Getting Started

What is VectorAxis?

VectorAxis is an AI gateway and control plane that sits between your application and every LLM provider. It gives your team one endpoint, unified credentials, intelligent routing, caching, guardrails, and deep observability — without changing how you write AI code. Point your OpenAI SDK at it and every feature is one header away.

Fully OpenAI SDK compatible. Change one line — your base URL — and caching, fallbacks, retries, guardrails, and cost tracking are all available immediately. No new SDK, no rewrite.
The problem it solves

As soon as a team ships more than a toy AI feature, the same pains show up: provider API keys scattered across services, no idea what any of it costs, an outage on one provider taking the product down, no way to enforce a budget, and no record of what was sent or returned. VectorAxis puts a single, governed layer in front of all of it — so those become configuration, not engineering projects.

  • For engineers: one base URL, the OpenAI wire format everywhere, and per-request control over routing, caching, retries, and guardrails via headers.
  • For the business: real-time cost visibility, hard spend caps, provider redundancy, and an audit trail — the controls needed to run AI in production safely.
How it works

Your app sends a normal OpenAI-format request to VectorAxis with a virtual key in place of a provider secret. VectorAxis resolves the real credential, applies whatever policies you’ve configured — routing, cache lookup, guardrails, retry/fallback — calls the provider (translating to its wire format when needed), and returns an OpenAI-format response. Along the way it records the request’s tokens, cost, and latency. Nothing proprietary goes in the request body; every capability is a header.

What’s in the platform
🔑
Virtual Keys
Encrypted credential proxying with rate limits, credit budgets, allowlists, and lifecycle management.
🔀
Routing Configs
Load-balance, fail over, and conditionally route across providers with a visual graph builder.
Simple & Semantic Cache
Exact-match and pgvector semantic caching — save cost and latency on repeated or similar queries.
🛡️
Guardrails
PII detection, regex matching, JSON-schema validation, and LLM-judge checks on inputs and outputs.
📝
Prompt Management
A versioned prompt library with draft/publish workflow, labels, test suites, and analytics.
🎛️
Multimodal
Audio, images, files, batches, and the Responses API — the same gateway for every modality.
📊
Observability
Per-request logs, cost tracking, and latency percentiles.
👥
Access Control
Organizations, workspaces, roles, and va_ platform keys for team-wide governance.
Ready to send your first request? Head to the Quick Start — you’ll be routing through VectorAxis in four steps. Prefer to see the surface first? Browse Providers and the Endpoints reference.