Most apps that call LLMs have the same setup: hardcoded API keys, manual provider switching, bespoke retry logic.
Aurora Gateway - is The fastest LLM gateway you integrate into your backend.
npm install -g iaurora
aurorallm.online/
github.com/aurorallm/aurora
Aurora Gateway — OSS (Apache 2.0) + Enterprise
npm: npm install -g iaurora
GitHub: github.com/aurorallm/aurora
Docs: aurorallm.online/docs
NPM: npmjs.com/package/iaurora
55x faster than litellm python.
OSS edition (Apache 2.0) ships with auto-discovery, provider pools,
caching, guardrails, audit logging, and analytics — everything you need
for production. Enterprise adds SSO, RBAC, and tenant isolation when
your team scales.
Built in Go (Echo v5). Storage: SQLite, Postgres, or Mongo.
Caching: Redis. Vector stores: Qdrant, pgvector, Pinecone.
Dashboard: React SPA embedded in the binary.
Deploy via single binary, Docker, Docker Compose, or Helm.
Observability and cost control:
• Full request/response audit logging with live SSE stream
• Per-model token counting and cost tracking
• Scoped API keys with rate limits and budgets
• Token saver — policy-driven output compression
One place to see every dollar spent.
Guardrails built in:
• PII redaction (email, phone, SSN, credit cards)
• Prompt injection detection
• Regex content filters
• System prompt protection
• LLM-based message rewriting
All configured through the dashboard. No extra services.
Three caching layers:
• Exact — SHA-256 hash match, Redis-backed, sub-ms
• Semantic — vector similarity (Qdrant, pgvector, Pinecone)
• Prompt — native Anthropic/OpenAI prompt caching
Each one means fewer API calls and lower latency.
Provider pools:
Group multiple keys and endpoints. Round-robin or weighted routing.
Health-aware failover — if one provider returns 5xx or 429, traffic moves to the next. Zero config changes at the app level.
