Container Images¶

SentinelGuard can run as a containerized LLM gateway in front of OpenAI, Anthropic, Gemini, DeepSeek, Mistral, MiniMax, Kimi/Moonshot, Ollama, Hugging Face, and other OpenAI-compatible providers.

Images¶

Published images are intended to be pulled by applications, platform teams, and Kubernetes deployments.

ghcr.io/<github-owner>/sentinelguard-gateway:<version>
ghcr.io/<github-owner>/sentinelguard-gateway:latest
aitechnav/sentinelguard:<version>
aitechnav/sentinelguard:latest

Use a version tag for production deployments so upgrades are explicit. Use latest only for local testing or demos.

Run Locally¶

docker run --rm -p 8080:8080 \
  -e OPENAI_API_KEY="$OPENAI_API_KEY" \
  -e SENTINELGUARD_GATEWAY_API_KEY="$SENTINELGUARD_GATEWAY_API_KEY" \
  aitechnav/sentinelguard:0.0.15

Then point OpenAI-compatible clients to:

Base URL: http://localhost:8080/v1
API key:  your SentinelGuard gateway client token

Docker Compose And Kubernetes¶

For local multi-container testing, use the repository docker-compose.yml. For Kubernetes, use examples/kubernetes or the Helm chart under examples/helm/sentinelguard.

The gateway should receive upstream provider keys through environment variables, Kubernetes Secrets, ECS task secrets, or your platform secret manager. Client apps and IDEs should receive only the SentinelGuard gateway token, not the upstream provider key.

Supply Chain Metadata¶

Release images may include SBOM, provenance, and signature metadata so security teams can verify where the image came from and inspect its package contents. Exact verification commands depend on the registry and release tag your organization uses.