@llamaindex/llama-cloud

Manages document upload, parsing, and indexing through Llama Cloud's managed infrastructure. The SDK provides client-side abstractions that handle document chunking, embedding generation, and vector storage on remote servers, eliminating the need for local infrastructure while maintaining TypeScript-native integration patterns for file handling and progress tracking.

Executes vector similarity search queries against documents indexed in Llama Cloud, translating natural language queries into embeddings and retrieving semantically relevant chunks. The SDK handles query embedding generation server-side and returns ranked results with relevance scores, abstracting the vector database mechanics behind a simple query interface.

Supports updating indexed documents and maintaining version history in Llama Cloud, allowing developers to modify document content and metadata while preserving previous versions. The SDK abstracts versioning mechanics, handling version tracking and retrieval without exposing underlying version control implementation.

Abstracts vector database operations by storing embeddings in Llama Cloud's managed infrastructure, automatically generating embeddings for indexed documents using Llama Cloud's default embedding model. The SDK provides CRUD operations for document collections without exposing vector database implementation details, handling embedding generation, storage, and retrieval transparently.

Provides CRUD operations for managing document collections in Llama Cloud, including creation, deletion, listing, and metadata updates. The SDK abstracts collection lifecycle through a fluent API that handles remote state synchronization, allowing developers to organize documents into logical collections and manage their indexing status without direct API calls.

Handles large document uploads through streaming APIs that report ingestion progress in real-time, allowing developers to monitor document processing without blocking on completion. The SDK abstracts streaming mechanics and provides callbacks or event emitters for progress updates, enabling responsive UIs and graceful error handling during long-running ingestion operations.

Provides a fully typed TypeScript client library for the Llama Cloud API, with compile-time type checking for all requests and responses. The SDK uses TypeScript generics and discriminated unions to model Llama Cloud's API surface, enabling IDE autocomplete, type inference, and compile-time error detection without runtime validation overhead.

Handles Llama Cloud API authentication through credential management abstractions, supporting API key-based authentication with environment variable loading and credential validation. The SDK abstracts authentication mechanics, allowing developers to configure credentials once and use them across all API operations without manual token management.

Implements automatic retry logic for transient failures and provides structured error handling for Llama Cloud API errors. The SDK abstracts retry strategies (exponential backoff, jitter) and error classification, allowing developers to handle different error types (rate limits, network errors, validation errors) with appropriate recovery strategies without manual retry implementation.

Supports batch ingestion and retrieval of multiple documents in a single operation, reducing API call overhead and improving throughput for bulk operations. The SDK abstracts batch mechanics, handling request batching, result aggregation, and partial failure scenarios without exposing underlying batch API details.

Enables filtering and querying documents by metadata attributes (tags, timestamps, custom fields) during search and retrieval operations. The SDK provides a query builder or filter DSL that translates metadata filters into Llama Cloud API queries, allowing developers to narrow search results without post-processing.