امبدینگ (Embeddings)

curl https://gisoo.pro/api/v1/embeddings \
  -H "Authorization: Bearer $GISOO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-3-small",
    "input": ["How long does shipping take?", "What is your return policy?"]
  }'

{
  "object": "list",
  "data": [
    { "object": "embedding", "index": 0, "embedding": [0.0182, -0.0041, 0.0333, "…"] },
    { "object": "embedding", "index": 1, "embedding": [-0.0107, 0.0219, 0.0045, "…"] }
  ],
  "model": "text-embedding-3-small",
  "usage": { "prompt_tokens": 14, "total_tokens": 14 }
}

# pip install openai numpy
import numpy as np
from openai import OpenAI

client = OpenAI(base_url="https://gisoo.pro/api/v1", api_key="sk-gisoo-v1-...")
MODEL = "text-embedding-3-small"

docs = [
    "ارسال سفارش به تهران یک تا دو روز کاری طول می‌کشد.",
    "کالای سالم را تا هفت روز پس از تحویل می‌توانید مرجوع کنید.",
    "پرداخت با همهٔ کارت‌های عضو شتاب انجام می‌شود.",
]

def embed(texts):
    res = client.embeddings.create(model=MODEL, input=texts)
    vectors = np.array([d.embedding for d in res.data])
    return vectors / np.linalg.norm(vectors, axis=1, keepdims=True)   # normalise once

doc_vectors = embed(docs)                        # do this once and store it

def search(question: str, k: int = 2):
    q = embed([question])[0]
    scores = doc_vectors @ q                      # cosine similarity (vectors are normalised)
    best = np.argsort(-scores)[:k]
    return [(docs[i], float(scores[i])) for i in best]

for text, score in search("چند روزه می‌رسه؟"):
    print(round(score, 3), text)

const docs = ["Shipping to Tehran takes 1–2 business days.", "Items can be returned within 7 days.", "All Shetab cards are accepted."];

const embed = async (input) =>
  (await client.embeddings.create({ model: "text-embedding-3-small", input })).data.map((d) => d.embedding);

const cosine = (a, b) => {
  let dot = 0, na = 0, nb = 0;
  for (let i = 0; i < a.length; i++) { dot += a[i] * b[i]; na += a[i] ** 2; nb += b[i] ** 2; }
  return dot / Math.sqrt(na * nb);
};

const docVectors = await embed(docs);
const [q] = await embed(["How long is delivery?"]);
const ranked = docs
  .map((text, i) => ({ text, score: cosine(q, docVectors[i]) }))
  .sort((a, b) => b.score - a.score);
console.log(ranked[0]);

def answer(question: str) -> str:
    context = "\n".join(text for text, _ in search(question, k=3))
    reply = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": "Answer ONLY from the context below, in Persian. If the answer is not there, say you don't know.\n\nContext:\n" + context},
            {"role": "user", "content": question},
        ],
        max_tokens=300,
    )
    return reply.choices[0].message.content

res = client.embeddings.create(model="text-embedding-3-large", input="...", dimensions=768)
print(len(res.data[0].embedding))   # 768 instead of the full size: smaller index, a little less precise