How to Convert Text Into Dense Vector Representations
Last verified 16 Oct 2026
Inference provides a single control plane for managing inference workflows. It includes a Model Catalog where you can view available foundation models, including both DigitalOcean-hosted and third-party commercial models, compare model capabilities and pricing, use routing to match inference requests to the best-fit model, and run inference using serverless or dedicated deployments.
Use the Embeddings API to convert text into dense vector representations for use in semantic search, retrieval-augmented generation (RAG), clustering, classification, and similarity matching. Embedding models include Qwen3 Embedding 0.6B, BGE-M3, and E5-Large. They offer a range of options across multilingual support, token window sizes, and dimensionality to match your use case.
For embedding requests, send a POST request to the https://inference.do-ai.run base URL using your existing Model Access Key. Embeddings are returned synchronously as float arrays that you can store in a vector database or use directly in your application pipeline.
The following cURL and Python PyDo examples generates a vector embedding using the Qwen3 Embedding 0.6B model:
Create a model access key and save it for use with the API.
Using PyDo, the official DigitalOcean API client for Python:
import os
from pydo import Client
client = Client(token=os.environ.get("DIGITALOCEAN_TOKEN"))
resp = client.embeddings.create(
model="qwen3-embedding-0.6b",
input=["hello world", "goodbye world"],
encoding_format="float",
user="user-1234",
)
for item in resp.data:
print(item.index, item.embedding[:8])Using Godo, the official DigitalOcean API client for Go:
package main
import (
"context"
"encoding/json"
"fmt"
"os"
"github.com/digitalocean/godo"
)
func main() {
client := godo.NewFromToken(os.Getenv("DIGITALOCEAN_TOKEN"))
ctx := context.Background()
resp, _, err := client.Embeddings.New(ctx, &godo.EmbeddingNewParams{
Model: "qwen3-embedding-0.6b",
Input: []string{"hello world", "goodbye world"},
EncodingFormat: "float",
User: "user-1234",
})
if err != nil {
panic(err)
}
for _, item := range resp.Data {
var vec []float32
if err := json.Unmarshal(item.Embedding, &vec); err != nil {
panic(err)
}
head := vec
if len(head) > 8 {
head = head[:8]
}
fmt.Println(item.Index, head)
}
}Using dots, the official DigitalOcean API client for JavaScript:
import { InferenceClient } from "@digitalocean/dots";
const client = new InferenceClient({
apiKey: process.env.DIGITALOCEAN_TOKEN,
});
const resp = await client.embeddings.create({
model: "qwen3-embedding-0.6b",
input: ["hello world", "goodbye world"],
encoding_format: "float",
user: "user-1234",
});
for (const item of resp.data) {
console.log(item.index, item.embedding.slice(0, 8));
}Send a POST request to https://inference.do-ai.run/v1/embeddings.
Using cURL:
curl -X POST \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $DIGITALOCEAN_TOKEN" \
-d '{"model":"qwen3-embedding-0.6b","input":["hello world","goodbye world"],"encoding_format":"float","user":"user-1234"}' \
"https://inference.do-ai.run/v1/embeddings"You can also use the Go SDK (Godo):
package main
import (
"context"
"encoding/json"
"fmt"
"os"
"github.com/digitalocean/godo"
)
func main() {
client := godo.NewFromToken(os.Getenv("MODEL_ACCESS_KEY"))
ctx := context.Background()
resp, _, err := client.Embeddings.New(ctx, &godo.EmbeddingNewParams{
Model: "qwen3-embedding-0.6b",
Input: []string{"hello world", "goodbye world"},
})
if err != nil {
panic(err)
}
for _, item := range resp.Data {
var vec []float32
if err := json.Unmarshal(item.Embedding, &vec); err != nil {
panic(err)
}
fmt.Println(item.Index, len(vec))
}
}