Documentation
Build on AhuraSense
The inference API serves frontier and open-source models behind one OpenAI-compatible endpoint. The app platform builds and hosts your code from a git repository. The cloud API manages compute, storage, databases, Kubernetes and the rest of the platform.
Inference API
Call any model with one key
Chat completions and an Anthropic-compatible route, streaming, tool calling, JSON mode, per-key limits and caps. Works with the OpenAI and Anthropic SDKs unchanged.
https://api.ahurasense.com/v1Read the docsApp Platform
Deploy from a git repository
Connect GitHub, GitLab or Bitbucket and every push builds and deploys. Preview deployments for branches, custom domains, and deploy hooks for your CI.
git push origin mainRead the docsCloud API
Manage the platform
Servers, GPU instances, Kubernetes, object storage, databases, domains and billing, as an OpenAPI reference with a request builder.
https://ahurasense.com/api/v1Open the referenceInference API, by topic
Get started
Overview
What the inference API is, the base URL, and a first request in under a minute.
Authentication
API keys, the headers every request carries, and what a key can be restricted to.
Models
The live catalog: ids, context windows, capabilities and per-token prices.
SDKs & frameworks
Use the OpenAI or Anthropic SDK, LangChain, or plain HTTP.
API reference
Chat completions
POST /v1/chat/completions: every parameter, the response, and the headers we add.
Messages
POST /v1/messages: the Anthropic-compatible endpoint for existing Anthropic code.
Images
POST /v1/images/generations: one image per request, returned as base64, billed per image.
Videos
POST /v1/videos and the job lifecycle: submit, poll, download, billed per second on completion.
Streaming
Server-sent events, how usage arrives on a stream, and reasoning tokens.
Errors
The error envelope and every code the gateway returns, with what to do about each.
Features
Tool calling
Function calling in the OpenAI format, including the follow-up turn.
Structured outputs
JSON mode and how to get reliably parseable answers.
Reasoning effort
How much a hosted model thinks before it answers, from none to unbounded, and what it costs.
Caching
The exact-match response cache: when it applies, how to control it, what it costs.
Guardrails
Prompt-injection detection: warn, block, or off, per request.
Presets
Named model defaults you can switch without redeploying.
Account
App Platform, by topic
Get started
Deploying
Configure