sglang

Plugin sglang from marketplace air-gapped-marketplace in air-gapped/skills. Commands read from the repo's manifest — 3 days ago.

Skills (77)

• aiperf

  NVIDIA AIPerf — vendor-neutral generative-AI inference benchmarking (genai-perf successor). `aiperf profile`: concurrency / request-rate / fixed-schedule trace replay / user-centric / multi-run confidence, 19 endpoint types (chat, completions, embeddings, rankings, responses, messages, audio transcr

• airgap-vetting

  Vet an open-source product for air-gap readiness BEFORE adoption. Answers eight questions — telemetry, does-opt-out-actually-work, proxy-in-disguise over a hosted API, runtime downloads, custom-CA support, feature-level content egress / offline degradation, day-two sustainment (feed mirroring, stale

• analytical-troubleshooting

  Structured live troubleshooting for deviation problems: something worked (or should work) and now doesn't, and the cause is unknown. Staged method — quick likelihood-weighted tests first, escalating to a comparative IS/IS-NOT specification with distinctions-and-changes analysis when quick tests fail

• ansible-idrac-9-10

  Run and debug `dellemc.openmanage` Ansible playbooks against Dell PowerEdge **iDRAC 9** (14G–16G) and **iDRAC 10** (17G — R670, R770, R870, R970, XE9780, XE9785). Covers the iDRAC 10 / iDRAC 9 ≥ 7.30.10.50 `BasicAuthState: Unadvertised` default that silently 401s `ansible.builtin.uri` (Dell KB 00043

• argo-cd-apps

  Author and maintain Argo CD `Application` and `ApplicationSet` manifests as a GitOps consumer (publisher), targeting Argo CD v3.4 / v3.5, v3.6 in RC (Oct 2026). Covers source types (Helm, Kustomize, OCI, multi-source, plugin), sync policies + options + waves + hooks, ApplicationSet generators (List,

• autoresearch

  Autonomous experiment loops that hill-climb a measurable metric — apply one change, measure, keep it only if the number improved, revert if not, repeat unattended. Also deep multi-perspective research producing a saved report, and research-then-optimize when no metric exists yet. Use whenever work s

• baml-expert

  BAML (Boundary ML) expert for projects defining LLM calls as typed functions in .baml files with a generated Python client. Use whenever the repo contains baml_src/, baml_client/, baml-cli commands, or imports from baml_py / baml_client. Covers .baml syntax (function, class, enum, client, test, retr

• ceph-performance

  Tune and benchmark a healthy Ceph cluster, with Rook-on-Kubernetes specifics: mClock profiles and the NVMe capacity-fallback trap (osd_mclock_max_capacity_iops), osd_memory_target from pod requests, BlueStore allocation and cache defaults, PG autoscaler targets (mon_target_pg_per_osd, bulk, target_s

• ceph-s3

  Run Ceph RGW as an S3 endpoint under Rook (CephObjectStore, CephObjectStoreUser, ObjectBucketClaim, COSI, CephBucketTopic, multisite realms/zones). Core knowledge: the 19.2.6 / 20.2.4 RGW security fixes and their fallout (SigV4 hardening rejecting presigned PUTs and RGW's own multisite forwarding; r

• ceph-troubleshooting

  Diagnose and recover an unhealthy Ceph cluster, with Rook-on-Kubernetes specifics: slow or stuck recovery/backfill under the mClock scheduler (why osd_max_backfills "does nothing"), nearfull/backfillfull/full, inconsistent/unfound/incomplete/down PGs, BLUESTORE_SLOW_OP_ALERT vs SLOW_OPS, OSD OOM kil

• chat-completions-api

  Reference for the OpenAI Chat Completions API (/v1/chat/completions) and legacy /v1/completions as the lingua-franca compatibility protocol — the official spec incl. deprecation timeline and Responses-only feature delta, how 7 local servers (vLLM, SGLang, llama.cpp, Ollama, mistral.rs, Llama Stack/O

