⚡ Standard Vector Embeddings

Embeddings API (`/v1/embeddings`)

Convert text into dense vector representations for semantic search, retrieval-augmented generation (RAG), and clustering. 100% OpenAI SDK compatible with multi-provider failover and Zero Data Retention.

1. Quick Example (OpenAI SDK)

Point your standard OpenAI client to https://api.nicrron.ai/v1. No custom SDK required.

TypeScript / Node.js

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.nicrron.ai/v1",
  apiKey: process.env.NICRRON_API_KEY,
});

const response = await client.embeddings.create({
  model: "openai/text-embedding-3-small",
  input: ["The quick brown fox jumps over the lazy dog", "Semantic search with Nicrron"],
  dimensions: 512, // Optional: override the model's default width
});

console.log(response.data[0].embedding); // Float array
console.log(response.usage); // { prompt_tokens: 18, total_tokens: 18 }

2. Supported Embedding Models

Model Identifier	Provider	Default Dimensions	Max Dimensions	Pricing (per 1M tokens)
openai/text-embedding-3-small	OpenAI	1536	1536	$0.02
openai/text-embedding-3-large	OpenAI	3072	3072	$0.13
mistral/mistral-embed	Mistral	1024	1024	$0.10
mistral/codestral-embed	Mistral	1536	1536	$0.15
voyage/voyage-4-lite	Voyage	1024	2048	$0.02
voyage/voyage-4	Voyage	1024	2048	$0.06
voyage/voyage-4-large	Voyage	1024	2048	$0.12
voyage/voyage-code-4	Voyage	1024	2048	$0.12
voyage/voyage-finance-2	Voyage	1024	1024	$0.12
voyage/voyage-law-2	Voyage	1024	1024	$0.12

Size your vector index against the default column unless you pass dimensions explicitly. The two differ only on the Voyage -4 family, which returns 1024 by default but supports up to 2048; the -2 domain models reject anything above 1024.

3. Multi-Model Embedding Fallbacks

Specify a fallback array with models in your request body. If the primary provider encounters a rate-limit (429) or outage (5xx), Nicrron automatically cascades to the secondary provider.

cURL Request

curl https://api.nicrron.ai/v1/embeddings \
  -H "Authorization: Bearer $NICRRON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "models": ["openai/text-embedding-3-small", "mistral/mistral-embed"],
    "input": "High-throughput retrieval augmented generation"
  }'

4. Zero Data Retention

🛡️ Strict In-Memory Streaming

Your document embeddings and raw input chunks are never saved to databases or disk. Only token usage and latency telemetry are stored for cost settlement. Read our Data Privacy Guarantee.