Awesome AI AgentsAI Application Development

lgrammel/js-agent

⭐ 272 TypeScript added to this list on 2025-04-19 repository created 2023-04-06

JS Agent is a composable and extensible framework designed for building AI agents and applications using JavaScript and TypeScript. It provides robust building blocks and tooling to help developers create reliable and robust AI agents more efficiently. The framework supports various features such as configurable agent run properties, observation of agent runs for console output or UI integration, recording of all language model calls, cost calculation for these calls, and stopping agent runs based on specific criteria like step limits. It supports multiple language models including OpenAI's text completion models (e.g., text-davinci-003), chat completion models (e.g., GPT-4, GPT-3.5-turbo), and embedding models (e.g., text-embedding-ada-002). JS Agent includes an HTTP server to start, stop, and observe agent runs via API and can host multiple agents simultaneously. It offers a variety of actions and tools such as reading and writing files, running CLI commands, using programmable search engines, extracting information from web pages, user input prompts, and calling sub-agents. The framework supports agent loops like BabyAGI-style task planning and generate-next-step loops. The project provides prompt templates for chat and text prompts, utility functions for prompt combination and conversion, and text functions for information extraction, text splitting by characters or tokens, and helpers for loading and generating text. It also supports data sources like web pages and files, and data converters for HTML and PDF to text. General utilities include retry mechanisms for language model calls with exponential backoff. JS Agent emphasizes strong typing, composability, extensibility, and functional programming principles for immutable objects, while using object-oriented design for mutable states. It aims to support progressive refinement of agent specifications and is built with production readiness in mind, including logging and cost tracking. Although JS Agent is no longer actively developed, it served as an early library for experimenting with AI agent concepts. The author recommends ModelFusion as a more general and actively maintained AI library for TypeScript that incorporates learnings from JS Agent. The repository includes several example agents demonstrating various use cases such as a Wikipedia question-answering agent, a JavaScript/TypeScript developer agent running in Docker, a BabyAGI implementation, a PDF to Twitter thread converter, and text splitting and embedding generation.

https://github.com/lgrammel/js-agent

agentagent-loopsai-agentsautogptbabyagicomposable-frameworkdata-convertersextensiblefunctional-programminggpt-3gpt-3.5-turbogpt-4http-serverjavascriptlanguage-modelsopenaiproduction-readyprompt-templatestext-davinci-003text-embedding-ada-002text-functionstypescript

Also in AI Application Development

langgenius/dify

Dify is an open-source platform for developing and deploying AI applications using large language models, featuring workflow building, model management, agent capabilities, and observability tools.

daytonaio/daytona

Daytona is a secure and elastic infrastructure platform designed for safely running AI-generated code with high performance and scalability.

e2b-dev/awesome-ai-agents

A curated and comprehensive list of AI autonomous agents categorized into open-source and closed-source projects, serving as a valuable resource for exploring, building, and contributing to AI agent development.

mlflow/mlflow

MLflow is an open-source platform that manages the full machine learning lifecycle, providing tools for experiment tracking, model packaging, deployment, evaluation, and observability in one integrated solution.

pydantic/pydantic-ai

PydanticAI is a Python agent framework that simplifies building production-grade Generative AI applications by providing type-safe, structured, and model-agnostic integration with multiple LLMs, along with features like dependency injection and real-time debugging.

GoogleCloudPlatform/generative-ai

A comprehensive repository of sample code, notebooks, and resources for developing and managing generative AI workflows on Google Cloud using Gemini models on Vertex AI.

0xPlaygrounds/rig

Rig is a Rust library for building scalable, modular, and ergonomic fullstack agents powered by large language models, supporting multiple LLM providers and vector stores for AI integration.

mnfst/manifest

Manifest is an open-source components library built on shadcn/ui for creating customizable and production-ready UI components to develop ChatGPT and conversational AI applications efficiently.