1. Confirm model availability
Choose an embedding model visible in your TKEN account and keep its vector dimension stable for each index.
Embeddings API
Build embeddings through the TKEN gateway
Point OpenAI-compatible SDK clients at https://www.tken.shop/v1, run a small vector smoke test, then use the same server-side pattern for search, clustering and retrieval jobs.
Node.js
embeddings smoke test
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.TKEN_API_KEY,
baseURL: "https://www.tken.shop/v1"
});
const result = await client.embeddings.create({
model: process.env.TKEN_EMBEDDING_MODEL,
input: [
"TKEN is an OpenAI-compatible API gateway.",
"Vector search works best with consistent dimensions."
]
});
console.log(result.data.map(item => item.embedding.length));Use one gateway base URL
Keep vector dimensions consistent
Store keys outside client-side code
Python backend job
Treat embedding generation as a backend workflow. Capture only non-sensitive request metadata in logs, and keep raw documents out of shared debugging output.
Python
server-side key
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["TKEN_API_KEY"],
base_url="https://www.tken.shop/v1",
)
response = client.embeddings.create(
model=os.environ["TKEN_EMBEDDING_MODEL"],
input="Create a vector for retrieval testing.",
)
vector = response.data[0].embedding
print(len(vector))Embedding rollout checklist
2. Test batch size
Start with a short input batch, then increase throughput only after latency, error rate and billing behavior are understood.
3. Protect document data
Redact customer identifiers, raw private documents and API keys before storing logs or sharing examples.
Disclosure: TKEN is an independent third-party API gateway, not an
official endpoint for OpenAI or other model providers.
Run an embeddings smoke test
Verify one request, confirm vector shape, then move the same https://www.tken.shop/v1 base URL into your retrieval workflow.