Skip to content
← Back to projects Active · AI Gateway / Multi-provider Platform

Helmora AI

Multi-provider AI gateway, ops & chat workspace. Requests route across cloud and local providers — with protocol translation, tools, memory, knowledge bases, and cost controls — behind OpenAI- and Anthropic-compatible APIs.

View version 2 Helmora-AI org ↗

Overview

Helmora AI is a personal ecosystem of AI gateway + control-plane + chat software. Requests are routed across cloud and local providers with protocol translation, tools, memory, knowledge bases, and cost controls — behind OpenAI- and Anthropic-compatible APIs. It is a real, split-into-repos personal platform build, not a demo.

Helmora is grouped into two generations: v2 is the active rebuild (2.0.0-alpha), while v1 is paused. Every repo is public under the Helmora-AI organization.

System Flow

Request lifecycle

User RequestHub GatewayCapability RouterTools & MCPMemory & KnowledgeWeb SearchAgent LoopProvider ExecutorsSecurity & CostFinal Response
  1. 01

    A request enters the Hub

    A client — Web console, SDK, IDE or CLI — hits an OpenAI- or Anthropic-compatible endpoint on Helmora Hub.

  2. 02

    Protocol translation & routing

    The Capability Router plans execution: capability-aware direct or virtual routing with bounded physical attempts, cooldown and concurrency state.

  3. 03

    Enrichment

    Tools and MCP, memory and knowledge bases, plus web search and a bounded agent loop can enrich the request before generation.

  4. 04

    Provider execution

    Upstream executors — OpenAI, Anthropic, Gemini, Ollama, or any OpenAI-compatible endpoint — generate with adaptive streaming and a visible-output commitment boundary.

  5. 05

    Security & cost guard

    Scoped API keys, HttpOnly sessions, rate limits, SSRF protection, and monthly/daily cost ceilings gate the result before it leaves.

  6. 06

    Final response

    The normalized, safely streamed response returns to the client through the same compatible interface.
Active / v2.0.0-alpha

Version 2

Current rebuild — a portable multi-provider gateway and a chat/ops web console. Strict TypeScript, SQLite/WAL, encrypted upstream credentials, protocol translation, routing, tools, MCP, memory, knowledge bases, and cost ceilings.

Helmora-AI GitHub ↗

HelmoraAI-Hub-v2

Multi-provider AI gateway & operations backend — routing, protocol translation (OpenAI + Anthropic compatible), tools, MCP, memory & knowledge, cost ceilings, encrypted credentials, versioned management API.

  • Node.js 22
  • TypeScript
  • SQLite/WAL
  • MCP
  • Routing
Helmora-AI GitHub ↗

HelmoraAI-Web

Chat workspace and operations console for the Hub — React 19 + Vite 8, talks only to the Hub's public HTTP contract. Conversations, memories, knowledge bases, providers, models & routes, audit and usage.

  • React 19
  • Vite 8
  • Chat UI
  • Ops Console
Paused / Deprecated

Version 1

First-generation Helmora Hub — layered AI gateway with providers, modes, hybrid storage (local SQLite / Supabase) and OpenAI-compatible routing, plus an admin SPA.

Paused — Development of version 1 has been paused to focus on completing version 2.

Helmora-AI GitHub ↗

HelmoraHub

Layered AI gateway for the Helmora ecosystem — providers, modes, hybrid storage (local SQLite / Supabase) and OpenAI-compatible routing. Paused in favor of v2.

  • Node.js
  • TypeScript
  • Supabase
  • OpenAI API
Helmora-AI GitHub ↗

HelmoraHub-Frontend

Admin control center for HelmoraHub — Vite + React SPA deployable to Cloudflare Pages. Paused in favor of v2 Web.

  • React
  • Vite
  • Cloudflare Pages

Tech stack

  • Node.js / TypeScript
  • SQLite (WAL)
  • OpenAI-compatible API
  • Anthropic-compatible API
  • MCP (JSON-RPC over HTTP)
  • React 19 + Vite 8
  • Provider routing & fallback
  • Encrypted credential vault
  • Cost ceilings & monitoring
  • Docker / systemd / Pterodactyl

Future direction

  • Multi-node deployments
  • Durable webhook retry
  • Completed MCP streaming semantics
  • More advanced memory controls
  • Deeper provider / platform validation
  • Recovery & deployment-doctor UI