Report Abuse

Basic Information

DemoGPT is an open-source, all-in-one library and toolkit for building, customizing, and running LLM-powered agents and agent-driven applications. It provides an AgentHub that developers can use to create agents that call tools, perform multi-step reasoning, and assemble LangChain code into runnable Streamlit applications. The repository describes a plan-based pipeline that generates a plan from user instructions, creates tasks, turns tasks into code snippets, and assembles final code. DemoGPT includes ready agent types such as ToolCallingAgent and ReactAgent, a Retrieval-Augmented Generation component (BaseRAG) for document grounding, a suite of built-in tools for search, weather, Python execution, YOLO object detection and domain-specific sources, and examples showing how to combine RAG, tools, and LLMs. The project can be used as an installable package or run from source to iterate on generated apps and agent behaviors.

Links

Categorization

App Details

Features
DemoGPT bundles features for agent development and orchestration including an AgentHub for creating and registering custom tools and agents, built-in agent classes like ToolCallingAgent and ReactAgent that expose stepwise reasoning and tool-calling workflows, and a BaseRAG module that supports vector stores such as chroma, pinecone and faiss. It ships many prebuilt tools including DuckDuckGoSearchTool, WeatherTool, WikipediaTool, BashTool, PythonTool, ArxivTool, YouTubeSearchTool, StackOverFlowTool, RequestUrlTool, WikiDataTool, PubmedTool and a YoloTool for object detection. The repo demonstrates how to combine RAG with agents, use multi-tool chains, persist vectorstores, and generate LangChain code that is turned into Streamlit apps. Installation and usage are provided for both pip package and source modes and examples show detailed agent runs, tool calls, and output formatting.
Use Cases
For developers and teams building LLM-driven applications, DemoGPT accelerates prototyping and production workflows by automating many agent-building tasks. It produces LangChain code from high-level instructions, converts code into interactive Streamlit apps, and offers prebuilt tools and agent templates to handle search, computation, vision, and domain retrieval. The RAG integration enables document-grounded reasoning across multiple vectorstores and the ReactAgent and ToolCallingAgent examples illustrate composable multi-step decision making and tool orchestration. DemoGPT can be installed via pip or run from source, making it easy to integrate into existing Python projects, test generated code phases, and iterate on agent logic. Planned additions include autonomous API usage, Gorilla integration for external API calls, and a public example database to speed refinement.

Please fill the required fields*