modules/deploying-a-llama-model-with-kserve.adoc (1)

13-16: Anchor/terminology mismatch between “activated” wording and linked section

The prerequisite now says activated the Llama Stack Operator, but the link that follows still points to #installing-the-llama-stack-operator_rag.
Consider renaming the target anchor (and page heading) to activating for consistency, or adjust the prose here back to installed to avoid sending readers to a differently-worded section.

No functional impact, but improving this avoids cognitive friction.

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

Learnt from: bredamc
PR: opendatahub-io/opendatahub-documentation#860
File: modules/using-llm-as-a-judge-metrics-with-lmeval.adoc:62-65
Timestamp: 2025-07-16T09:05:50.422Z
Learning: In the opendatahub-documentation project, bredamc prefers comprehensive error detection in code reviews. Continue to point out potential errors whether they are related to the code changes in the PR or not, but clearly distinguish between errors introduced in the PR versus errors in existing code.

Learnt from: jiridanek
PR: opendatahub-io/opendatahub-documentation#888
File: modules/about-workbench-images.adoc:105-105
Timestamp: 2025-07-31T15:09:05.126Z
Learning: In the opendatahub-documentation project, ROCm workbench images come pre-integrated with additional frameworks and dependencies that users can utilize, rather than requiring users to integrate frameworks themselves. When describing these images, the focus should be on the pre-integrated capabilities rather than user integration tasks.

modules/deploying-a-llamastackdistribution-instance.adoc (1)

114-114: Using distribution.name is correct and resolves the earlier placeholder-image concern

Specifying spec.server.distribution.name: rh-dev (instead of distribution.image) aligns with the operator-managed resolution approach discussed in this PR and addresses the “invalid image reference” feedback from earlier reviews.

Please confirm that the documented Operator version in this release supports resolving the rh-dev alias (so readers on that version won’t see reconciliation failures). If there are version constraints, consider noting them briefly in the surrounding text.

modules/deploying-a-llamastackdistribution-instance.adoc (2)

7-7: Tighten abstract phrasing; keep key directive about distribution.name

Current sentence is repetitive (“You can… You can…”). Consider a concise, active rewrite.

-You can integrate LlamaStack and its retrieval-augmented generation (RAG) capabilities with your deployed Llama 3.2 model served by vLLM. You can use this integration to build intelligent applications that combine large language models (LLMs) with real-time data retrieval, providing more accurate and contextually relevant responses for your AI workloads. When you create a `LlamaStackDistribution` custom resource (CR), specify `rh-dev` in the `spec.server.distribution.name` field. 
+Integrate LlamaStack’s retrieval-augmented generation (RAG) with a vLLM‑served Llama 3.2 model to build applications that combine LLMs with real‑time data retrieval for more accurate, contextually relevant responses. When creating a `LlamaStackDistribution` custom resource (CR), set `spec.server.distribution.name` to `rh-dev`.

---

117-120: Clarify what ‘rh-dev’ is and add guidance for pinning and disconnected clusters

Good note. Minor tightening and adding two reader‑helpful points: it’s an Operator‑managed alias, pinning implications, and a reminder about mirroring for disconnected setups (you already link to mirroring guidance above).

-[NOTE]
-====
-The `rh-dev` value is an internal image reference. When you create the `LlamaStackDistribution` custom resource, the {productname-short} Operator automatically resolves `rh-dev` to the container image in the appropriate registry. This internal image reference allows the underlying image to update without requiring changes to your custom resource.
-====
+[NOTE]
+====
+`rh-dev` is an Operator‑managed distribution alias. When you apply the `LlamaStackDistribution` custom resource, the {productname-short} Operator resolves this alias to the correct registry image.
+
+This indirection lets the Operator deliver image updates without modifying your custom resource. If you require a pinned version for reproducibility, follow the Operator’s guidance for selecting a fixed release/channel.
+
+For disconnected clusters, ensure the resolved image is mirrored to your internal registry and that image source policies are configured, as described in the mirroring documentation above.
+====

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro