Awesome GitHub RepositoriesDurable Execution Persistence

Conductor is a durable workflow engine designed to orchestrate complex, long-running business processes and autonomous agent loops. It functions as a stateful execution platform that persists the entire history of a process, ensuring that workflows remain reliable and recoverable across infrastructure failures, system restarts, and transient network errors. By managing task lifecycles, worker polling, and state transitions, it provides a centralized coordination layer for distributed systems. The platform distinguishes itself through its specialized support for AI agent orchestration, allowin

Conductor is a distributed workflow manager and microservices orchestration engine. It coordinates long-running business processes across a network of services, using a stateful orchestration model to ensure consistent execution order and reliable state management. The system defines task sequences and conditional business logic using structured JSON configuration files. It manages the execution of complex operations across distributed systems with built-in persistence and failure recovery, tracking the progress of processes through a backend database. The platform provides capabilities for

Cadence is a distributed workflow orchestration engine designed to execute long-running, asynchronous business logic with built-in durability and resilience across distributed systems. It functions as a stateful process manager that ensures processes resume from their last known state following system crashes or network outages. The platform utilizes a distributed task queue to manage work across independent worker nodes and supports persistence via SQL or Cassandra backend storage. It includes a workflow visualization dashboard for inspecting execution histories and state traces, alongside a

This project is a comprehensive framework for the orchestration, evaluation, and context management of large language model agents. It provides a set of architectural patterns and standards for designing agent interactions, integrating external tools, and establishing memory architectures to persist knowledge across sessions. The system focuses on optimizing the limited memory of language models through token-aware context compression and filesystem-based context offloading. It incorporates secure execution environments using sandboxed virtual machines and isolated containers to safely run ba

Flyte is a distributed machine learning pipeline manager and MLOps workflow engine. It functions as a Kubernetes-native orchestrator used to coordinate data, models, and compute resources for executing machine learning pipelines and autonomous agents at scale. The platform provides specialized infrastructure for the full machine learning lifecycle, including a dedicated model serving platform to deploy trained models as scalable production-ready inference services. It also enables the coordination and state management of autonomous AI agents. The system manages scalable pipeline execution th

Inngest is a durable execution framework and event-driven automation engine designed to orchestrate background workflows. It enables developers to build resilient, stateful processes by memoizing function steps, ensuring that long-running tasks can automatically resume from the last successful operation after failures, timeouts, or infrastructure restarts. The platform distinguishes itself through its event-driven architecture, which uses a schema-validated bus to trigger functions and coordinate complex, multi-step logic. It employs an onion-model middleware approach for cross-cutting concer

Open Canvas est un système pour gérer les flux de travail d'agents IA avec état via un éditeur collaboratif et un framework d'orchestration. Il fournit un espace de travail partagé où les humains et les grands modèles de langage co-rédigent des documents et écrivent du code en temps réel, pris en charge par un éditeur de texte structuré avec rendu en direct. Le projet se distingue en intégrant un gestionnaire d'état qui suit le contexte de session, les mémoires utilisateur et les instantanés historiques à travers les fils de conversation. Il emploie un modèle d'exécution durable qui permet des interventions humaines et maintient un système de suivi de version pour les documents et les blocs de code. Le framework couvre une vaste surface de capacités, incluant des interfaces de modèle agnostiques au fournisseur pour le chat et les embeddings, une composition de flux de travail pilotée par middleware pour les garde-fous et le routage, et une gestion de contexte personnel pour stocker les préférences utilisateur. Il inclut également des outils pour le débogage de l'exécution d'agent et la possibilité de déployer le serveur via des images de conteneur. Un serveur d'agent local est fourni pour gérer les exécutions, les fils de discussion et le stockage persistant.

ZenML is an extensible machine learning orchestration framework designed to manage the end-to-end lifecycle of data pipelines and AI agent workflows. It functions as a durable orchestrator that executes machine learning tasks as directed acyclic graphs, ensuring that every step is containerized for consistent performance across local, cloud, and hybrid infrastructure. By decoupling pipeline code from underlying compute and storage backends, the platform allows developers to define infrastructure-agnostic stacks that remain portable across diverse environments. The project distinguishes itself

ZenML is an orchestration platform designed for building, deploying, and monitoring reproducible machine learning pipelines and agentic workflows. It provides a unified framework that manages the entire lifecycle of machine learning assets, from data processing and model training to the deployment of persistent inference services. By decoupling pipeline logic from underlying compute and storage, the platform enables teams to transition workflows seamlessly from local development environments to production-grade cloud infrastructure. The platform distinguishes itself through a service-oriented

OpenSquilla est un framework d'orchestration d'agents LLM conçu pour coordonner des workflows IA multi-étapes et l'exécution d'outils via des graphes orientés acycliques (DAG). Il fonctionne comme un système centralisé pour gérer des packages de compétences spécialisés et exécuter des séquences de raisonnement complexes. Le projet se distingue par une passerelle de routage qui dirige les tâches vers différents fournisseurs d'IA en fonction de la complexité, du coût et de la performance. Il utilise un système de mémoire IA à plusieurs niveaux qui organise les connaissances de travail, épisodiques et sémantiques à l'aide d'embeddings locaux et de SQLite, ainsi qu'un bac à sable d'exécution sécurisé qui isole le code généré par l'agent via des profils de permission basés sur les risques. La plateforme couvre un large éventail de capacités, incluant le déploiement multicanal vers le web et les plateformes de messagerie, la planification automatisée des tâches via cron, et un pont Model Context Protocol pour se connecter à des outils externes. Elle fournit également des outils complets de surveillance et d'observabilité pour suivre les coûts en jetons, auditer les décisions d'exécution et gérer un catalogue de compétences réutilisables. Le système inclut des utilitaires en ligne de commande pour l'initialisation de l'espace de travail et la gestion du cycle de vie des compétences.

Workflow is a platform for executing long-running, stateful processes that automatically persist progress and recover state after system interruptions. It functions as a durable orchestration engine designed to manage complex, multi-step asynchronous operations within distributed systems. The framework distinguishes itself through event-sourcing and checkpointing mechanisms that record execution steps as an immutable log, allowing for precise state reconstruction after failures. It provides sandboxed task isolation to run side-effect-heavy logic in restricted environments, ensuring that sensi