Aurora Gateway OSS — Open-Source AI Gateway | OpenAI & Anthropic Compatible API

Open-source LLM gateway (OSS edition). One API for every AI provider.

Self-hosted. No vendor lock-in. 14 provider types, 30+ LLM providers supported.

Aurora open-source AI gateway admin dashboard showing provider stats and usage metrics

Quick Start

Start routing AI traffic in 60 seconds using any of these deployment methods:

CLI (npm)

npm install -g iaurora
mkdir my-gateway && cd my-gateway
aurora init        # creates config.yaml, .env, data/

Option A — via .env file (recommended):

AURORA_MASTER_KEY=your-secure-key
GROQ_API_KEY=gsk_...

aurora

Option B — via inline env vars (no .env needed):

# Linux / macOS
AURORA_MASTER_KEY=your-secure-key GROQ_API_KEY=gsk_... aurora

# Windows PowerShell
$env:AURORA_MASTER_KEY="your-secure-key"; $env:GROQ_API_KEY="gsk_..."; aurora

# Windows CMD
set AURORA_MASTER_KEY=your-secure-key && set GROQ_API_KEY=gsk_... && aurora

Docker

docker run -d --name aurora -p 8080:8080 \
  -e AURORA_MASTER_KEY="your-secure-key" \
  -e GROQ_API_KEY="gsk_..." \
  aurorahq/aurora

Kubernetes (Helm)

helm install aurora ./helm \
  --namespace aurora --create-namespace \
  --set auth.masterKey="your-secure-key" \
  --set providers.groq.apiKey="gsk_..." \
  --set providers.groq.enabled=true \
  --set redis.enabled=false

Test your gateway

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $(grep AURORA_MASTER_KEY .env | cut -d= -f2)" \
  -d '{"model":"groq/llama-3.3-70b-versatile","messages":[{"role":"user","content":"Hello!"}]}'

Dashboard: http://localhost:8080/admin/dashboard

Setup guides: npm · Docker · Helm/Kubernetes · Source

Enterprise Deployments

Aurora supports enterprise-grade deployments for teams running production AI systems at scale. In addition to private networking, custom security controls, and governance, Aurora Enterprise unlocks advanced capabilities including SSO, RBAC, tenant isolation, budget enforcement, compliance workflows, and production support.

The Enterprise edition is a separate distribution with a signed license.

Kubernetes (Helm)

Deploy on any Kubernetes cluster with the Helm chart.

# Quick dev — Groq, no Redis, no auth
helm install aurora ./helm \
  --namespace aurora --create-namespace \
  --set image.tag=1.0.25 \
  --set providers.groq.apiKey="gsk_your_key_here" \
  --set providers.groq.enabled=true \
  --set redis.enabled=false \
  --set auth.masterKey=""

# Production — multiple providers, auth, Redis
helm upgrade --install aurora ./helm \
  --namespace aurora --create-namespace \
  --set image.tag=1.0.25 \
  --set auth.masterKey="your-secure-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

Full Helm docs: helm/README.md

Providers

Providers are auto-discovered from environment variables. Set any provider's _API_KEY and restart — the provider and its default models appear automatically.

Security note for public docs: The env var names listed below are documentation references. Actual secrets go into your .env file (in .gitignore) or your deployment secrets manager — never commit them.

