semantic-router

Guides coding agents through a session-isolated install of the vllm-sr CLI. The install is scoped to a temporary directory so it does not modify the user's global PATH or system packages. Use when an agent or user wants to try vllm-sr without a permanent install.

Calibrates routing changes against a live router endpoint with executable probes, local DSL validation, versioned deploys, and structured failure review. Use when tuning signals, projections, decisions, or maintained route examples against a real apiserver.

Manages GitHub issue and pull-request lifecycle including creation, updates, triage labelling, and closeout metadata using canonical templates and repository taxonomy. Use when a maintainer asks to create, update, close, or triage GitHub issues or PRs, or when issue creation requires codebase analysis for scope, labels, or acceptance criteria.

Runs an explicit structural review for boundary-sensitive edits that need dependency, module, and structure-rule scrutiny without forcing that material into every default context pack. Use when a task crosses hotspot boundaries, adds new seams, or needs a deliberate architecture pass.

Synchronizes config representations across router config, Python CLI schema, and dashboard config UI. Use when adding or changing a config concept that spans those surfaces or addressing config representation debt before Kubernetes-facing translation.

Maintains dashboard frontend and backend implementation slices that present, configure, or expose router behavior through the console. Use when the task changes dashboard config flows, console backend handlers, topology rendering, playground reveal flows, or other user-visible routing metadata in the dashboard.

Modifies Kubernetes-facing operator, CRD, deployment-profile, or DSL translation behavior for semantic-router platform integration. Use when changing operator APIs or controllers, deployment stack manifests, profile-owned platform wiring, or router-to-Kubernetes translation layers.

Maintains Kubernetes-facing operator, deployment-profile, and DSL translation implementation slices for semantic-router platform integration. Use when the task changes operator APIs or CRDs, deployment manifests, profile-owned stack resources, or router-to-Kubernetes translation behavior.

Builds Docker images, starts local servers, and runs smoke tests for the AMD/ROCm development environment. Use when validating changes locally on AMD hardware, building AMD container images, or running AMD-specific smoke and explicit runtime checks.

Manages Python CLI command orchestration including startup, serve, bootstrap, status checks, and local Docker image management. Use when modifying CLI startup sequences, changing how local images are built or managed, updating serve behavior, or adjusting status command output.

Modifies routing policy after signal extraction, including matched-decision logic, candidate-model selection, and downstream looper behavior. Use when changing decision predicates, thresholds, priorities, model ranking, cost or latency routing, or other post-signal routing policy.

Maintains matched-decision predicates, looper policy, and candidate-model selection behavior after signal extraction. Use when the task changes decision trees, thresholds, downstream model ranking, or other routing policy that runs after signals are already produced.

Implements end-to-end signal changes spanning router config, signal extraction, CLI schema, optional bindings, router-owned metadata headers, and E2E test coverage. Use when adding a new signal type, changing signal configuration or extraction logic, updating CLI schema for signal parameters, or modifying router-owned signal metadata contracts.

Modifies training-stack workflows, selector or embedding pipelines, evaluation scripts, and runtime-facing artifact expectations. Use when a primary skill touches src/training, tools/make/models.mk, or training docs for runtime-fed artifacts.

Selects which local and CI end-to-end test profiles are affected by a code change, using the repo-local profile map. Use when a change could affect E2E test behavior and the correct test profiles need to be identified and executed.

Modifies the fleet simulator package, API surface, release workflow, and simulator-owned docs or assets. Use when a primary skill touches src/fleet-sim or the simulator subsystem's owned website/release surfaces.

Implements end-to-end plugin changes spanning router config, post-decision processing, optional CLI/UI exposure, and E2E test coverage. Use when adding a new plugin type, changing plugin config schema or execution semantics, updating plugin chain behavior, or modifying plugin-exposed metadata across surfaces.

Modifies router-side API, authz, memory, provider, storage, or runtime service modules outside config, decision, selection, and extproc plugin chains. Use when changing apiserver endpoints, authz or rate-limit policy code, memory or response storage flows, provider adapters, or other router service-platform modules.

Modifies router-side API, authz, memory, provider, storage, and service-support modules outside config, decision, selection, and extproc plugin chains. Use when a primary skill touches apiserver, authz, rate-limit, memory, response API, provider adapters, or related runtime service-support packages.

Modifies the local startup chain including image build, container serve/bootstrap logic, and canonical smoke test behavior. Use when changing `vllm-sr serve` behavior, image selection or pull policy, container startup sequences, local Docker/Make bootstrap, or canonical smoke config.

Bridges native skill discovery into the vLLM Semantic Router repository harness, routing tasks through the canonical agent-report flow, repo-local skill registry, and validation commands. Use when starting any task inside the vLLM Semantic Router repository to resolve the correct primary skill, read canonical docs, and run harness validation.

Builds and maintains native Rust/C bindings and FFI layers that connect router-side classifiers and signal evaluation to compiled model runtimes. Use when adding or modifying native model bindings, updating FFI interfaces, or changing how the router calls into compiled classifier code.

Builds Docker images, starts local servers, and runs smoke tests for the CPU-only development environment. Use when validating changes locally on CPU, building CPU container images, or running CPU-specific smoke and E2E tests.

Modifies post-decision plugin hooks that transform requests or responses in the routing pipeline. Use when adding, changing, or debugging plugin-controlled request filtering, response transformation, or middleware processing in the Go router.

Maintains the Python CLI argument schema, parser, validator, merger, and config-translation layer that mirrors the router config contract. Use when modifying CLI argument definitions, updating config validation rules, changing how CLI inputs are merged, or adjusting config translation between CLI and router formats.

Modifies signal extraction and text-understanding runtime code that produces facts consumed by decision logic. Use when adding new signals, changing classifier runtime behavior, or updating how the router extracts features from input text.