vllm-project/semantic-router

docs: add reference recipes and tutorials for common routing patterns

Open

#2,334 opened on Jul 5, 2026

View on GitHub
 (1 comment) (0 reactions) (1 assignee)Go (699 forks)github user discovery
area/docsarea/user-experiencedocumentationenhancementhelp wantedroadmap

Repository metrics

Stars
 (4,293 stars)
PR merge metrics
 (PR metrics pending)

Description

Summary

Add maintained reference recipes and end-to-end tutorials for common vLLM Semantic Router deployment patterns so adopters can start from known-good baselines instead of assembling configs from scattered examples.

Parent roadmap: #2287 Related: #2326, #2330, #2331, #2294 Related focused issue: #2378 for out-of-the-box preset tuning across balance, cost-first, accuracy-first, speed-first, privacy-first, and grounding-first.

Motivation

Adopters need clear production-facing examples for common use cases. Current documentation explains pieces of the system, but there is not yet a maintained set of reference configurations that can be copied, validated, and adapted.

Scope

Create tutorials and reference configs for at least these patterns:

  • complexity routing;
  • model/provider failover;
  • PII and guardrail routing;
  • multi-model selection;
  • custom signal extension;
  • recipe/entrypoint based virtual models;
  • agent-oriented routing and learning diagnostics when available.

Each reference should include:

  • a complete config or recipe;
  • expected request examples;
  • expected routing behavior;
  • validation commands;
  • known limitations and production caveats.

Non-goals

  • Do not add new runtime features only to make examples work.
  • Do not use docs to hide missing validation or broken config surfaces.

Acceptance criteria

  • Maintained reference configs exist under the repo's recipe/config surfaces.
  • Website tutorials explain the supported path for each pattern.
  • Examples pass the relevant config validation gate.
  • The docs link to the related roadmap issues for features that are still experimental.

Validation

make agent-report ENV=cpu CHANGED_FILES="config deploy/recipes website/docs"
make agent-lint CHANGED_FILES="config deploy/recipes website/docs"
make agent-ci-gate CHANGED_FILES="config deploy/recipes website/docs"

Contributor guide