All supported providers

Provider	Env var	Default base URL	Requires base URL	API key required	Default models
OpenAI	`OPENAI_API_KEY`	`https://api.openai.com/v1`	No	Yes	`gpt-4o`, `gpt-4o-mini`
Anthropic	`ANTHROPIC_API_KEY`	`https://api.anthropic.com/v1`	No	Yes	`claude-sonnet-4`, `claude-opus-4`
Google Gemini	`GEMINI_API_KEY`	`https://generativelanguage.googleapis.com/v1beta/openai`	No	Yes	`gemini-2.5-pro`, `gemini-2.5-flash`
Groq	`GROQ_API_KEY`	`https://api.groq.com/openai/v1`	No	Yes	`llama-3.3-70b`, `qwen3-32b`, `whisper`
DeepSeek	`DEEPSEEK_API_KEY`	`https://api.deepseek.com`	No	Yes	`deepseek-chat`, `deepseek-reasoner`
OpenRouter	`OPENROUTER_API_KEY`	`https://openrouter.ai/api/v1`	No	Yes	300+ models
xAI (Grok)	`XAI_API_KEY`	`https://api.x.ai/v1`	No	Yes	`grok-3`, `grok-3-mini`
Z.ai	`ZAI_API_KEY`	`https://api.z.ai/api/paas/v4`	No	Yes	`glm-4.5`
MiniMax	`MINIMAX_API_KEY`	`https://api.minimax.io/v1`	No	Yes	`minimax-m1`
Azure OpenAI	`AZURE_API_KEY`	—	Yes	Yes	Your deployments
Oracle	`ORACLE_API_KEY`	—	Yes	Yes	`cohere.command-r-plus`
Ollama	`OLLAMA_API_KEY`	`http://localhost:11434/v1`	No	No (optional)	Any local model
vLLM	`VLLM_API_KEY`	`http://localhost:8000/v1`	No	No (optional)	Any served model
Jina (reranker)	`JINA_API_KEY`	—	Yes	Yes	`jina-embeddings-v3`

Per-provider extras

Every provider supports *_MODELS to override auto-discovered models:

OPENAI_MODELS=gpt-4o,gpt-4o-mini,gpt-4-turbo

OpenRouter extras:

OPENROUTER_SITE_URL=https://github.com/aurorallm/aurora
OPENROUTER_APP_NAME=Aurora Gateway

Azure requires API version:

AZURE_API_VERSION=2024-10-21

Custom base URL

Override any provider's endpoint:

OPENAI_BASE_URL=https://my-corp-openai-proxy.example.com/v1

Multiple instances of the same provider

Use suffix notation (underscores become hyphens in the provider name):

OPENAI_EAST_API_KEY=sk-...     # → provider: openai-east
OPENAI_WEST_API_KEY=sk-...     # → provider: openai-west
JINA_API_KEY1=...              # → provider: jina-1

Configuration

The gateway loads settings in this priority order (later wins):

code defaults → config.yaml → .env / environment variables

Generated by aurora init, every section of config.yaml is documented inline:

Section	What it controls
`server`	Port, base path, master key, passthrough, Anthropic ingress
`admin`	Dashboard API and UI
`models`	Discovery, overrides, allowlisting
`storage`	SQLite (default), PostgreSQL, or MongoDB
`logging`	Audit logging of requests/responses
`usage`	Token tracking, pricing, retention
`metrics`	Prometheus endpoint
`guardrails`	Content safety filters
`cache`	Model cache, response cache (exact + semantic)
`combos`	Multi-model combo definitions
`token_saver`	Output compression
`fallback`	Provider failover rules
`resilience`	Retry + circuit breaker

The tables below document every supported env var for reference. In practice you configure these in your .env file (CLI deployments), -e flags (Docker), or --set values (Helm). Never commit secrets to version control.

Essential `.env` variables

Minimal set to get the gateway running:

# ── REQUIRED ──────────────────────────────────────────────
AURORA_MASTER_KEY=replace-with-secure-random-key

# ── AT LEAST ONE PROVIDER ────────────────────────────────
OPENAI_API_KEY=sk-...
# or: GROQ_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, etc.

