Embeddings and rerank - pre.dev

POST https://api.pre.dev/v1/embeddings takes the OpenAI-compatible embeddings request and returns the embeddings response. GET /v1/embeddings/models lists embedding models with credit prices; it is free and not rate-limited.

MODEL=$(curl -s https://api.pre.dev/v1/embeddings/models \
  -H "Authorization: Bearer $PREDEV_API_KEY" | jq -r '.data[0].id')

curl --fail-with-body https://api.pre.dev/v1/embeddings \
  -H "Authorization: Bearer $PREDEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"model\": \"$MODEL\", \"input\": [\"pre.dev builds software from a spec\", \"credits pay for model calls\"]}"

import OpenAI from 'openai';

const client = new OpenAI({ baseURL: 'https://api.pre.dev/v1', apiKey: process.env.PREDEV_API_KEY! });

const { data: models } = await client.models.list(); // /v1/models also lists embedding models
const embeddingModel = models.find((m: any) => m.architecture?.output_modalities?.includes('embeddings'))?.id;

const result = await client.embeddings.create({
  model: embeddingModel!,
  input: ['pre.dev builds software from a spec', 'credits pay for model calls'],
});
console.log(result.data[0].embedding.length, result.usage);

import os
import requests
from openai import OpenAI

client = OpenAI(base_url="https://api.pre.dev/v1", api_key=os.environ["PREDEV_API_KEY"])

catalog = requests.get(
    "https://api.pre.dev/v1/embeddings/models",
    headers={"Authorization": f"Bearer {os.environ['PREDEV_API_KEY']}"},
).json()["data"]

result = client.embeddings.create(
    model=catalog[0]["id"],
    input=["pre.dev builds software from a spec", "credits pay for model calls"],
)
print(len(result.data[0].embedding), result.usage)

The response is the standard embeddings response: data[].embedding, model, and usage. Credits are computed from usage.cost and reported in x-predev-credits-charged.

Rerank

POST /v1/rerank ranks documents against a query. Send a model, a query, and documents; the response returns the documents ordered by relevance with a score for each.

curl --fail-with-body https://api.pre.dev/v1/rerank \
  -H "Authorization: Bearer $PREDEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$RERANK_MODEL\",
    \"query\": \"how are gateway calls billed\",
    \"documents\": [
      \"Credits are computed from the model's metered cost per call.\",
      \"Browser tasks have a 0.1-credit floor.\",
      \"Catalog reads are free.\"
    ]
  }"

Set RERANK_MODEL to a rerank-capable id from the catalog. Rerank calls are billed like any other inference call.