docs: update vLLM loader registration docs

[zhengluo-nv]

Summary

• Document --load-format modelexpress as the canonical vLLM load format and mx as the backward-compatible alias.
• Remove stale mx-source / mx-target references and obsolete --worker-cls quick-start usage.
• Remove unsupported MX_REGISTER_LOADERS docs and point users to vLLM plugin registration, VLLM_PLUGINS=modelexpress when explicit plugin selection is needed, or manual register_modelexpress_loaders().

Bugs

• NVBug 6235739
• NVBug 6235844

Validation

• git diff --check
• rg -n MX_REGISTER_LOADERS . returns no matches
• python3 -m py_compile modelexpress_client/python/modelexpress/vllm_worker.py
• Parsed touched YAML manifests with Ruby YAML

Summary by CodeRabbit

• Documentation
  • Standardized vLLM loader format naming: modelexpress is now the primary format with mx as a backward-compatible alias
  • Updated all example configurations and deployment manifests to reflect the new terminology
  • Removed deprecated MX_REGISTER_LOADERS environment variable
  • Added MODEL_EXPRESS_URL environment variable documentation

[AndyDai-nv]

LGTM, let's wait for CI

[coderabbitai]

Actionable comments posted: 0

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 53fe0eb3-78f5-4d3c-a9cb-0fe3e7b94ac2

📥 Commits

Reviewing files that changed from the base of the PR and between 26bd0cd and 8d04e09.

📒 Files selected for processing (23)

• CONTRIBUTING.md
• README.md
• docs/ARCHITECTURE.md
• docs/DEPLOYMENT.md
• docs/metadata.md
• examples/dynamo_p2p_transfer_k8s/README.md
• examples/dynamo_p2p_transfer_k8s/vllm/vllm-multi-node-aggregated.yaml
• examples/dynamo_p2p_transfer_k8s/vllm/vllm-single-node-disaggregated.yaml
• examples/k8s_service_sources/sources-tp2-single-pod.yaml
• examples/k8s_service_sources/sources-tp2.yaml
• examples/k8s_service_sources/target.yaml
• examples/model_streamer_k8s/README.md
• examples/model_streamer_k8s/client/vllm/README.md
• examples/model_streamer_k8s/client/vllm/vllm-single-node-streamer-azure.yaml
• examples/model_streamer_k8s/client/vllm/vllm-single-node-streamer-local.yaml
• examples/model_streamer_k8s/client/vllm/vllm-single-node-streamer-s3.yaml
• examples/p2p_transfer_k8s/README.md
• examples/p2p_transfer_k8s/client/vllm/aws_efa/vllm-aws-efa.yaml
• examples/p2p_transfer_k8s/client/vllm/vllm-multi-node.yaml
• examples/p2p_transfer_k8s/client/vllm/vllm-single-node-p2p.yaml
• examples/p2p_transfer_k8s/client/vllm/vllm-single-node.yaml
• modelexpress_client/python/README.md
• modelexpress_client/python/modelexpress/vllm_worker.py

💤 Files with no reviewable changes (1)

• CONTRIBUTING.md

---

Walkthrough

Documentation and Kubernetes deployment examples updated to standardize vLLM load-format terminology from mx (primary) to modelexpress (with mx as backward-compatible alias). Core project documentation, deployment guides, Python client docs, and example manifests across all P2P and streaming deployment patterns reflect the new naming convention.

Changes

Load Format Naming Standardization

Layer / File(s)	Summary
High-level project documentation updates README.md, CONTRIBUTING.md, docs/ARCHITECTURE.md	Core README, CONTRIBUTING.md, and ARCHITECTURE.md updated to reflect modelexpress as the primary load-format value, with mx documented as backward-compatible alias. Documentation for register_modelexpress_loaders() updated to describe registering both loaders. vllm_worker.py reworded as compatibility worker for manual-registration workflows. MX_REGISTER_LOADERS environment variable entries removed from multiple docs.
Deployment and metadata documentation docs/DEPLOYMENT.md, docs/metadata.md	docs/DEPLOYMENT.md and docs/metadata.md updated to show --load-format modelexpress in environment variable tables, configuration guidance, and CLI examples. VLLM_PLUGINS documentation refined to document both modelexpress and mx loaders.
Python client library documentation and module docstring modelexpress_client/python/README.md, modelexpress_client/python/modelexpress/vllm_worker.py	modelexpress_client/python/README.md updated to emphasize vllm.general_plugins discovery with --load-format modelexpress and remove legacy MX_REGISTER_LOADERS guidance. vllm_worker.py module docstring reworded from "Usage:" to "Compatibility usage:" to mark it as a legacy compatibility extension.
Kubernetes example manifest updates examples/p2p_transfer_k8s/..., examples/model_streamer_k8s/..., examples/k8s_service_sources/..., examples/dynamo_p2p_transfer_k8s/...	Kubernetes YAML manifests across all example directories updated with --load-format modelexpress instead of --load-format mx. Example README files for each deployment type updated to document the new format with mx backward-compatibility note.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Poem

🐰 A format by any other name—
From mx we dance to modelexpress fame,
With backward-compatible grace,
The aliases embrace,
And YAML configs sparkle the same! ✨

🚥 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 'docs: update vLLM loader registration docs' accurately reflects the main objective of the pull request—updating documentation to standardize vLLM loader registration terminology (from mx to modelexpress with mx as an alias) and removing obsolete configuration references.
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.}

---

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.}