 30 https://serverless.hostnotgpu.ae/v1/models \
  -H "Authorization: Bearer $HOSTNOTGPU_API_KEY"

This GET discovers models; it does not run generation. Read the data array and select a text model's public id. Record modality, context length and pricing. Empty results and failed discovery are different states. The response has no streaming/tool flags and does not reserve capacity.

Prepare one bounded chat operation

The following JSON is a body template for POST /v1/inference/chat on the same host. Replace the model placeholder. Sending it performs billable generation. Add Content-Type application/json, bearer authentication and a non-empty Idempotency-Key of at most 255 characters. Store that key for this logical operation before dispatch.

{"model":"MODEL_ID_FROM_DISCOVERY","messages":[{"role":"user","content":"Explain one GPU use in one sentence."}],"max_tokens":64,"stream":false}

Chat permits 1–100 messages using system, user or assistant roles and string content, at most 50,000 characters per message. The output bound is 1–8192 tokens; the model's actual context can be more restrictive. Streaming and unknown fields are rejected. Input tokens plus generated output must fit the selected model.

Verify success and interpret failure

Check status and Content-Type before JSON parsing. Successful chat uses inference.result with output.type: text and output.text, plus usage, billing and latency. Do not read an assumed choices array. Render generated text as untrusted input and preserve money as decimal strings.

A 400 requires corrected input; 401/403 requires credential or account review; 402 requires available-funds review; 404/409 requires model or pricing review. Reduce admission on 429 and honor Retry-After when supplied. A timeout or 5xx may leave delivery uncertain: retain the original key and request ID and reconcile before retrying. Aborting your client is not confirmed remote cancellation.

Complete the first evaluation

Compare output quality, end-to-end latency and returned billing against your acceptance criteria. Inspect workspace records for unresolved reservations; GPU-only /usage is not an inference export. Test failure paths locally before increasing paid concurrency. Continue with the exact API contract, error decision tree and wallet terminology.

Published · Updated