llamacloud

Guided LlamaCloud onboarding — account, API key, region choice, SDK install, and a first-call smoke test.

Router for the LlamaCloud / LlamaIndex plugin — describe what you're building and get pointed to the right product, skill, and next step.

Use this agent when the user wants to find existing LlamaIndex/LlamaParse examples or cookbooks that match a scenario. Typical triggers include "is there an example for building RAG over X", "find a cookbook for this extraction use case", "what notebooks cover agentic workflows", and "show me a sample close to what I'm building". See "When to invoke" in the agent body for worked scenarios. This agent is read-only — it finds and summarizes recipes, it does not write application code.

Use this agent when the user wants to design, draft, or validate a LlamaExtract extraction schema from real documents. Typical triggers include "design an extraction schema for these invoices", "what schema should I use to pull fields from this PDF", "review/validate my LlamaExtract Pydantic schema", and "turn this sample document into an extraction schema". See "When to invoke" in the agent body for worked scenarios. Do not use for raw text parsing (that is Parse) or for running production extraction jobs.

Use this agent when the user wants a designed RAG or retrieval pipeline and needs the architecture decided. Typical triggers include "design a RAG pipeline for my docs", "should I use Cloud Index or build retrieval myself", "choose chunking/embedding/retrieval settings for my corpus", and "architect retrieval for this app". See "When to invoke" in the agent body for worked scenarios. Do not use to run ingestion jobs or to write production app code end to end.

Use this agent when a self-hosted / BYOC LlamaCloud deployment is misbehaving and needs diagnosis. Typical triggers include "my self-hosted LlamaParse pods are crashing", "diagnose my BYOC LlamaCloud Helm deployment", "why is my self-hosted LlamaCloud failing to start", and "check my LlamaCloud Kubernetes setup". See "When to invoke" in the agent body for worked scenarios. This agent is strictly read-only — it diagnoses and recommends, never mutates the cluster.

Use this agent when the user wants to scaffold a LlamaIndex Workflow or an agent deployment from a description. Typical triggers include "scaffold a LlamaIndex Workflow that does X", "build me an event-driven agent workflow", "turn this process into a Workflow", and "set up a workflow I can deploy". See "When to invoke" in the agent body for worked scenarios. Use for authoring/scaffolding workflow code; defer pure architecture decisions to rag-architect and deployment operations to the llamaindex-deploy skill.

This skill should be used when the user wants to "use LiteParse", "parse locally without an API key", do "offline/OSS document parsing", or work with the "LiteParse library/CLI/server/WASM". Owns local/OSS no-key document parsing and the LiteParse-vs-cloud-LlamaParse decision; defers cloud/agentic/scaled parsing, typed extraction, and managed RAG to sibling skills.

This skill should be used when the user wants to "classify documents", "route documents by type", define "document classification rules", or "categorize incoming files" with LlamaCloud Classify. Owns natural-language classification and intake routing; defers segmenting one file into sub-docs to llamacloud-split, typed field extraction to llamacloud-extract, and raw text to llamacloud-parse.

This skill should be used when the user wants to "extract structured data from a document", "design/validate an extraction schema", write a "Pydantic/JSON/Zod extract schema", configure "LlamaExtract options/citations", or "pull typed fields from PDFs". Owns LlamaExtract v2 schema design and configuration judgment; defers raw text/markdown parsing, RAG/retrieval, and account/key/region setup to sibling skills.

This skill should be used when the user wants to "build a Cloud Index", set up a "managed RAG pipeline on LlamaCloud", configure "hybrid search / rerank / metadata filter on LlamaCloud", "ingest documents into a LlamaCloud index", or expose a "managed retrieval endpoint". Owns the end-to-end managed Cloud Index (source → parse/transform → chunk → embed → sink) and retrieval tuning; defers OSS/self-managed RAG to `llamaindex-framework`, raw parsing to `llamacloud-parse`, typed extraction to `llamacloud-extract`, and account/key setup to `llamacloud`.

This skill should be used when the user wants to "parse a PDF/document", "use LlamaParse", choose between "LlamaParse tiers / agentic vs fast vs cost", "configure parse output (markdown/JSON)", or "retrieve/parse results". Owns LlamaParse v2 (GA) parsing decisions and config; defers typed-field extraction, managed retrieval, and local no-key parsing to sibling skills.