Open in Playground Download design mesheryctl design import -f https://raw.githubusercontent.com/meshery/meshery.io/master/catalog/0d10efaf-433e-4e3f-a2ff-ec6a4178f52f/0.0.2/design.yml

What This Design Does:

Kubernetes deployment design for Kratos.ai, an intelligent market intelligence platform. Includes a Node.js API gateway (portfolio management, trading simulation, live P&L tracking), an async FastAPI ML service running an LSTM/GRU PyTorch encoder and FinBERT for market regime detection (Stable Growth, Volatile, Crisis) with confidence scoring, and a Redis cache layer. MongoDB Atlas is used as the external data store and is not deployed in-cluster.

Design Snapshot

Kratos.ai design snapshot

Caveats and Consideration:

MongoDB connection is external (Atlas) — the Secret in this design is a placeholder and must be replaced with a real connection string or wired through Meshery Connections/Credentials before deploy. Redis runs as a single replica with no persistence configured — cache loss on pod restart is expected; not suitable as a source of truth. The FastAPI service assumes model artifacts (LSTM/GRU encoder, FinBERT) are pre-loaded onto the mounted PVC (kratos-model-pvc); the design does not include a model-sync/init step. No GPU resource requests are defined for the ML service — PyTorch inference will run on CPU unless you add nvidia.com/gpu limits and a compatible node pool. Resource requests/limits (CPU, memory, storage) are placeholder values and haven't been load-tested against real inference throughput. No liveness/readiness probes, HPA, NetworkPolicy, or Ingress are defined — this design is deploy-ready for a dev/staging cluster, not production-hardened. Image references (kratos-ai/node-gateway: latest, kratos-ai/fastapi-service: latest) are placeholders — replace with your actual registry paths and pin to a specific tag rather than latest.

Compatibility:

Design Details

Related Designs

deployment
vault-0.30.1.tgz
MESHERY40c7
`