# ── OPTIONAL BUT RECOMMENDED ─────────────────────────────
PORT=8080                       # Listening port (default: 8080)
STORAGE_TYPE=sqlite             # sqlite (default), postgresql, mongodb
LOG_LEVEL=info                  # debug, info, warn, error
LOG_FORMAT=text                 # text or json

Configuring the same variable across all deployment methods

Each env var below works in all three deploy modes — here is an example:

Variable	`.env` file	Docker `-e` flag	Helm `--set` value
`PORT`	`PORT=9090`	`-e PORT=9090`	`--set server.port=9090`
`AURORA_MASTER_KEY`	`AURORA_MASTER_KEY=...`	`-e AURORA_MASTER_KEY=...`	`--set auth.masterKey=...`
`OPENAI_API_KEY`	`OPENAI_API_KEY=sk-...`	`-e OPENAI_API_KEY=sk-...`	`--set providers.openai.apiKey=...`
`STORAGE_TYPE`	`STORAGE_TYPE=postgresql`	`-e STORAGE_TYPE=postgresql`	`--set storage.type=postgresql`
`REDIS_URL`	`REDIS_URL=redis://...`	`-e REDIS_URL=redis://...`	`--set cache.redis.url=...`

Complete env var reference

All configurable environment variables, organized by subsystem:

Server & Security

Env var	Default	Description
`PORT`	`8080`	HTTP listening port
`BASE_PATH`	`/`	URL path prefix to mount under
`AURORA_MASTER_KEY`	`""`	Master API key for auth
`BODY_SIZE_LIMIT`	`10M`	Max request body size (e.g. `10M`, `1G`, `500K`)
`SWAGGER_ENABLED`	`false`	Enable Swagger UI at `/swagger/index.html`
`PPROF_ENABLED`	`false`	Enable pprof at `/debug/pprof/`
`ENABLE_PASSTHROUGH_ROUTES`	`true`	Provider-native passthrough at `/p/{provider}`
`ALLOW_PASSTHROUGH_V1_ALIAS`	`true`	Allow `/p/{provider}/v1/...` alias routes
`ENABLED_PASSTHROUGH_PROVIDERS`	`openai,anthropic,openrouter,zai,vllm`	Provider types for passthrough
`ENABLE_ANTHROPIC_INGRESS`	`false`	Expose `/v1/messages` for native Anthropic clients
`DISABLE_REQUEST_LOGGING`	`false`	Turn off request logging
`DISABLE_REQUEST_BODY_SNAPSHOT`	`false`	Don't snapshot request bodies
`DISABLE_PASSTHROUGH_SEMANTIC_ENRICHMENT`	`false`	Disable semantic enrichment on passthrough

HTTP Client & Proxy

Env var	Default	Description
`HTTP_TIMEOUT`	`600`	Upstream request timeout (seconds)
`HTTP_RESPONSE_HEADER_TIMEOUT`	`600`	Timeout for upstream response headers (seconds)
`HTTP_PROXY`	—	HTTP proxy URL for upstream calls
`HTTPS_PROXY`	—	HTTPS proxy URL
`NO_PROXY`	—	Hosts to exclude from proxy

Storage

Env var	Default	Description
`STORAGE_TYPE`	`sqlite`	Backend: `sqlite`, `postgresql`, or `mongodb`
`SQLITE_PATH`	`data/aurora-oss.db`	SQLite database file path
`POSTGRES_URL`	—	PostgreSQL connection string
`POSTGRES_MAX_CONNS`	`10`	PostgreSQL connection pool max
`MONGODB_URL`	—	MongoDB connection string
`MONGODB_DATABASE`	`aurora`	MongoDB database name

Model Registry

Env var	Default	Description
`MODEL_LIST_URL`	`https://aurorallm.github.io/aurora/assets/models.json`	External model metadata registry (empty = disabled)
`MODEL_LIST_LOCAL_PATH`	`data/models.local.json`	Local model registry snapshot path
`MODEL_LIST_USER_OVERRIDES_PATH`	`data/user_pricing.yaml`	User pricing override file
`MODELS_ENABLED_BY_DEFAULT`	`true`	Default enabled state for provider models
`MODEL_OVERRIDES_ENABLED`	`true`	Allow per-model overrides (dashboard editing)
`KEEP_ONLY_ALIASES_AT_MODELS_ENDPOINT`	`false`	Hide provider models from `GET /v1/models`, show only aliases
`CONFIGURED_PROVIDER_MODELS_MODE`	`fallback`	`fallback` or `allowlist` — how configured model lists affect inventory

