omnigraph-best-practices

Operate a locally or remotely deployed Omnigraph graph database. Use this skill whenever you see Omnigraph CLI commands (omnigraph init/query/mutate/load/ingest/schema/lint/embed/branch/commit), .pg schema files, .gq query files, RustFS S3 URIs (s3://omnigraph-local/...), remote bearer-authed graph endpoints, 504 errors against a graph, work inside a folder containing omnigraph.yaml or cluster.yaml, or cluster commands (omnigraph cluster validate/plan/apply/approve, omnigraph-server --cluster). Covers cluster-mode declarative deployments (cluster.yaml, plan/apply loop, approval-gated deletes), local RustFS setup, project layout, schema authoring and evolution (plan before apply), query linting, data changes (mutate vs load vs ingest, --mode merge vs overwrite), branches for data review, embeddings, aliases for automation, HTTP server operation, Cedar policy, remote graph operations (504 verification ritual, ingest vs load tradeoffs, version drift), and common gotchas. Especially important BEFORE running schem

By ModernRelay schedule Updated 6/10/2026