• claude-code-ci-harness

  Run Claude Code headlessly as an isolated CI harness against any Anthropic Messages-compatible endpoint (local engine, gateway or hosted API) or a Claude subscription token. Covers credential isolation, stream-json transcripts, tool-call assertions, multi-turn scenarios, and endpoint/model compatibi

• confluence-best-practices

  Advise on USING Confluence well, not operating it: make the structural call — is this a space, a page, or a child page? — diagnose why a wiki is a dread (can't find anything, content rots, duplicates, hidden by permissions, unreadable), and recommend the lean fix. Built FIRST for an agent that ACTS

• gitlab-best-practices

  Operate self-managed GitLab on the official Helm chart — multi-hop upgrade campaigns across required stops, the chart-10 wall (bundled PostgreSQL, Redis and MinIO removed and now mandatory), PostgreSQL 17 being the only major both GitLab 18.x and 19.x accept, default flips invisible to a values diff

• gpu-host-tuning

  Audit AND tune Linux/GPU inference hosts — read-only host snapshot

• harvester-upgrade

  Plan and run a controlled, COMMUNITY-edition Harvester HCI upgrade off an EOL line up to latest stable — the no-skip minor ladder (1.5→1.6→1.7→1.8→1.9; embedded RKE2/KubeVirt/Longhorn/SLE-Micro ride along), gated at each hop on first upgrading the EXTERNAL Rancher + a matching Harvester UI-extension

• helm

  This skill should be used for Helm chart work in either direction. Authoring: creating charts, templates and _helpers.tpl, values.yaml patterns, Chart.yaml, values.schema.json, helm-docs, library charts, chart CI/CD and signing. Consuming: carrying a values file across chart versions with a three-wa

• jinja-expert

  Author, read, and debug Jinja2 templates across the three places Jinja lives in 2026 — HuggingFace `chat_template.jinja` (rendered by `apply_chat_template` for vLLM / sglang), Ansible playbooks + `.j2` files, and Jinja-adjacent Kubernetes workflows (`values.yaml.j2`, `kubernetes.core.k8s + template`

• jira-best-practices

  Advise on USING Jira well, not operating it: make the structural call — is this an epic, a story, a task, or a sub-task? — and diagnose why a Jira is a dread, then recommend the lean fix. Adapt to the organisation's OWN hierarchy names, conventions, and working language instead of imposing a methodo

• jira-cli

  Drive Atlassian Jira from the terminal with the `jira` CLI (jira-cli, v1.7.0) against ANY Jira — Cloud or on-premise/Data Center. Covers the full command surface (issue / epic / sprint / board / project / release), the non-interactive automation contract (`--no-input` + `--plain`/`--raw`/`--csv` for

• jira-confluence-mcp

  Install, configure, secure, and troubleshoot the mcp-atlassian MCP server (sooperset/mcp-atlassian) that connects an agent to Jira/Confluence — including AIR-GAPPED setup (mirror the prebuilt image by digest; no PyPI/git mirror) and internal-CA / TLS handling (mount the CA vs JIRA_SSL_VERIFY=false).

• k8s-components-checker

  Survey an RKE2 community cluster against an embedded compatibility registry of 20 stack components and produce a verdict for upgrade-readiness, drift-review, and version-skew questions. Components: RKE2, Rancher, Harvester, Cilium, Tetragon, cert-manager, Kyverno, KEDA, Argo CD, Harbor, Traefik, Roo

• keda

  Configure, operate, and master KEDA (Kubernetes Event-driven Autoscaling) — ScaledObject, ScaledJob, TriggerAuthentication CRDs, 80+ scalers, HPA behavior tuning, scale-to-zero, the KEDA HTTP Add-on, production hardening, multi-trigger semantics, scalingModifiers formulas, GitOps integration, and tr

