Velaxe
AI Hub — Unified LLM Gateway, Chat, Embeddings & Jobs | Velaxe

AI Hub

Create embeddings for RAG & similarity search

Use /api/handler/execute with task=encode to get vectors via Hugging Face.

10 min Intermediate Developer Updated Sep 19, 2025
View App All Guides

Overview

Use /api/handler/execute with task=encode to get vectors via Hugging Face.

Prerequisites

  • Hugging Face token in KeyVault ("huggingface")

Permissions required

ai.backend.call

Downloads & Templates

Embedding request sample

Spec

Steps (2)

Estimated: 10 min
  1. 1

    Call the bridge

    Developer 4 min Back to top

    POST /api/handler/execute {"task":"encode","texts":["hello","world"],"model":"sentence-transformers/all-mpnet-base-v2"}

    Tips

    —

    Validation

    • {"ok":true,"embeddings":[[…],[…]]} shape matches input count.

    Success criteria

    —

  2. 2

    Persist vectors

    Developer 6 min Back to top

    Store vectors in your app’s index (e.g., SQLite/pg/vector store). Keep an id↔embedding mapping.

    Tips

    —

    Validation

    —

    Success criteria

    —

About this guide

AI Hub centralizes generative AI for your workspace with a single, policy-aware gateway to multiple providers. Teams get a streamlined chat experience with searchable history and feedback, a minimal Query API for quick prompts, and embeddings for retrieval workflows. Operators gain visibility with usage & cost tracking, quotas, and exportable audit logs.

Choose the best model for each task, fail over between providers, and moderate inputs/outputs with block/warn/allow policies. Keys are encrypted at rest and scoped per workspace. Long-running tasks run on a background worker and broadcast events so other apps can react in real time.

Designed for safety and speed: opinionated defaults, least-privilege access, and drop-in APIs that make it easy to bring AI to every surface of Velaxe.