Agents

The MathGPT is a web application built with Streamlit that allows users to input mathematical problems and receive solutions. The app supports both text and LaTeX input, making it versatile for various mathematical expressions.

Local AI Agent Runtime

In this solution accelerator, we demonstrate how generative AI, retrieval augmented generation (RAG) and multi stage reasoning can be used to better navigate through the complexities of regulatory filings, bringing more transparency for companies to disclose their societal and environmental impacts.

Embodied-OS：面向具身智能的分层运行时——Driver·Capability·Skill·Brain 全栈参考实现

This sample application demonstrates the practical implementation and usage patterns of the LLM Agents library.

It automatically describes images in PDF files and generates questions from these descriptions. With its advanced RAG structure, it directs these questions directly to PDF text content, providing comprehensive information extraction and analysis.

This repository contains sample code demonstrating how to implement a verified semantic cache using Amazon Bedrock Knowledge Bases to prevent hallucinations in Large Language Model (LLM) responses while improving latency and reducing costs.

HumanStudy-Bench: Towards AI Agent Design for Participant Simulation

A modular, LLM-agnostic fullstack agent framework.

Implementation of Anthropic agentic workflow's using Google Gemini 2.0 Flash Experimental

Fast & less costly AI decision making and intelligent processing of multi-modal data.

This is an official repository for "Exploring Multi-Modal Data with Tool-Augmented LLM Agents for Precise Causal Discovery".

Pocket Flow: A minimalist LLM framework. Let Agents build Agents!

ByeDB.AI is an innovative AI-powered platform transforming natural language into data insights, all within a familiar chat interface. It enables non-technical users to effortlessly query databases, visualize results with charts, and export data, eliminating the need for SQL knowledge.

This Agentic system is designed to handle complex data analysis tasks by breaking them down into smaller subtasks and delegating them to specialized agents. The system uses a multi-agent architecture to improve performance and scalability.

A declarative meta-language for orchestrating multi-agent workflows. Readable by humans. Executable by LLMs. Portable across models.

🐙 Interfaces (aka `into`) is the quickest way to make Large Language Models do things. Get started in 3 lines of code!

The 3rd Eye is a modular OSINT (Open Source Intelligence) framework built on an agent-based, graph-driven architecture. It automates public information discovery, identity correlation, and exposure analysis across multiple platforms, and generates structured intelligence reports. The system follows a LangGraph agent design.

Beyond Better with code and docs

ToolRM: Towards Agentic Tool-Use Reward Modeling

Powering Agent Chains by Constraining LLM Outputs

A Python-based logical reasoning system combining rule-based engines with Large Language Models (LLMs) for business automation. Supports deduction and hypothesis testing, partial fact completion, and automated fact retrieval via documents or chat. Features include explainable reasoning, workflow integration, and customizable knowledge bases.

An interactive and comprehensive legal agent benchmark

Metadata filtering in RAG application