HN-DB AI Engineai.hn-db.fun
The multi-model AI inference layer powering every intelligent feature across the HN Group network.
Visit live websiteInterface gallery
Executive summary & purpose
ai.hn-db.fun is the central intelligence pillar of the HN-DB ecosystem: every assistant, scoring engine and content generator in the network routes through it, so model choice, cost and latency are governed in one place.
The challenge
Each product was calling AI providers directly, duplicating prompts, leaking keys into clients and making cost impossible to track across 20+ domains.
The HN Group solution
HN Group built a single gateway with server-side key custody, per-tenant quotas, streaming responses and automatic model fallback — designed, deployed and connected to its custom domain in under 36 hours.
Technical features
- Multi-model routing with automatic fallback
- Streaming SSE responses
- Per-tenant token quotas and budgets
- Prompt template registry with versioning
- Realtime usage analytics dashboard
- Server-side key custody, zero client exposure
API specifications
- POST /v1/chat/completions — OpenAI-compatible streaming
- POST /v1/embeddings — vector generation for RAG
- GET /v1/usage — per-tenant token and cost metering
- WebSocket /v1/stream — low-latency token push
- Bearer auth via HN-DB Auth JWT
Technology stack
- React
- TanStack Start
- Supabase
- Edge Functions
- Tailwind
- AI Gateway
Impact & results
Inference cost cut by 41%, p95 latency down to 780ms, and every new product ships AI features in hours instead of days.
Need an enterprise-grade web application or custom infrastructure like this built for your company?
Request your build now on hnsite24.online — production-ready in 24–48 hours, custom domain included.
Request your build now