mancube/forge-chart repository overview

⁠forge-chart

OCI Helm chart for forge⁠ — the batch-inference coordinator. The container image lives at mancube/forge⁠.

helm install forge oci://registry-1.docker.io/mancube/forge-chart --version 0.1.1 \n  --set 'serveBatch.workers={http://vllm-0.engines.svc:8000}' \n  --set serveBatch.apiKey=<bearer key>

Two shapes: serve-batch (OpenAI Batch REST front door) as a Deployment, and one-shot forge run as a Job whose PVC-backed checkpoint turns pod evictions into resumes. Engines (vLLM/SGLang/llama.cpp) are BYO Services — the chart deploys the coordinator only.

Values reference: deploy/helm/forge⁠.