0063df0a87
- Бэкенд: FastAPI + SQLAlchemy async + PostgreSQL, JWT-аутентификация, Qdrant multi-tenant векторное хранилище, Celery воркер - Фронтенд: React + Vite + TypeScript SPA (логин, регистрация, дашборд, поиск по коллекциям, настройки) - Инфраструктура: docker-compose, Helm-чарт, GitHub Actions CI, Prometheus/Grafana мониторинг - Embedding через OpenRouter (nvidia/llama-nemotron-embed-vl-1b-v2:free) - LLM: google/gemma-4-31b-it:free через OpenRouter
35 lines
917 B
Python
35 lines
917 B
Python
from openai import AsyncOpenAI
|
|
|
|
from app.config import settings
|
|
|
|
client = AsyncOpenAI(
|
|
base_url="https://openrouter.ai/api/v1",
|
|
api_key=settings.openrouter_api_key,
|
|
)
|
|
|
|
EMBEDDING_MODEL = settings.openrouter_embedding_model
|
|
IS_NVIDIA = "nvidia" in EMBEDDING_MODEL
|
|
|
|
|
|
def _prefix_text(text: str, prefix: str) -> str:
|
|
if IS_NVIDIA:
|
|
return f"{prefix}: {text}"
|
|
return text
|
|
|
|
|
|
async def embed_texts(texts: list[str]) -> list[list[float]]:
|
|
prefixed = [_prefix_text(t, "passage") for t in texts]
|
|
response = await client.embeddings.create(
|
|
model=EMBEDDING_MODEL,
|
|
input=prefixed,
|
|
)
|
|
return [item.embedding for item in response.data]
|
|
|
|
|
|
async def embed_query(text: str) -> list[float]:
|
|
prefixed = _prefix_text(text, "query")
|
|
response = await client.embeddings.create(
|
|
model=EMBEDDING_MODEL,
|
|
input=[prefixed],
|
|
)
|
|
return response.data[0].embedding
|