Oleksandr Kuzmenko, EditorPublished 3 min readAI-assisted · editor-reviewed
TL;DR
- Deploy a single self-hosted Go binary to manage model fallbacks, semantic caching, and token costs.
- Works natively with OpenAI and Anthropic SDKs by modifying only the base URL parameter.
Key facts
- Supported Providers
- 14 (OpenAI, Anthropic, Gemini, DeepSeek, Groq, Ollama, etc.)
- License
- Apache 2.0
- Tech Stack
- Go, React (Admin SPA), Redis, Postgres
Unified Wire Protocol and Multi-Provider Support
Aurora routes incoming OpenAI or Anthropic SDK calls to 14 provider backends, including OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, and local deployments like Ollama or vLLM. Switching your backend requires changing only the base_url parameter in your standard SDK setup.
Enterprise Resilience and Traffic Control
Aurora supports provider pools with health-aware load balancing, circuit breaking, and automatic fallback rules triggered by rate limits (429) or server errors (5xx). Built-in middleware handles PII redaction, regex filtering, and token compression (TOKEN_SAVER_ENABLED) to reduce response payloads.
Try it in 2 minutes
docker run -d --name aurora -p 8080:8080 -e AURORA_MASTER_KEY="sk-aurora-master" -e ANTHROPIC_API_KEY="sk-ant-..." aurorahq/auroraWhen to use
- You want to centralize LLM API keys, rate limits, and audit logs across teams
- You need multi-provider fallback and load balancing without changing SDK code
When NOT to use
- You are using a single LLM provider in a simple serverless script without caching needs
What to do today
- Clone or pull `aurorahq/aurora` via Docker.
- Set up environment API keys and test swapping `base_url` in your local dev script.
Step names are taken from What to do today.
Step names are taken from What to do today. Clone or pull `aurorahq/aurora` via Docker.. Set up environment API keys and test swapping `base_url` in your local dev script.
- Clone or pull `aurorahq/aurora` via Docker.
- Set up environment API keys and test swapping `base_url` in your local dev script.