bitflicker64 commented on code in PR #393: URL: https://github.com/apache/hugegraph-ai/pull/393#discussion_r4459865848
########## docker/charts/README.md: ########## @@ -0,0 +1,72 @@ +--- +version: 1.8.0 +appVersion: "1.8.0" +--- + +# Kubernetes deployment + +HugeGraph RAG is deployed through the canonical HugeGraph Helm chart in the +[`apache/hugegraph`](https://github.com/apache/hugegraph/tree/master/helm/hugegraph) +repository. The former `docker/charts/hg-llm` chart duplicated the system topology +and has been retired. + +The front matter above remains synchronized by `scripts/version.py` so release +metadata stays testable without retaining a second Helm chart. + +The integration is introduced by +[`hugegraph/hugegraph#365`](https://github.com/hugegraph/hugegraph/pull/365). +Until that change is merged upstream, use its branch when evaluating the +migration. + +## Single-node HugeGraph with RAG + +```bash +git clone https://github.com/hugegraph/hugegraph.git Review Comment: Important: this clones the `hugegraph/hugegraph` fork and checks out `feat/helm-rag-integration`, and line 17 links fork PR `hugegraph/hugegraph#365`; this PR merges after or with apache/hugegraph#3450, so once that branch is deleted these steps fail. Point the clone, checkout and link at `apache/hugegraph` `master`, matching the link on line 9. ########## docker/charts/README.md: ########## @@ -0,0 +1,72 @@ +--- +version: 1.8.0 +appVersion: "1.8.0" +--- + +# Kubernetes deployment + +HugeGraph RAG is deployed through the canonical HugeGraph Helm chart in the +[`apache/hugegraph`](https://github.com/apache/hugegraph/tree/master/helm/hugegraph) +repository. The former `docker/charts/hg-llm` chart duplicated the system topology +and has been retired. + +The front matter above remains synchronized by `scripts/version.py` so release +metadata stays testable without retaining a second Helm chart. + +The integration is introduced by +[`hugegraph/hugegraph#365`](https://github.com/hugegraph/hugegraph/pull/365). +Until that change is merged upstream, use its branch when evaluating the +migration. + +## Single-node HugeGraph with RAG + +```bash +git clone https://github.com/hugegraph/hugegraph.git +cd hugegraph +git checkout feat/helm-rag-integration + +helm upgrade --install hugegraph ./helm/hugegraph \ + -n hugegraph --create-namespace \ + -f helm/hugegraph/values-single.yaml \ + -f helm/hugegraph/values-rag.yaml +``` + +## HugeGraph cluster with RAG + +```bash +helm upgrade --install hugegraph ./helm/hugegraph \ + -n hugegraph --create-namespace \ + -f helm/hugegraph/values-cluster.yaml \ + -f helm/hugegraph/values-rag.yaml +``` + +Both topologies use the same RAG templates. The topology preset controls the +replica count, while `values-rag.yaml` is the explicit on/off overlay. RAG is +disabled by default. + +## Migration mapping + +| Former `hg-llm` value | Canonical chart value | +| --- | --- | +| `image.repository`, `image.tag` | `rag.image.repository`, `rag.image.tag` | +| `replicaCount` | `rag.replicas` | +| `service.type` | `rag.service.type` | Review Comment: Minor: the retired chart defaulted to `service.type: NodePort` with `service.nodePort: 8039`, while the canonical chart defaults to `ClusterIP`, so users who relied on the defaults lose external access with no hint here. Add a `service.nodePort` row mapping to `rag.service.nodePort` that states the `ClusterIP` default. ########## docker/charts/hg-llm/values.yaml: ########## @@ -1,139 +0,0 @@ -# Default values for hg-llm. Review Comment: Minor: `hugegraph-llm/config.md` line 292 still lists `docker/charts/hg-llm/values.yaml` as the Helm chart configuration file. Point it at `docker/charts/README.md`. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