• keycloak-iam

  Operate, configure, deploy, secure, and integrate with Keycloak (open-source IAM) — the modern Quarkus distribution (24.x–26.8.x), the Keycloak Operator with `Keycloak` and `KeycloakRealmImport` CRDs, realm/client/identity-provider configuration, plus legacy 16.x–23.x instances (incl. WildFly "-lega

• litellm-api

  Administer a LiteLLM proxy via its management REST API (keys, teams, users, orgs, models, budgets, spend) — a 447-path control plane where ~63% of paths are undocumented, 91 are hidden from /openapi.json, and the OpenAPI schema documents wrong field names. Centerpiece: four "empty means UNLIMITED" f

• litellm-valkey

  Run LiteLLM proxy multi-pod with Redis/Valkey (standalone, Sentinel, or Cluster) so rate limits, budgets, and router state are actually shared across pods. Centerpiece: LiteLLM's dominant failure mode is SILENT fallback from Redis to per-pod in-memory enforcement (≥6 independent code paths, log-leve

• lmcache-mp

  LMCache multiprocess (MP) mode — standalone LMCache server in its own pod/process that vLLM connects to over ZMQ. Gives process isolation, no GIL contention on the inference path, one cache shared by multiple vLLM pods per node, and CPU-memory scaling independent of GPU memory. Covers the `LMCacheMP

• logging-operator

  Configure and operate the kube-logging logging-operator (formerly Banzai Cloud) on Kubernetes — the CRD-driven log pipeline: Fluent Bit collector → fluentd or syslog-ng aggregator → outputs. Covers the 16-CRD model (Logging, Flow/ClusterFlow, Output/ClusterOutput, FluentbitAgent, SyslogNG*, LoggingR

• makefile-best-practices

  Makefile best practices, patterns, and templates for GNU Make 4.x — dependency graphs, task-runner workflows, parallel-safe recipes, self-documenting help targets, and language-specific patterns (Go, Python, Node, Docker, Helm, POSIX).

• messages-api

  Reference for the Anthropic Messages API (/v1/messages) as a third-party compatibility protocol — the 7 inference servers that implement it natively (vLLM, SGLang, llama.cpp, Ollama, mistral.rs, Llama Stack/OGX, Lemonade), gateways that adapt it (LiteLLM, Bifrost, Superagent Gateway), client behavio

• mimir-upgrade

  Plan and run a controlled, COMMUNITY-edition Grafana Mimir upgrade on the `mimir-distributed` Helm chart, air-gap first — the chart↔app co-pinned ladder (5.7→5.8→6.0.6→6.1.0 = app 2.16→2.17→3.0.4→3.1.2), the classic-vs-ingest-storage decision (the chart ships a supported `classic-architecture.yaml`;

• netbox-best-practices

  NetBox 4.2-4.7 deployment and upgrade knowledge the official netboxlabs/skills marketplace does not cover — deploying or upgrading NetBox on Kubernetes with the netbox-community helm chart (netbox-chart), external PostgreSQL/valkey wiring, API token bootstrap on 4.5+ (nbt_ v2 tokens), plugin install

• nvidia-datacenter-bringup

  Bring up NVIDIA HGX/DGX datacenter GPU hosts on Ubuntu 24.04 LTS — air-gapped or connected, Secure Boot enabled. Covers B300/B200/H100/A100/L40S/L4 driver+fabricmanager+NVLSM+DOCA-OFED install order and exact package set from NVIDIA CUDA repo + DOCA repo. Triggers on B300/B200/HGX/DGX install, "fabr

• nvidia-nixl

  NVIDIA Inference Xfer Library (NIXL) operator + developer reference. Point-to-point KV-cache and tensor transport for distributed inference (Dynamo, vLLM, SGLang). Covers the agent API (full Python reference; C++/Rust via upstream pointers), all 15 backend plugins (UCX, GDS, GDS_MT, libfabric, moonc

