Embeddings

Embeddings convert text into vectors that capture semantic meaning. Use them for search, retrieval-augmented generation (RAG), clustering, recommendations, deduplication, and similarity scoring. The Venice embeddings endpoint is OpenAI-compatible. Send one string or an array of strings to /embeddings, then store the returned vectors in your database or vector index.

Venice accepts text only. Send a string or an array of strings. Token ID arrays are rejected with HTTP 400. LangChain OpenAIEmbeddings must set check_embedding_ctx_length=False, or it tokenizes locally with tiktoken and sends integer arrays.

Basic Usage

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["VENICE_API_KEY"],
    base_url="https://api.venice.ai/api/v1",
)

response = client.embeddings.create(
    model="text-embedding-bge-m3",
    input="Privacy-first AI infrastructure for semantic search",
)

vector = response.data[0].embedding
print(len(vector), vector[:5])

import OpenAI from "openai";

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

const response = await client.embeddings.create({
  model: "text-embedding-bge-m3",
  input: "Privacy-first AI infrastructure for semantic search",
});

const vector = response.data[0].embedding;
console.log(vector.length, vector.slice(0, 5));

curl https://api.venice.ai/api/v1/embeddings \
  -H "Authorization: Bearer $VENICE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-bge-m3",
    "input": "Privacy-first AI infrastructure for semantic search",
    "encoding_format": "float"
  }'

Batch Inputs

Pass an array of strings to embed multiple texts in one request. Each item must be text, not a token ID array. Two limits apply, and both are enforced before the request reaches a model:

Limit	Value	Error when exceeded
Items per request	2,048	`400` — “Array must contain at most 2048 element(s)“
Tokens per item	8,192	`400` — token count is checked per array entry, not across the batch

Size your ingest batches against the 2,048 cap and chunk long documents below 8,192 tokens per entry. The request body is also strict, so an unrecognized field returns a 400 rather than being ignored.

{
  "model": "text-embedding-bge-m3",
  "input": [
    "Venice supports private chat completions.",
    "Embeddings help retrieve relevant documents.",
    "Vector search powers RAG applications."
  ]
}

The response preserves input order. Store each vector with the source text ID, metadata, and embedding model ID.

Common Workflow

Split source documents into chunks.
Generate embeddings for each chunk.
Store vectors and metadata in a vector database.
Embed the user’s query.
Retrieve nearby chunks.
Send the retrieved context to a chat model.

For a complete implementation, see Building a Private RAG Bot.

Model Selection

Use the Embedding Models page to compare current embedding models, dimensions, and pricing.

Use the same embedding model for indexing and querying. Mixing models can make similarity scores unreliable because vector spaces are not interchangeable.

⌘I

Docs

Getting Started

Text & Chat

Image

Video

Audio

Search & RAG

Agents & Integrations

Coding Tools

Agent Tooling

SDKs & Frameworks

Basic Usage

Batch Inputs

Common Workflow

Model Selection

​Basic Usage

​Batch Inputs

​Common Workflow

​Model Selection

​Related Resources

Basic Usage

Batch Inputs

Common Workflow

Model Selection

Related Resources