docs: migrate Speculative Decoding docs to three-tier structure

[dagil-nvidia]

Summary

• Create docs/features/speculative_decoding/README.md with feature overview and backend support matrix
• Add deprecation notice to docs/backends/vllm/speculative_decoding.md pointing to new location
• Add cross-reference from vLLM README to new feature documentation
• Add redirect in conf.py for backward compatibility
• Add toctree entry in hidden_toctree.rst

Part of the docs hierarchy refactoring effort to organize cross-cutting features.

Test plan

• Docs build passes with standard commands:

  # One-time setup
  uv venv .venv-docs
  uv pip install --python .venv-docs --group docs
  
  # Build docs
  uv run --python .venv-docs --no-project docs/generate_docs.py

• No broken links in new content
• Redirect configured for old path

Summary by CodeRabbit

• Documentation
  • Added new documentation pages covering speculative decoding and vLLM backend integration with setup instructions and examples
  • Added cross-references between related documentation sections
  • Implemented documentation redirects for improved navigation

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Walkthrough

Adds a new Speculative Decoding feature guide and vLLM-specific page, inserts cross-references from existing vLLM docs, adds a Sphinx redirect, and registers the new pages in the hidden toctree. All changes are documentation-only.

Changes

Cohort / File(s)	Summary
Feature Overview docs/features/speculative_decoding/README.md, docs/features/speculative_decoding/speculative_decoding_vllm.md	New comprehensive Speculative Decoding feature guide and a vLLM-specific page outlining workflow, prerequisites, quick-start (vLLM + Eagle3), examples, and backend notes.
vLLM Backend Docs docs/backends/vllm/README.md, docs/backends/vllm/speculative_decoding.md	Added cross-reference links to the new feature guide in Advanced Examples/Quickstart; added notes that content moved and will be removed in future.
Docs Configuration docs/conf.py, docs/hidden_toctree.rst	Added a Sphinx redirect mapping backends/vllm/speculative_decoding → ../../features/speculative_decoding/speculative_decoding_vllm.html and registered new feature pages in the hidden toctree.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐇 I hopped through docs to find a trail,
Left notes of drafts and a verification tale,
Eagles and vLLM peeking through the trees,
Tokens scurry fast on speculative breeze,
A tiny rabbit cheers: "Read on, please!" 🥕

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	❓ Inconclusive	The description provides a clear summary of changes, includes a test plan with verification steps, but is missing a 'Related Issues' section with issue references as specified in the template.	Add a 'Related Issues' section citing the relevant GitHub issue(s) using action keywords like 'Closes' or 'Relates to' as shown in the template.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main change: migrating speculative decoding documentation to a three-tier structure, which aligns with the substantive changes across multiple documentation files.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ 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.}

[dagil-nvidia]

/ok to test 3dd4010

[dagil-nvidia]

/ok to test 0d6bbcb

[dagil-nvidia]

@coderabbitai full review

[coderabbitai]

✅ Actions performed

Full review triggered.

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@docs/features/speculative_decoding/speculative_decoding_vllm.md`:
- Around line 89-105: The example response JSON for the /v1/chat/completions
example is using choices[].text; update it to the OpenAI-compatible chat schema
by replacing choices[].text with choices[].message containing a role (e.g.,
"assistant") and content (string or object) fields—i.e., change the choice
object structure used in the example (keep id, model, usage) so that each choice
uses message.role and message.content instead of text to match the vLLM/Dynamo
integration and OpenAI API.

🧹 Nitpick comments (1)

docs/features/speculative_decoding/speculative_decoding_vllm.md (1)

51-55: Avoid a fixed approval-time promise.

Hugging Face approval latency varies; a softer statement reduces the risk of stale guidance.

📝 Suggested tweak

-Approval usually takes around **5 minutes**.
+Approval time can vary depending on Hugging Face review/traffic.

[dagil-nvidia]

/ok to test a3df9f9