• open-webui-api

  Administer Open WebUI entirely via its REST API (v0.11+): user/group lifecycle, permissions, model catalog GitOps (export/import/sync), knowledge/RAG pipelines, config-as-code, SCIM provisioning, event webhooks, and backup surfaces. Grounded in the v0.11.0 source — covers the 476-path surface the of

• open-webui-embeddings

  Choose and wire open-weights embedding + reranker models into Open WebUI's RAG pipeline, across TEI, vLLM, OpenRouter or LiteLLM. Covers model selection on current MTEB data (quality, zero-shot contamination, and a FLOPs-based throughput model, since no leaderboard publishes speed), the query/docume

• open-webui-valkey-websocket

  Deploy Open WebUI multi-pod with WebSockets and Valkey/Redis Sentinel at 1000+ user scale on Kubernetes. Centerpiece is the structural Socket.IO+Redis frame-amplification bug (#23733) that crippled multi-pod streaming before v0.11.1, and the maintainer-endorsed mitigation (`CHAT_RESPONSE_STREAM_DELT

• opencode-ci-harness

  Run opencode (v2) CLI as a headless, fully-isolated test harness against any OpenAI-, Responses- or Messages-compatible endpoint (local inference engine or remote API), or against any provider credential already stored in an opencode profile (API key or login). Drives multi-step tool-calling scenari

• openshell-best-practices

  Run NVIDIA OpenShell 0.1.x — the sandboxed runtime for autonomous AI agents (gateway, per-sandbox supervisor, Landlock/seccomp/netns isolation, L7 network policy, credential placeholders via providers/profiles) — the way current practitioners do. Core knowledge: the 0.1.0 breaking cutover and which

• openshift-app

  Package applications for OpenShift deployment: container images (UBI, arbitrary UID, multi-stage builds), packaging formats (Helm, Kustomize, Operators, OLM v1), CI/CD (Tekton, ArgoCD, Shipwright, Conforma), security (SCC, PSA, supply chain, image signing, secrets), operations (Routes, probes, scali

• patch

  [4/4 defending-code] Generate candidate fixes for verified security findings. Last step of the find-and-fix loop (/threat-model -> /vuln-scan -> /triage -> /patch); run /triage first so fixes aren't written for false positives. Consumes TRIAGE.json (preferred), VULN-FINDINGS.json, or an execution-ha

• postgres-operator-best-practices

  Run, configure, upgrade and troubleshoot the Zalando postgres-operator (acid.zalan.do `postgresql` CRs, Spilo, Patroni, WAL-G) on Kubernetes. Core knowledge: the v1.x -> v2.x major upgrade and its three changed defaults that bite silently (`kubernetes_use_configmaps` true, spilo-17 -> spilo-18, scra

• postgres-operator-cloudnative-pg-migration

  Migrate PostgreSQL clusters from the Zalando postgres-operator (acid.zalan.do `postgresql` CRs, Spilo/Patroni, WAL-G) to CloudNativePG (CNPG) on Kubernetes, incl. fully air-gapped clusters. Core knowledge: the two walls (Spilo↔CNPG glibc/collation divergence that corrupts physically-copied indexes;

• prometheus-mimir-grafana

  Query Prometheus and Grafana Mimir, write and debug PromQL, and build or fix Grafana dashboards — for agents solving problems from metrics. Covers the Prometheus HTTP API (`/api/v1/query`, `query_range`, `series`, `labels`, `metadata`), Mimir multi-tenancy (`X-Scope-OrgID`, federation `a|b|c`, per-t

• rancher-logging-exit

  Migrate off the Rancher-bundled `rancher-logging` chart (cattle-logging-system, rancher/mirrored-kube-logging-* images) to the upstream kube-logging logging-operator ≥6.7.0 — air-gap-first. Rancher 2.11 through 2.16-dev all bundle a frozen operator 4.10.0 that is inside the affected range of CVE-202

