# NVIDIA FLARE

> NVIDIA FLARE is a domain-agnostic, open-source, extensible SDK for federated learning and other federated-computing applications, from local development through production deployment.


2025 NVIDIA

## Pages in this subsection

- [Choose an NVFLARE API Path](api_selection.md): NVFLARE offers two high-level ways to build federated applications, plus
- [Client API Usage](client_api_usage.md): The FLARE Client API provides an easy way for users to convert their centralized,
- [HuggingFace Client API](hf_client_api.md): The HuggingFace Client API lets you federate an existing HuggingFace
- [Getting Started with Recipes](job_recipe.md): This task-oriented tutorial shows how to create and run an NVFlare job with a
- [Recipe API Reference](recipe_api.md): This page is the authoritative reference for the stable Recipe APIs you can
- [Available Recipes](available_recipes.md): NVFlare provides a variety of pre-built recipes for common federated learning algorithms and workflo...
- [Collaboration API (Technical Preview)](collab_api.md): The Collaboration API (“Collab API”) is a **Technical Preview** introduced
- [FLARE API](flare_api.md): `FLARE API` is the FLAdminAPI redesigned for a better user experience in version 2.3. It is a
- [Running Flower in NVIDIA FLARE](flower_integration/flower_integration.md): **Run your existing Flower applications in FLARE – no code changes needed.**
- [Initial Integration](flower_integration/flower_initial_integration.md): Architecturally, Flower uses client/server communication. Clients communicate with the server
- [Flower Job Structure](flower_integration/flower_job_structure.md): Even though Flower Programming is out of the scope of FLARE/Flower integration, you need to have a g...
- [Run Flower Application as FLARE Job](flower_integration/flower_run_as_flare_job.md): Before running Flower applications with FLARE, you must have both FLARE and Flower frameworks
- [FLARE Multi-Job Architecture](flower_integration/flare_multi_job_architecture.md): To maximize the utilization of compute resources, FLARE supports multiple jobs running at the
- [Detailed Design](flower_integration/flower_detailed_design.md): Flower uses gRPC as the communication protocol. To use FLARE  as the communicator, we route Flower’s...
- [Reliable Messaging](flower_integration/flower_reliable_messaging.md): The interaction between the FLARE Clients and Server is through reliable messaging.
- [Federated XGBoost with NVFlare](federated_xgboost/federated_xgboost.md): XGBoost (https://github.com/dmlc/xgboost) is an open-source project that
- [Federated Learning for XGBoost](federated_xgboost/secure_xgboost_user_guide.md): This guide demonstrates how to use NVIDIA FLARE (NVFlare) to train XGBoost models in a federated lea...
- [Reliable Federated XGBoost Design](federated_xgboost/reliable_xgboost_design.md): NVFLARE serves as a launchpad to start the XGBoost system.
- [Reliable Federated XGBoost Timeout Mechanism](federated_xgboost/reliable_xgboost_timeout.md): NVFlare introduces a tightly-coupled integration between XGBoost and NVFlare.
- [Secure Federated XGBoost Design](federated_xgboost/secure_xgboost_design.md): For horizontal XGBoost, each party holds “equal status” - whole feature and label for partial popula...
- [Data Preparation & Heterogeneity](data_preparation.md): Data preparation in federated learning differs from centralized ML because:
- [POC: Prove of Concept: Simulate Production deployment locally](poc.md): To get started with a proof of concept (POC) setup after Installation, run this command to generate ...
- [Data Scientist Guide](index.md): * Choose an NVFLARE API Path
- [Recipe Metrics Artifacts](recipe_metrics_artifacts.md): Built-in training aggregation recipes write standard metrics artifacts when the

## Optional

- [Top-level llms.txt](../../llms.txt): Complete documentation index.