aurorahq/aurora

By aurorahq

•Updated about 20 hours ago

Aurora OSS — open-source OpenAI-compatible AI gateway for multi-provider LLM routing

Image

Networking

API management

Machine learning & AI

808

Overview Tags

aurorahq/aurora repository overview

⁠Aurora OSS — AI Gateway

Aurora OSS is an open-source, OpenAI-compatible AI gateway that routes requests across multiple LLM providers. It acts as a unified API endpoint for OpenAI, Anthropic, Gemini, Groq, xAI, Z.ai, vLLM, and custom providers — with built-in caching, guardrails, usage tracking, provider pools, and a management dashboard.

⁠Quick Start

docker pull aurorahq/aurora:1.0.30

Run with a single provider:

docker run -d -p 8080:8080 \
  --name aurora \
  -e GROQ_API_KEY=gsk_your_key_here \
  -e AURORA_MASTER_KEY=your-admin-key \
  aurorahq/aurora:1.0.30

# Verify it works
curl http://localhost:8080/health
# → {"status":"ok"}

# List available models
curl http://localhost:8080/v1/models
# → {"object":"list","data":[{"id":"groq/llama-3.3-70b-versatile",...}]}

# Chat completion
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"groq/llama-3.3-70b-versatile","messages":[{"role":"user","content":"Hello!"}]}'

⁠Features

Feature	Description
OpenAI-compatible API	Drop-in replacement — use existing OpenAI SDKs by changing `base_url`
Multi-provider routing	Route requests to OpenAI, Anthropic, Gemini, Groq, xAI, Z.ai, vLLM, and more
Provider pools	Load-balance across multiple API keys with round-robin, least-loaded, or weighted strategies
Response caching	Redis or local SQLite cache to reduce costs and latency
Guardrails	Content filtering and safety checks on requests/responses
Usage tracking	Per-user and per-model token counting, pricing, and billing
Admin dashboard	Web UI and REST API for managing providers, models, and usage
Prometheus metrics	`/metrics` endpoint for monitoring with Grafana
Token optimization	Token saver reduces token usage on long conversations
Workflows	Automate multi-step LLM pipelines
Authentication	API key-based auth, optional master key for admin
Anthropic ingress	Native Anthropic API format support alongside OpenAI format

⁠Configuration

Aurora uses a 3-layer config chain: code defaults → config.yaml → environment variables (env vars win).

⁠Via Environment Variables

The image ships with ${VAR:-default} syntax throughout. Override any setting at runtime without rebuilding:

⁠Core Settings

Env Variable	Default	Description
`PORT`	`8080`	HTTP server port
`BASE_PATH`	`/`	URL path prefix for reverse proxy
`AURORA_MASTER_KEY`	(none)	Admin API key (empty = no auth)
`SQLITE_PATH`	`data/aurora-oss.db`	SQLite storage path
`STORAGE_TYPE`	`sqlite`	Storage backend (`sqlite` or `redis`)
`LOG_FORMAT`	(auto)	Log format: `json` or `text`
`LOG_BODIES`	`false`	Log request/response bodies
`LOGGING_ENABLED`	`false`	Enable structured logging

⁠Features

Env Variable	Default	Description
`METRICS_ENABLED`	`false`	Enable Prometheus metrics at `/metrics`
`METRICS_ENDPOINT`	`/metrics`	Metrics endpoint path
`GUARDRAILS_ENABLED`	`false`	Enable content guardrails
`USAGE_ENABLED`	`true`	Enable usage tracking and pricing
`COMBOS_ENABLED`	`true`	Enable model combos/routing
`TOKEN_SAVER_ENABLED`	`false`	Enable token optimization
`ADMIN_ENDPOINTS`	`true`	Enable admin API and dashboard
`MODELS_ENABLED_BY_DEFAULT`	`true`	Auto-register provider models
`EDITION`	`oss`	Edition tag

⁠Model Caching

Env Variable	Default	Description
`MODEL_LIST_URL`	(none)	URL to fetch model list
`MODEL_LIST_LOCAL_PATH`	`data/models.local.json`	Local model list path
`MODEL_LIST_USER_OVERRIDES_PATH`	`data/user_pricing.yaml`	User pricing overrides
`MODEL_REFRESH_INTERVAL`	`3600`	Model list refresh interval (seconds)

