Awesome AI Agents › AI Agent Operating System

arakoodev/EdgeChains

⭐ 428 JavaScript added to this list on 2025-04-19 repository created 2023-03-20

EdgeChains.js is a comprehensive full-stack Generative AI (GenAI) library designed to simplify the deployment, configuration, and management of GenAI applications. It supports front-end, back-end, APIs, prompt management, and distributed computing, with all core prompts and chains managed declaratively using Jsonnet, a configuration language developed by Google. This approach avoids embedding prompts in code classes, making them versionable, diffable, and easier to maintain. EdgeChains is built on top of robust technologies like honojs and Jsonnet, ensuring stability and production readiness. The library addresses the complexity of prompt engineering in Generative AI, where different large language models (LLMs) require specific prompt formats and chains. EdgeChains tackles the problem of "prompt explosion"—the proliferation of prompts needed for different LLMs—and "prompt drift," where prompt effectiveness changes over time. By managing prompts declaratively and externally, EdgeChains enables easier version control, testing, and iteration in production environments. Key features include automatic parallelism that distributes LLM chains and chain-of-thought tasks across CPUs, GPUs, and TPUs using WebAssembly runtime, fault tolerance with retry and backoff mechanisms, and scalability to handle large numbers of APIs, prompt lengths, and vector datasets. The library also incorporates token cost tracking and measurement to optimize prompt usage and reduce operational costs. EdgeChains emphasizes simplicity, allowing users to run production-ready GenAI applications with just one script file and one Jsonnet file. It supports integration with vector databases like Supabase for advanced use cases such as chat with PDF documents. The project provides detailed setup instructions and examples, fostering a friendly developer community via Discord. Overall, EdgeChains offers a unique, configuration-driven approach to building and managing Generative AI applications, focusing on deployment, scalability, and maintainability rather than just UI or library design patterns.

https://github.com/arakoodev/EdgeChains

agentaiai-agentsautogptchain-of-thoughtchatbotdistributed-computingfault-tolerancefull-stackgenaigenerative-aigpthonojshtmljavascriptjsonnetkubernetes-config-managementllmllm-chainsopenaiparallelismproduction-readyprompt-driftprompt-engineeringprompt-managementprompt-versioningrest-apiscalabilitysupabasetoken-cost-trackingtypescriptvectorvector-databaseweb-developmentwebassemblywebdevycombinator