Docs/vllm ascend modelcar permissions

[fyuan1316]

Summary by CodeRabbit

Documentation

• Updated installation guidance with latest Knative version and clarified Model Catalog password configuration steps.
• Enhanced custom inference runtime documentation with detailed resource sizing guidance and Ascend 910 deployment mode specifications.
• Expanded troubleshooting section with Ascend 910 vLLM-ascend permission mode guidance and cross-references.

[coderabbitai]

Walkthrough

This PR updates documentation across cluster installation, custom inference runtime guidance, and troubleshooting to improve clarity on Kubernetes configuration. Changes include a Knative version bump, clarification of Model Catalog secret setup, callout annotations for resource sizing across multiple runtimes, and comprehensive new guidance on vLLM-ascend security and permission modes for Ascend hardware.

Changes

Documentation Updates: Cluster Installation and Inference Runtime Guidance

Layer / File(s)	Summary
Cluster setup and Model Catalog configuration docs/en/installation/ai-cluster.mdx	Knative KnativeServing version updated to 1.19.6. Model Catalog secret namespace/name fields clarified with guidance to create the secret before instance creation. Kubernetes stringData.password encoding behavior documented to clarify no manual base64-encoding is required.
Resource sizing callouts for standard runtimes docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx	Xinference, MLServer, Triton, and MindIE YAML examples annotated with CPU/memory resource limit callout markers and corresponding explanation blocks added for each runtime.
vLLM-ascend permission modes and security configuration docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx	New "Modelcar Permission Modes for Ascend 910" subsection added with non-root vs root deployment guidance and cluster-level warning. vLLM-ascend ClusterServingRuntime YAML updated with HOME/USER environment variables and runAsUser 1000 with callout markers. InferenceService example reordered with updated resource limits and expanded fsGroup/supplementalGroups guidance. Runtime comparison table updated to reflect new requirements.
Troubleshooting updates and conclusion alignment docs/en/model_inference/inference_service/how_to/using_modelcar.mdx	Ascend 910 vLLM-ascend permission mode troubleshooting item added with link to permission modes documentation. Conclusion section consolidated to single continuous paragraph.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• alauda/aml-docs#210: Both PRs modify docs/en/installation/ai-cluster.mdx in the Model Catalog setup section around PostgreSQL password secret configuration.

Poem

🐰 A cluster takes shape, with Knative up high,
Ascend modes unfold beneath the K8s sky—
Resource callouts dance, from Xinference to vLLM,
Secrets encoded smooth, just add them and fill 'em!
Documentation blooms, let troubleshooting be clear,
The inference path brightens—deployment's right here! ✨

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title directly relates to the main changes in the pull request, which focus on documenting vLLM-ascend modelcar permission modes and Ascend-specific security configurations across multiple documentation files.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch docs/vllm-ascend-modelcar-permissions

Warning

There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure.

🔧 ESLint

If the error stems from missing dependencies, add them to the package.json file. For unrecoverable errors (e.g., due to private dependencies), disable the tool in the CodeRabbit configuration.

ESLint skipped: no ESLint configuration detected in root package.json. To enable, add eslint to devDependencies.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[cloudflare-workers-and-pages]

Deploying alauda-ai with    Cloudflare Pages

Latest commit:	2ad292c
Status:	✅  Deploy successful!
Preview URL:	https://4ee37638.alauda-ai.pages.dev
Branch Preview URL:	https://docs-vllm-ascend-modelcar-pe.alauda-ai.pages.dev

View logs

[coderabbitai]

🧹 Nitpick comments (1)

docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx (1)

1010-1010: 💤 Low value

Consider improving table cell readability.

The "Special Requirements" cell for vLLM-ascend is comprehensive but quite long (103 words). For better readability in the comparison table, consider either:

1. Breaking this into bullet points within the cell (if your MDX table renderer supports it), or
2. Using a shorter summary with a reference link to the detailed section, such as:

   **Must** set `HOME` and `USER` environment variables and configure security context according to the [Modelcar permission mode](`#modelcar-permission-modes-for-ascend-910`) (UID 1000 group settings for non-root mode only)
   

The current content is accurate and complete, so this is an optional readability improvement.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In
`@docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx`
at line 1010, Shorten the long "Special Requirements" cell for vLLM-ascend in
the MDX table by either replacing the 103-word sentence with a concise summary
and a link to the detailed section (e.g., reference the "Modelcar permission
mode" anchor) or by splitting it into 2–3 bullet points if the table renderer
supports inline lists; keep key symbols referenced exactly
(ClusterServingRuntime, HOME, USER, Modelcar permission mode, UID 1000) so
readers can find the detailed instructions in the longer section.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In
`@docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx`:
- Line 1010: Shorten the long "Special Requirements" cell for vLLM-ascend in the
MDX table by either replacing the 103-word sentence with a concise summary and a
link to the detailed section (e.g., reference the "Modelcar permission mode"
anchor) or by splitting it into 2–3 bullet points if the table renderer supports
inline lists; keep key symbols referenced exactly (ClusterServingRuntime, HOME,
USER, Modelcar permission mode, UID 1000) so readers can find the detailed
instructions in the longer section.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 0eeccdd4-401a-43aa-b31c-b62754a2ff18

📥 Commits

Reviewing files that changed from the base of the PR and between c766bc1 and 2ad292c.

📒 Files selected for processing (3)

• docs/en/installation/ai-cluster.mdx
• docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx
• docs/en/model_inference/inference_service/how_to/using_modelcar.mdx