• rancher-upgrade

  Plan and sequence COMMUNITY-edition Rancher upgrades across air-gapped multi-cluster fleets — a management/"hosting" Rancher cluster plus the downstream RKE2/K3s clusters it provisions. Covers the community release model (2.11→2.15, community-vs-Prime cadence, EOL), the Kontainer Driver Metadata (KD

• redis-to-valkey

  Migrate Redis deployments (especially Bitnami Redis Helm charts in Sentinel HA mode) to Valkey on Kubernetes, including fully air-gapped clusters. Core knowledge: the RDB-version wall (Valkey replicates/loads only from Redis ≤ 7.2.x; Redis 7.4+ writes RDB v12 which Valkey rejects; Valkey 9 writes it

• responses-api

  Reference for the OpenAI Responses API (/v1/responses), OpenResponses open standard, and Codex CLI. Covers the request/response schema, previous_response_id, Conversations API, server-side compaction, WebSocket transport, hosted Shell tool, Skills, tool_search, MCP connectors, prompt caching, phase

• rook-ceph-best-practices

  Operate, configure and upgrade Rook-managed Ceph on Kubernetes (CephCluster CR; rook-ceph, rook-ceph-cluster and ceph-csi-drivers charts; ceph-csi-operator). Core knowledge: the Rook -> Ceph -> key-rotation upgrade order and its version gates; the CVE-2025-30156 fix (new AES256K CephX key type, daem

• secure-boot-cert-rotation

  Triage and fix the Microsoft Secure Boot 2011→2023 UEFI certificate rotation across Dell PowerEdge / iDRAC9 bare metal, Ubuntu/Linux servers, and Harvester HCI / KubeVirt guest VMs. All three 2011 CAs expired (two in June 2026, Windows Production PCA 2011 on 2026-10-19) — but UEFI firmware ignores c

• sglang-hicache

  SGLang HiCache (hierarchical KV cache) — three-tier prefix cache: GPU HBM (L1) → pinned host DRAM (L2) → distributed L3 (Mooncake / 3FS / NIXL / AIBrix / EIC / SiMM / file / LMCache). Covers `--enable-hierarchical-cache`, all `--hicache-*` flags, write policies, page_first* layouts, prefetch policy

• sglang-model-gateway

  SGLang Model Gateway (`sgl-model-gateway`, formerly `sgl-router`) — Rust router fronting vLLM and SGLang inference workers on Kubernetes. Covers first-class vLLM gRPC backend plus HTTP transparent-proxy for vanilla vLLM, the policy set (6-8 `--policy` values by entrypoint, `cache_aware` default), to

• skill-improver

  Improve, test and maintain Claude Code skills. Modes: `fix` runs a skill's eval cases and fixes what the failed answers show is missing; `improve` hill-climbs a 10-dimension quality rubric with blind scoring and an A/B comparator; `score` rates a skill without editing it; `freshen` verifies the skil

• snmp-exporter

  Best practices for Prometheus snmp_exporter (v0.30.x): writing generator.yml modules, curating MIB walks, SNMPv2c/v3 auth, timeout tuning, Kubernetes deployment (Probe/ScrapeConfig CRDs, secrets, UDP egress), local docker testing, and debugging failed scrapes. Includes worked device references for D

• threat-model

  [1/4 defending-code] Build a threat model for a target codebase. Optional first step of the find-and-fix loop (/threat-model -> /vuln-scan -> /triage -> /patch); its THREAT_MODEL.md scopes and severity-calibrates every later step, but /vuln-scan runs without it. Three modes: "interview" walks an app

• traefik-hardening

  Harden a Traefik 2.x/3.x reverse proxy against abusive or unwanted traffic — per-identity rate + concurrency limiting, IP allowlisting, request-body buffering, middleware chaining, keying limits on a JWT claim/header instead of raw source IP, air-gapped plugin loading (localPlugins/WASM), and detect