Caching

Model cache:

Env var	Default	Description
`CACHE_REFRESH_INTERVAL`	`3600`	Model registry cache refresh (seconds)
`AURORA_CACHE_DIR`	`.cache`	Local filesystem cache directory
`REDIS_URL`	—	Redis connection URL (enables Redis-backed model cache)
`REDIS_KEY_MODELS`	`aurora:models`	Redis key for model cache
`REDIS_TTL_MODELS`	`86400`	Redis model cache TTL (seconds)

Response cache (exact match):

Env var	Default	Description
`RESPONSE_CACHE_SIMPLE_ENABLED`	`false`	Enable Redis exact-response cache
`REDIS_KEY_RESPONSES`	`aurora:response:`	Redis key prefix for responses
`REDIS_TTL_RESPONSES`	`3600`	Response cache TTL (seconds)

Semantic cache (vector similarity):

Env var	Default	Description
`SEMANTIC_CACHE_ENABLED`	`false`	Enable semantic cache
`SEMANTIC_CACHE_THRESHOLD`	`0.92`	Similarity threshold (0-1)
`SEMANTIC_CACHE_PROMPT_SIMILARITY`	`0.90`	Prompt similarity threshold (0-1)
`SEMANTIC_CACHE_TTL`	`3600`	Entry TTL (seconds)
`SEMANTIC_CACHE_MAX_CONV_MESSAGES`	`3`	Recent conversation messages to embed
`SEMANTIC_CACHE_EXCLUDE_SYSTEM_PROMPT`	`false`	Exclude system prompt from cache key
`SEMANTIC_CACHE_EMBEDDER_PROVIDER`	`openai`	Embedder provider name
`SEMANTIC_CACHE_EMBEDDER_MODEL`	`text-embedding-3-small`	Embedder model
`SEMANTIC_CACHE_VECTOR_STORE_TYPE`	`qdrant`	Backend: `qdrant`, `pgvector`, `pinecone`, `weaviate`
`SEMANTIC_CACHE_QDRANT_URL`	`http://localhost:6333`	Qdrant URL
`SEMANTIC_CACHE_QDRANT_COLLECTION`	`aurora_semantic`	Qdrant collection name
`SEMANTIC_CACHE_QDRANT_API_KEY`	—	Qdrant API key
`SEMANTIC_CACHE_PGVECTOR_URL`	—	pgvector connection string
`SEMANTIC_CACHE_PGVECTOR_TABLE`	`aurora_semantic_cache`	pgvector table name
`SEMANTIC_CACHE_PGVECTOR_DIMENSION`	`1536`	pgvector embedding dimension
`SEMANTIC_CACHE_PINECONE_HOST`	—	Pinecone host URL
`SEMANTIC_CACHE_PINECONE_API_KEY`	—	Pinecone API key
`SEMANTIC_CACHE_PINECONE_NAMESPACE`	—	Pinecone namespace
`SEMANTIC_CACHE_PINECONE_DIMENSION`	`1536`	Pinecone embedding dimension
`SEMANTIC_CACHE_WEAVIATE_URL`	—	Weaviate URL
`SEMANTIC_CACHE_WEAVIATE_CLASS`	`AuroraSemanticCache`	Weaviate class name
`SEMANTIC_CACHE_WEAVIATE_API_KEY`	—	Weaviate API key

Audit Logging

Env var	Default	Description
`LOGGING_ENABLED`	`false`	Enable audit log to storage
`LOGGING_LOG_BODIES`	`true`	Log request/response bodies (may contain PII)
`LOGGING_LOG_HEADERS`	`true`	Log headers (sensitive headers redacted)
`LOGGING_ONLY_MODEL_INTERACTIONS`	`true`	Skip health/metrics/admin endpoints
`LOGGING_BUFFER_SIZE`	`1000`	In-memory queue capacity
`LOGGING_FLUSH_INTERVAL`	`5`	Flush interval (seconds)
`LOGGING_RETENTION_DAYS`	`30`	Auto-delete after N days (0 = forever)