⁠Provider API Keys

Env Variable	Description
`OPENAI_API_KEY`	OpenAI API key
`OPENAI_BASE_URL`	Custom OpenAI-compatible endpoint
`ANTHROPIC_API_KEY`	Anthropic API key
`ANTHROPIC_BASE_URL`	Custom Anthropic endpoint
`GEMINI_API_KEY`	Google Gemini API key
`GROQ_API_KEY`	Groq API key
`XAI_API_KEY`	xAI (Grok) API key
`ZAI_API_KEY`	Z.ai API key
`ZAI_BASE_URL`	Custom Z.ai endpoint
`ORACLE_API_KEY`	Oracle AI API key
`ORACLE_BASE_URL`	Oracle OpenAI-compatible endpoint
`VLLM_API_KEY`	vLLM API key (optional)
`VLLM_BASE_URL`	vLLM endpoint URL

⁠Via Custom Config File

Mount your own config.yaml for full control:

docker run -d -p 8080:8080 \
  --name aurora \
  -v $(pwd)/my-config.yaml:/app/configs/config.yaml \
  -e AURORA_MASTER_KEY=your-admin-key \
  aurorahq/aurora:1.0.30

⁠Kubernetes Deployment (Helm)

A full Helm chart is available for production deployments.

⁠Prerequisites

Kubernetes cluster (minikube, EKS, AKS, GKE, etc.)
Helm v3+

⁠Quick Install

# Clone or add the Helm chart
helm install aurora ./helm \
  --namespace aurora --create-namespace \
  --set image.tag=1.0.25 \
  --set providers.openai.apiKey=sk-... \
  --set providers.openai.enabled=true \
  --set redis.enabled=true

⁠Minimal Dev Install (No Redis, No Auth)

helm install aurora ./helm \
  --namespace aurora --create-namespace \
  --set image.tag=1.0.25 \
  --set providers.groq.apiKey=gsk_... \
  --set providers.groq.enabled=true \
  --set redis.enabled=false \
  --set auth.masterKey=""

⁠Production Install

helm upgrade --install aurora ./helm \
  --namespace aurora --create-namespace \
  --set image.tag=1.0.25 \
  --set auth.masterKey=your-secure-master-key \
  --set providers.openai.apiKey=sk-... \
  --set providers.openai.enabled=true \
  --set providers.anthropic.apiKey=sk-ant-... \
  --set providers.anthropic.enabled=true \
  --set redis.enabled=true \
  --set replicaCount=3

⁠Custom Configuration

# Mount custom config.yaml from a file
helm upgrade aurora ./helm \
  --set-file config.content=my-custom-config.yaml

# Or use an existing ConfigMap
helm upgrade aurora ./helm \
  --set config.existingConfigMap=my-aurora-config

⁠Key Helm Parameters

Parameter	Default	Description
`image.tag`	(appVersion)	Docker image tag
`image.pullPolicy`	`IfNotPresent`	Image pull policy
`replicaCount`	`2`	Number of pod replicas
`server.port`	`8080`	Service port
`server.basePath`	`/`	URL path prefix
`auth.masterKey`	`""`	Master API key (empty = no auth)
`providers.*.apiKey`	`""`	Provider API keys
`providers.*.enabled`	`false`	Enable provider
`providers.*.baseUrl`	`""`	Custom provider URL
`redis.enabled`	`true`	Deploy Redis subchart
`metrics.enabled`	`true`	Enable Prometheus metrics
`securityContext.readOnlyRootFilesystem`	`true`	Read-only root filesystem
`config.content`	`""`	Custom config.yaml content
`config.existingConfigMap`	`""`	Existing config ConfigMap

⁠Links

GitHub repository: github.com/aurorallm/aurora⁠
npm package: npmjs.com/package/iaurora⁠
Helm chart: github.com/aurorallm/aurora/tree/main/helm⁠
Docker images: hub.docker.com/r/aurorahq/aurora⁠
Issue tracker: github.com/aurorallm/aurora/issues⁠
Changelog: github.com/aurorallm/aurora/releases⁠

⁠License

Apache 2.0⁠

Tag summary

Recent tags

Content type

Image

Digest

sha256:466ec5392…

Size

15.4 MB

Last updated

about 20 hours ago

docker pull aurorahq/aurora