• transformers-config-tokenizers-expert

  Preflight reference for HuggingFace snapshots — what vLLM, sglang, and transformers.generate see at runtime. Covers config-file precedence (tokenizer.json, tokenizer_config.json, generation_config.json, chat_template.jinja), transformers v5 tokenizer-class taxonomy (TokenizersBackend, PythonBackend,

• triage

  [3/4 defending-code] Triage a batch of raw security findings. Third step of the find-and-fix loop (/threat-model -> /vuln-scan -> /triage -> /patch); consumes /vuln-scan's VULN-FINDINGS.json, but accepts any scanner output, so it also stands alone on a third-party backlog. Verify each is real, colla

• ubuntu-autoinstall

  Author, validate, and debug Ubuntu Server autoinstall configuration (the Subiquity installer's `autoinstall:` schema, version 1) for Ubuntu Server LTS 24.04 and 26.04, focused on unattended on-premise and air-gapped installs — identity, storage (LVM/direct/ZFS/encryption/RAID), apt mirror-selection

• ubuntu-cloud-init

  Author, validate, and debug cloud-init configuration for Ubuntu Server LTS 24.04 and 26.04, focused on on-premise and air-gapped hosts via the NoCloud datasource — `#cloud-config` user-data, the cloud-config modules (users, ssh, write_files, runcmd, apt with local mirrors, ca_certs for internal CAs,

• ubuntu-netplan

  Author, validate, and debug netplan network configuration (`/etc/netplan/*.yaml`) for Ubuntu Server LTS 24.04 and 26.04, focused on on-premise and air-gapped hosts — static addressing, bonds, bridges, VLANs, VRFs, routing/policy-routing, DNS, interface matching/renaming, and the systemd-networkd ren

• vllm-benchmarking

  Run production vLLM benchmarks — `vllm bench` (serve, throughput, latency, sweep, startup, mm-processor, forward), request-rate vs max-concurrency semantics, TTFT/TPOT/ITL/E2EL percentiles, goodput SLO measurement, prefix-cache workloads, air-gapped operation (HF_ENDPOINT, ModelScope, hf-mirror, off

• vllm-caching

  vLLM tiered KV cache configuration for production H100/H200 clusters. Native CPU offload, LMCache (CPU+NVMe+GDS), NixlConnector (disaggregated prefill), MooncakeConnector (RDMA), MultiConnector composition. Version gates, sizing math (flag total across TP, not per-GPU — opposite of SGLang), KV-vs-we

• vllm-chat-templates

  vLLM chat-template (prompt-side Jinja) operator reference. Template resolution precedence (`--chat-template` → AutoProcessor → tokenizer default → bundled fallback), `chat_template_kwargs` allowlist silently dropping `add_generation_prompt`/`enable_thinking`/custom kwargs (PR 27622 fix), 27 shipped

• vllm-configuration

  Configure vLLM completely — YAML config file format, CLI arg precedence, full VLLM_*/HF_*/TRANSFORMERS_* env-var catalog, end-to-end recipe for air-gapped environments (internal HF mirrors, hf-mirror.com, ModelScope, HF_HUB_OFFLINE with pre-seeded cache, gated models offline, trust_remote_code suppl

• vllm-deployment

  Use this skill when authoring, reviewing, or fixing a vLLM Kubernetes manifest, Docker/Podman pod, or OpenShift ServingRuntime — even when the user does not say "vllm". Triggers on: lab cluster performance practices, cache mount + survival across pod restarts (/root/.cache, VLLM_CACHE_ROOT, TORCHIND

• vllm-gemma-4-31b

  Operating-point reference for serving Gemma 4 31B on vLLM — TP sizing, max_model_len, max_num_seqs, gpu_memory_utilization, kv_cache_dtype, EAGLE3 spec-dec, chat_template choice. When the user mentions Gemma 4 31B in the context of vLLM deployment, tuning, or performance.