Usage Tracking

Env var	Default	Description
`USAGE_ENABLED`	`true`	Enable token usage tracking
`USAGE_PRICING_RECALCULATION_ENABLED`	`true`	Allow admin pricing recalculation
`ENFORCE_RETURNING_USAGE_DATA`	`true`	Add `stream_options.include_usage=true` to streaming requests
`USAGE_BUFFER_SIZE`	`1000`	In-memory queue capacity
`USAGE_FLUSH_INTERVAL`	`5`	Flush interval (seconds)
`USAGE_RETENTION_DAYS`	`90`	Auto-delete after N days (0 = forever)

Guardrails

Env var	Default	Description
`GUARDRAILS_ENABLED`	`false`	Enable content safety filters globally
`ENABLE_GUARDRAILS_FOR_BATCH_PROCESSING`	`false`	Apply guardrails to `/v1/batches` items

Metrics

Env var	Default	Description
`METRICS_ENABLED`	`false`	Enable Prometheus `/metrics` endpoint
`METRICS_ENDPOINT`	`/metrics`	Metrics endpoint path

Token Saver (output compression)

Env var	Default	Description
`TOKEN_SAVER_ENABLED`	`false`	Enable caveman-style output compression
`TOKEN_SAVER_ENDPOINTS`	`chat_completions`	Endpoints to apply it to
`TOKEN_SAVER_APPLY_STREAMING`	`true`	Apply to streaming responses
`TOKEN_SAVER_OUTPUT_ENABLED`	`false`	Enable output style/profile
`TOKEN_SAVER_OUTPUT_PROFILE`	`concise`	Output profile name
`TOKEN_SAVER_MODELS_INCLUDE`	—	Models to include (comma-separated)
`TOKEN_SAVER_MODELS_EXCLUDE`	—	Models to exclude
`TOKEN_SAVER_PROVIDERS_INCLUDE`	—	Providers to include
`TOKEN_SAVER_PROVIDERS_EXCLUDE`	—	Providers to exclude
`TOKEN_SAVER_ON_ERROR`	`allow`	Behavior on error: `allow` or `block`
`TOKEN_SAVER_EMIT_HEADERS`	`true`	Emit token-saver headers in response
`TOKEN_SAVER_AUDIT_ENABLED`	`true`	Log token-saver actions

Resilience (retry + circuit breaker)

Env var	Default	Description
`RETRY_MAX_RETRIES`	`3`	Upstream retry count
`RETRY_INITIAL_BACKOFF`	`1s`	Initial backoff duration
`RETRY_MAX_BACKOFF`	`30s`	Maximum backoff duration
`RETRY_BACKOFF_FACTOR`	`2.0`	Exponential backoff multiplier
`RETRY_JITTER_FACTOR`	`0.1`	Random jitter fraction
`CIRCUIT_BREAKER_FAILURE_THRESHOLD`	`5`	Failures before circuit opens
`CIRCUIT_BREAKER_SUCCESS_THRESHOLD`	`2`	Successes before circuit closes
`CIRCUIT_BREAKER_TIMEOUT`	`30s`	Time before half-open retry

Fallback

Env var	Default	Description
`FEATURE_FALLBACK_MODE`	`manual`	Fallback mode: `auto`, `manual`, or `off`
`FALLBACK_MANUAL_RULES_PATH`	—	Path to manual fallback rules JSON

Admin & Features

Env var	Default	Description
`ADMIN_ENDPOINTS_ENABLED`	`true`	Enable `/admin/api/v1/*` REST endpoints
`ADMIN_UI_ENABLED`	`true`	Enable `/admin/dashboard` UI
`COMBOS_ENABLED`	`true`	Enable combo model calls
`CLI_TOOLS_ENABLED`	`true`	Enable CLI tools integration
`CLI_TOOLS_APPLY_ENABLED`	`false`	Allow admin/API to apply tool changes
`WORKFLOW_REFRESH_INTERVAL`	`1m`	Workflow refresh interval from storage
`EDITION`	—	Edition identifier (Enterprise use)

