Important notice — please read first​

The DEBtech AI Hub is not a standalone add-on and provides no visible functionality on its own. It is a technical foundation whose sole purpose is to supply our DEBtech AI add-ons with a unified, secure, and observable API layer.

In other words: Without at least one of the supported DEBtech add-ons, the AI Hub has no purpose. Installing it only makes sense if you are using (or plan to use) one of the following add-ons.

---

Add-ons this hub was built for​

• DEBtech Portal AI Suite Pro — AI news, AskBox, interest categories, cover images (hard dependency)
• DEBtech Advertising Suite Pro — AI-optimized product descriptions (next release)
• Future DEBtech AI add-ons — all of them route through this central hub

If you are not running any of these add-ons, do not install the AI Hub — it will sit idle.

---

What does the AI Hub do?​

The AI Hub provides a single unified interface to eight different AI providers. Instead of every add-on bringing its own API keys, retry logic, and cost tracking, the hub bundles it all centrally — with budget enforcement, structured logging, and an admin dashboard.

Supported providers:

• Cloud: OpenAI · OpenRouter · DeepSeek
• Local: Ollama · LocalAI · vLLM · LM Studio · SD.cpp

---

Core Features​

• Multi-provider management — All 8 providers usable simultaneously, each add-on picks its own
• Smart routing — Automatic retry on HTTP 429/500/502/503/504 with exponential backoff
• Fallback provider chain — If OpenAI goes down, OpenRouter/DeepSeek takes over automatically
• Budget tracking & enforcement — Monthly budget per provider; calls are blocked when exceeded
• Streaming support (SSE) — Live-typing output for long chat responses
• API log with filter & search — Every API call with tokens, cost, duration, and prompt excerpt
• Log export — CSV or JSON for analysis in Excel/BI tools
• Reasoning model detection — o-series, GPT-5+, DeepSeek-R1 get longer timeouts and proper parameters automatically
• Setup assistant — Detects OS, RAM, GPU, Docker, Python and recommends suitable models for your hardware
• Zero-load dashboard — Stats are pre-computed every 10 minutes via cron; dashboard loads with 0 SQL queries
• Model test tool — Test chat and image generation straight from the admin panel, including custom prompts

---

Security​

• Admin permission system — Per-admin "Manage AI Hub" permission control
• SSRF protection — Provider URLs validated against an allowlist (AWS/GCE metadata endpoints blocked)
• CSP-compliant JavaScript — No inline onclick, runs under strict Content Security Policies
• API key redaction — Bearer tokens and keys stripped from all error messages and logs
• CSV injection protection — Exported log cells hardened against Excel formula injection
• XSS hardening — All user/provider strings escaped; image URLs only via DOM API (no innerHTML)
• Input validation — Model tester with provider and model allowlist, prompt length cap

---

Admin Dashboard​

• KPI overview: Requests today/week/total · Monthly cost · Avg response time · 24h error rate
• Provider status: Live check for each provider with latency
• Usage per add-on: Which component consumes how many tokens
• Budget monitor: Percentage bar per provider, warning at 80 %
• Recent API calls: The last 10 calls at a glance

---

Technical Details​

• XenForo version: 2.3.9+
• PHP version: 8.1+
• Database tables: 2 (xf_deb_aihub_log, xf_deb_aihub_budget)
• Cron jobs: 3 (stats pre-compute, log cleanup, budget sync)
• Namespace: DEB\AiHub
• Languages: English (master) + German (fully translated, 210 phrases)
• No external dependencies beyond XenForo Core + Guzzle (already included in XF)

---

Currently Supported Image Models (April 2026)​

• GPT Image 1.5 (Flagship) — OpenAI's successor to DALL-E 3, 15 % cheaper
• GPT Image 1 Mini (Budget) — For high-volume use cases, approx. 67 % cheaper than Flagship
• Stable Diffusion 2.1 / SDXL Turbo / FLUX.1 via SD.cpp locally (GPU recommended)

Note: DALL-E 2 and DALL-E 3 have been deprecated by OpenAI as of May 12, 2026 and have already been fully replaced by GPT Image 1.5 / 1 Mini in this add-on.

---

Installation​

• Download the ZIP file
• Install via Admin → Add-ons → Install/Upgrade from Archive
• Under DEBtech AI Hub → Settings, enable at least one provider and enter an API key
• Optional: Visit the setup assistant for hardware analysis and model recommendations
• Install a matching DEBtech AI add-on (e.g. Portal AI Suite Pro)

---

Télécharger V1.0.1 :