flock

📃 Flock (Flexible Low-code Orchestrating Collaborative-agent Kits)

[Multiple language options and a "Getting Started" link are provided.]

[!NOTE]

Recent updates include MCP Tools support, Parameter Extractor Node, Subgraph Node Support, Human Node, and Multimodal Chat Support.

Flock is a workflow-based low-code platform for building chatbots, RAG applications, and coordinating multi-agent teams. Built on LangChain and LangGraph, it offers a flexible solution for collaborative agents, supporting chatbots, RAG, agents, and multi-agent systems, with offline operation capability.

[Video and image overviews are included.]

Flock's workflow system consists of node types like Input, LLM, Retrieval, Tool, Retrieval Tool, Intent Recognition, Answer, Subgraph, and Start/End Nodes. Future nodes include File Upload and Parameter Extraction.

Inspired by StreetLamb, Flock introduces new features and directions. It references Lobe-chat, Dify, and FastGPT for layout.

Project tech stack: LangChain, LangGraph, React, Next.js, Chakra UI, and PostgreSQL. It supports various model providers and custom tool integration.

Roadmap includes ChatBot, SimpleRAG, Hierarchical/Sequential Agent, Workflow, Intent Recognition Node, CrewAI Integration, OpenAI, ZhipuAI, Siliconflow, Ollama, and Qwen models, Tools Calling, and I18n.

Highlights: Persistent conversations, observability via LangSmith, tool calling, RAG, human-in-the-loop, open-source models, and multi-tenancy.

[Instructions for deploying with Docker Compose or local source code are provided.]