Architecture
This page provides an overview of the LlamaCloud architecture.
Overview
Section titled âOverviewâSelf-hosted LlamaCloud is an Enterprise-only feature, designed specifically to meet the needs of organizations that require a high degree of control over their data and infrastructure. Please contact us at https://www.llamaindex.ai/contact if youâre interested in learning more about self-hosting.
The following diagram shows the architecture of LlamaCloud:
Databases, Queues, and File Stores
Section titled âDatabases, Queues, and File Storesâ- Postgres: LlamaCloud uses Postgres as its primary, relational database for almost everything.
- MongoDB: LlamaCloud uses MongoDB as its secondary, document database for storing data around document ingestion and pipelines.
- Redis: LlamaCloud uses Redis as its primary, in-memory key-value store for queuing and caching operations.
- RabbitMQ: LlamaCloud uses RabbitMQ as its message queue. We leverage a series of queues to manage large-scale data processing jobs.
- S3Proxy: To support non-S3 object storage options, we allow users to deploy s3proxy, an S3-compatible proxy, to interact with other storage solutions.
Internal Services
Section titled âInternal ServicesâLlamaCloud Frontend
Section titled âLlamaCloud FrontendâThe frontend is the main user interface for LlamaCloud. We recommend exposing it through a reverse proxy like Nginx or Traefik for users to connect to in production.
LlamaCloud Backend
Section titled âLlamaCloud BackendâThis is the API entrypoint for LlamaCloud. It handles all requests from the frontend and the business logic of our platform. This service can also be used as a standalone API.
LlamaCloud Jobs Service
Section titled âLlamaCloud Jobs ServiceâThe jobs service is responsible for managing job processing and ingestion pipelines.
LlamaCloud Jobs Worker
Section titled âLlamaCloud Jobs WorkerâThe jobs worker works with the jobs service to process and ingest data.
LlamaCloud Usage Service
Section titled âLlamaCloud Usage ServiceâThis service tracks all parsing and ingestion usage across projects, indexes, and organizations.
LlamaParse Service
Section titled âLlamaParse ServiceâLlamaParse is the engine that powers LlamaCloudâs unstructured document parsing. It supports a variety of file formats, parsing modes, and output formats. For more information, please refer to the LlamaParse documentation.
LlamaParse OCR Service
Section titled âLlamaParse OCR ServiceâThis service works hand-in-hand with LlamaParse to increase the accuracy of our document parsing.