Config file path

Env var	Default	Description
`AURORA_CONFIG_PATH`	`configs/config.yaml`	Override path to config YAML

Quick config profiles

Aurora ships pre-built config profiles in configs/editions/:

Profile	File	Use case
OSS	`oss.env.example`	Minimal local — SQLite, no Redis
OSS Local Power	`oss.local-power.env.example`	SQLite + Redis exact cache
OSS Team	`oss.team.env.example`	Postgres + Redis + Qdrant — full team deployment

export AURORA_CONFIG_PATH=configs/editions/oss.team.example.yaml

CLI Reference

The aurora CLI is installed via npm install -g iaurora.

Command	Description
`aurora init`	Scaffolds `config.yaml`, `.env`, `data/` in the current directory
`aurora`	Starts the gateway server (default port 8080)
`aurora --help`	Show all CLI options

aurora init generates:

config.yaml — Full gateway configuration with inline docs for every section
.env — Environment file for secrets and runtime overrides (based on .env.template)
data/ — Directory for SQLite database and local model cache

The gateway merges config.yaml + .env/env vars at startup. Use config.yaml for structure and .env for secrets/keys.

Repository Structure

aurora/
├── apps/                  # Application entrypoints
│   └── aurora/            # Main gateway binary
├── internal/              # Internal packages
│   ├── api/               # HTTP handlers, middleware
│   ├── config/            # Configuration loading and validation
│   ├── providers/         # Provider implementations (OpenAI, Anthropic, etc.)
│   ├── storage/           # SQLite, PostgreSQL, MongoDB backends
│   ├── guardrails/        # Content safety filters
│   ├── cache/             # Exact and semantic caching
│   └── analytics/         # Usage tracking and metrics
├── dashboard-ui/          # React admin dashboard frontend
├── configs/               # Configuration profiles and examples
├── docs/                  # Documentation and assets
├── scripts/               # Build and release scripts
├── test/                  # Test suites
└── helm/                  # Kubernetes Helm charts

Documentation

Need Help?

Join our Discord for community support, setup help, and discussions.

Contributing

We welcome contributions of all kinds! Check out the repository to get started:

Setting up the development environment
Code conventions and best practices
How to submit pull requests
Building and testing locally

License

This project is licensed under the Apache 2.0 License — see the LICENSE file for details.

Built with ❤️ by the Aurora team.

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
.github/workflows		.github/workflows
apps/aurora		apps/aurora
configs		configs
configuration		configuration
dashboard-ui		dashboard-ui
docs-assets		docs-assets
docs		docs
helm		helm
internal		internal
monitoring		monitoring
.dockerignore		.dockerignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
Dockerfile		Dockerfile
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
SECURITY.md		SECURITY.md
artifacthub-repo.yml		artifacthub-repo.yml
docker-compose.yml		docker-compose.yml
go.mod		go.mod
go.sum		go.sum
package.json		package.json

Uh oh!

Folders and files

Latest commit

History

Repository files navigation

Aurora Gateway OSS — Open-Source AI Gateway | OpenAI & Anthropic Compatible API

Quick Start

CLI (npm)

Docker

Kubernetes (Helm)

Test your gateway

Enterprise Deployments

Kubernetes (Helm)

Providers

All supported providers

Per-provider extras

Custom base URL

Multiple instances of the same provider

Configuration

Essential .env variables

Configuring the same variable across all deployment methods

Complete env var reference

Server & Security

HTTP Client & Proxy

Storage

Model Registry

Caching

Audit Logging

Usage Tracking

Guardrails

Metrics

Token Saver (output compression)

Resilience (retry + circuit breaker)

Fallback

Admin & Features

Config file path

Quick config profiles

CLI Reference

Repository Structure

Documentation

Need Help?

Contributing

License

About

Topics

Resources

License

Code of conduct

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 2

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Essential `.env` variables

Packages