Vectros SDK for TypeScript / Node.js

The official TypeScript client for the Vectros API — hybrid search, document ingestion, structured records, and grounded inference for your application.

Installation

npm install @vectros-ai/sdk

Requires Node.js 18+. Ships with TypeScript types.

Quick start

import { VectrosClient } from "@vectros-ai/sdk";

const client = new VectrosClient({
  environment: "https://api.vectros.ai",
  token: process.env.VECTROS_API_KEY!, // sk_live_... or sk_test_...
});

// Hybrid (keyword + semantic) search over your indexed content
const results = await client.search.content({
  query: "patient intake form diabetes",
});

// Ingest a document — extracted, chunked, and indexed for search + RAG
const doc = await client.documents.ingestDocument({
  body: {
    title: "Patient Intake Form — Jane Doe",
  },
});

// Write a structured record against one of your schemas
const record = await client.records.createRecord({
  body: {
    typeName: "intake_form",
    schemaId: "6ba7b810-9dad-11d1-80b4-00c04fd430c8",
    payload: { first_name: "Jane", email: "[email protected]" },
  },
});

Authentication

The SDK sends whatever credential you pass in the Authorization: Bearer <token> header. Two credential types are accepted:

| Type | Prefix | Lifetime | Use from | |------|--------|----------|----------| | API key | sk_live_* / sk_test_* | Long-lived | Server only — full tenant access | | Scoped token | st_* | Short-lived | Server or browser — narrowed scope, auto-expiring |

Put an API key only on your server. For browser code, mint a short-lived scoped token on your backend and pass it (or a token-supplier function) to the client — never ship an API key to the browser. The token option accepts either a string or an async supplier, so refresh logic lives in one place:

const client = new VectrosClient({
  environment: "https://api.vectros.ai",
  token: async () => getFreshScopedToken(), // your backend mint call
});

See the authentication guide for the full browser token-minting pattern.

What you can do

• Hybrid search & RAG — client.search, client.inference — vector + keyword search and grounded document Q&A over your indexed corpus.
• Documents & folders — client.documents, client.folders — ingest, organize, retrieve, and look documents up by field.
• Structured records — client.records — create, read, update (full and partial), delete, and look records up by indexed field.
• Schemas — client.schemas — define and evolve record/document schemas.
• Identity & access — client.identity, client.auth — manage clients, organizations, and users; mint and revoke scoped credentials.

Rate limits

Requests are rate limited per account on a fixed one-minute window — writes, searches, and inference count against it; reads do not. When you exceed the limit the API returns HTTP 429 with a Retry-After header (seconds until the window resets) plus X-RateLimit-Limit and X-RateLimit-Remaining. Honor Retry-After (or back off exponentially with jitter), and pace bulk work so your steady rate stays under your plan's per-minute budget. See the rate limits guide for the per-plan limits.

Documentation

• Guides & reference: docs.vectros.ai
• Product: vectros.ai

Security & trust

Vectros enforces per-customer, fail-closed isolation and least-privilege scoped keys, with a tamper-evident audit and version history. Customer-facing surfaces are hardened through extensive adversarial security review. For the full trust posture, drawn plainly with its boundaries, see the compliance and trust guide.

License

Apache License 2.0.