AI Hub

Changelog / Latest Updates (June 2024)

• All LLM providers (OpenAI, Anthropic, Gemini) now route through the backend for chat and key validation.
• Provider portals: The "Run a Command" section has been removed for a cleaner, less confusing UI.
• API Key Testing: The "Test Key" button in each portal now accurately validates keys for all providers and displays success/failure.
• Security: API keys are never sent directly from the frontend to LLM providers; all requests are proxied through the backend.
• MCP Server Configuration: Ensure your Supabase mcp_servers table has correct apiurl entries for each provider (e.g., http://localhost:3001/api).
• Multi-Account Management: Add, remove, and set active accounts for each provider; the active account's key is used for all API calls.

Product Requirements (Updated, June 2024)

Current State

• Authentication: Google OAuth via Supabase is working for app login and Google integrations.
• Integrations: UI for OpenAI, Anthropic, Gemini, GitHub, Google Drive, and Gmail is present. Provider portals allow API key/token entry, but some tokens (notably GitHub) are not reliably saved to localStorage.
• Multi-Account Management: UI supports multiple accounts per provider, but only single-token logic is currently reliable. "Active" account logic is not fully robust.
• AI/LLM Integration: ChatBar supports slash commands and natural language, but command routing to MCP servers is still being debugged for some providers.
• Command Routing: /drive and /gmail commands now route to MCP servers using the Supabase session token, but require correct MCP server entries in Supabase and valid tokens.
• Provider Portal: Some portals (GitHub, OpenAI) have token save issues; UI may show "Saved" but not persist to localStorage.
• Command History: Up/down arrow navigation is implemented in the ChatBar, with improved styling for both dark and light modes.
• MCP Server Discovery: Relies on the mcp_servers table in Supabase; errors occur if entries are missing or duplicated.

---

Known Challenges

• Token Persistence: Some provider portals (notably GitHub) do not reliably save tokens to localStorage, causing command failures.
• MCP Server Lookup: Commands fail if the mcp_servers table is missing the correct provider entry or if there are duplicate/incorrect rows.
• Session Token Expiry: Supabase's Google access token (provider_token) may expire or be missing after page reload; robust token refresh is not yet implemented.
• UI/UX: Some status/error messages are missing or unclear; account management UI is "coming soon" for some providers.
• Command Routing: Some commands are still routed to OpenAI by default if MCP server lookup fails.

---

Development Priorities

1. Fix token persistence in provider portals (ensure tokens are reliably saved and loaded from localStorage).
2. Improve MCP server discovery and error handling (clearer errors, auto-heal for missing/duplicate entries).
3. Implement robust token refresh for Google integrations (backend or frontend).
4. Polish account management UI (multi-account, active account selection, error/status display).
5. Expand and document MCP protocol and server integration.
6. Add more provider integrations and polish command dropdowns.

---

Testing Checklist (Updated)

• Add and save tokens for all providers; confirm they persist in localStorage.
• Run /drive and /gmail commands; confirm they route to MCP server and use the correct token.
• Remove and re-add tokens; confirm UI and command routing update accordingly.
• Add/Remove MCP server entries in Supabase; confirm error handling in the UI.
• Test command history navigation and styling in both dark and light modes.

---

How to Contribute

• See the original Contributing section below.

Overview

AI Hub is an integration platform and marketplace for AI-powered developer tools, focused on free and open tools, multi-account management, and Model Context Protocol (MCP) support. The platform features a modern gallery UI, always-on AI ChatBar (supporting OpenAI, Anthropic, Gemini), secure API key management (via Supabase), and command dropdowns for each provider.

---

Key Features

• Authentication: Google OAuth via Supabase.
• Integrations: Multi-account support for GitHub, Google Drive, Gmail, Zapier, and LLMs (OpenAI, Anthropic, Gemini). Each provider has a management panel for adding/removing accounts and managing API keys.
• Multi-Account Management: Add, select, and remove multiple accounts per provider, with active account selection for API calls.
• AI/LLM Integration: ChatBar supports OpenAI, Anthropic (via a Node/Express proxy for CORS/security), and Gemini. Command routing allows both slash commands and natural language queries.
• Command Dropdowns: Each provider panel includes a comprehensive dropdown of available commands (chat, integration actions, etc.), with autofill and "Run" support for the ChatBar.
• Gallery as Main Interaction Point: The gallery is the central place for users to manage integrations, API keys, and accounts. API key management is handled per provider—no global ChatBar API key field.
• Security: API keys are stored securely in Supabase with RLS, and optional encryption is discussed for production readiness.
• Modern UI/UX: Responsive, accessible, and minimal design inspired by Obsidian. Dark/light mode support.
• Branding: Footer displays 'automationalien.com' in Orbitron font.
• MCP Focus: Designed for MCP server/client support, with plans for MCP server discovery, command routing, and a centralized MCP marketplace.

---

Technical Stack

• Frontend: React, TypeScript, Tailwind CSS
• Backend: Supabase (Postgres, Auth, Storage)
• Integrations: REST APIs for GitHub, Google Drive, Gmail, Zapier
• AI/LLM: OpenAI, Anthropic, Google Gemini (pluggable)
• Deployment: Vercel/Netlify or similar (optional)

---

Getting Started

Prerequisites

• Node.js (v16+ recommended)
• npm or yarn

Installation

npm install
# or
yarn install

Running the App

npm run dev
# or
yarn dev

The app will be available at http://localhost:5173 by default.

---

Usage

• Sign in with Google to access the marketplace and manage integrations.
• Use the gallery to add/manage multiple accounts per provider and securely store API keys.
• Use the ChatBar for natural language or slash command actions across providers (API keys are sourced from the active account in each provider portal).
• Use command dropdowns in each provider panel for quick actions.

---

AI/LLM Model Context Protocol

• All AI/LLM servers implement a standard protocol for context exchange:
  • Request: { prompt: string, context: object, metadata: object }
  • Response: { output: string, tokens_used: number, model_version: string, ... }
• Ensures secure, auditable, and consistent LLM interactions.

---

Manual Testing Checklist: Multi-Account Management & Integrations

1. For each provider (OpenAI, Anthropic, Google Drive, Gmail, etc.):

   • Add multiple accounts/keys.
   • Set different accounts as "Active."
   • Remove accounts and ensure the UI updates.
   • Confirm the "Active" account is used for API calls (check localStorage and network requests).

2. Edge Cases:

   • Remove the active account—does another become active, or does the UI prompt you to add/select one?
   • Add an invalid key—does the UI show an error?
   • Switch between providers and ensure the correct account is always active.

---

Contributing

1. Fork the repo
2. Create your feature branch (git checkout -b feature/YourFeature)
3. Commit your changes (git commit -am 'Add new feature')
4. Push to the branch (git push origin feature/YourFeature)
5. Open a pull request

---

Roadmap

• Debug and polish provider command dropdowns and account scanning
• Implement real backend for server installation and management
• Integrate live AI/LLM providers (OpenAI, Anthropic, etc.)
• Finalize and document the model context protocol
• Add billing and premium server support
• Expand analytics and audit logging
• Expand gallery and provider integrations

---

License

MIT

---

For questions or support, contact Senti Labs