• vllm-input-modalities

  vLLM non-chat inference surfaces — text embeddings (`/v1/embeddings`, `/v2/embed`), reranking/scoring (`/rerank`, `/score`), speech-to-text (`/v1/audio/transcriptions`, `/v1/audio/translations`), document OCR via VLMs. Covers 2026 `--runner pooling` (replacing `--task embed`), v0.20 deprecations (`s

• vllm-nvidia-hardware

  NVIDIA AI-hardware + vLLM-platform reference covering Hopper (H100/H200), Blackwell (B100/B200/B300) and Blackwell Ultra, Grace-Blackwell superchips and NVL72 racks (GB200, GB300), Vera Rubin (R100/R300) with VR200 NVL144 and Kyber NVL576, Dell PowerEdge XE family and IR5000/IR7000/IR9048 racks. Per

• vllm-observability

  Observe production vLLM from the data it already emits — the V1 `/metrics` Prometheus surface, SLO alerts on TTFT/ITL/queue/KV/preemption/aborts/corrupted logits, the shipped Grafana dashboards, OTLP tracing (`--otlp-traces-endpoint`, `--collect-detailed-traces`), DCGM pairing (`SM_OCCUPANCY`, not `

• vllm-omni

  vLLM-Omni output-side multimodal generation — image (FLUX.1/2, Qwen-Image, GLM-Image, BAGEL, SD3.5, HunyuanImage-3.0), video (Wan2.1/2.2, LTX-2, HunyuanVideo-1.5), TTS (Qwen3-TTS, CosyVoice3, Voxtral-TTS), any-to-any omni (Qwen3-Omni, Qwen2.5-Omni, MiMo-Audio) via `vllm serve --omni`. Stage-based di

• vllm-performance-tuning

  vLLM performance-tuning operator reference — tuning workflow (baseline → bottleneck → knob → re-bench), fused-MoE kernel autotune (`benchmark_moe.py` generates `E=N,N=M,device_name=X.json` configs), DeepEP all-to-all + expert parallelism + EPLB, CUDA graph modes (FULL_AND_PIECEWISE default), torch.c

• vllm-quantization

  vLLM datacenter-GPU quantization — picking, configuring, troubleshooting NVFP4, FP8, MXFP4, MXFP8, AWQ, GPTQ, INT8, compressed-tensors, modelopt, quark on H100/H200/B200/B300/GB200/GB300; every `--quantization` flag value, KV-cache dtypes (fp8_e4m3, nvfp4, per-token-head, turboquant), MoE backend se

• vllm-reasoning-parsers

  vLLM reasoning-parser operator + developer reference. `--reasoning-parser` CLI wiring, `ReasoningParser` contract (non-streaming `extract_reasoning` + per-delta `extract_reasoning_streaming`), `is_reasoning_end` xgrammar gating, `--structured-outputs-config.enable_in_reasoning` bypass, 42 built-in p

• vllm-speculative-decoding

  Pick, configure, tune, monitor vLLM speculative decoding in production. Eleven SpeculativeMethod options at v0.33.0 (ngram, ngram_gpu, draft_model, suffix, custom_class, eagle, eagle3, dflash, dspark, mtp, extract_hidden_states; medusa and mlp_speculator removed), `--speculative-config` JSON schema,

• vllm-tool-parsers

  vLLM tool-calling operator reference — picking `--tool-call-parser` per model family, writing custom parsers via `--tool-parser-plugin`, navigating vLLM source + GitHub tracker to debug any specific tool-call question. Pointer map, not source paraphrase. All 40+ built-in parsers (JSON-sentinel, pyth

• vuln-scan

  [2/4 defending-code] Static source-code vulnerability scan. Second step of the find-and-fix loop (/threat-model -> /vuln-scan -> /triage -> /patch), and the usual entry point when no threat model exists. Reads a target directory (and THREAT_MODEL.md if present), spawns parallel review subagents per

Similar plugins