Flowise: An Open-Source No-Code Tool for Visually Building LLM Workflows 118

Flowise is an open-source low-code/no-code tool mainly used for visually building LLM workflows. Through its drag-and-drop interface, users can quickly build LLM applications without writing complex code. It supports multiple workflow modes (such as Assistant, Chatflow, and Agentflow) and a rich set of components (such as models, prompt templates, chains, etc.). At the same time, it also supports multiple LLM models, extension features, and multi-platform deployment.

Flowise Features

• Low-code/No-code operation: Through the drag-and-drop interface, users can quickly build LLM applications without writing complex code.

• Visual construction: Provides an intuitive graphical interface, supporting the drag-and-drop and connection of multiple components (such as LLM models, prompt templates, chains, etc.).

• Supports multiple workflow modes:

• Assistant: Suitable for beginners, used to create simple chatbots.

• Chatflow: Used to build single-agent systems and simple LLM flows.

• Agentflow: Supports more complex multi-agent systems and workflow orchestration.

• Comprehensive model support: Seamlessly integrates with hundreds of open-source or proprietary LLMs, supporting HuggingFace, Ollama, Localai, etc.

• Rich extension features: Supports integration with components such as memory, data loaders, caching, and moderation.

• Developer-friendly: Provides API, SDK, and embedded chat features, making it convenient for developers to extend and integrate.

• Multi-platform support: Supports running on local environments and cloud platforms (such as AWS, Azure, GCP).

Flowise Tutorial

Installation

Install via Docker:

• Clone the Flowise project.

• Go to the docker folder in the project's root directory.

• Copy the .env.example file and rename it to .env.

• Run docker compose up -d to start the service.

• Visit http://localhost:3000 to use Flowise.

Install via Node.js:

• Make sure Node.js is installed (version >= 18.15.0).

• Use the command npx flowise start --FLOWISE_USERNAME=user --FLOWISE_PASSWORD=1234 to start Flowise.

• Visit http://localhost:3000 to use Flowise.

Building a Workflow

• Create a new workflow: After logging into Flowise, click the "Add New" button in the upper-right corner of the page to add a new workflow.

• Add components: Select the required components (such as LLM models, chains, prompt templates, etc.) from the component panel on the left and drag them onto the canvas.

• Connect components: Connect the output port of one component to the input port of another to form a workflow.

• Configure component parameters: Fill in the necessary parameters in the component's configuration panel, such as the LLM's API key, prompt template content, etc.

• Save and test: After saving the workflow, click the chat dialog in the upper-right corner to test it.

Examples

• Chatbot: By connecting OpenAI models, prompt templates, and LLM chains, you can quickly build a simple chatbot.

• Question-answering system: Combined with data loaders and retrieval components, you can build a knowledge-base-based question-answering system.

Technical Architecture

• Frontend: Uses React + TypeScript, with visualization built on ReactFlow.

• Backend: Uses Node.js + Express, supporting RESTful API and WebSocket.

• Deployment solution: Supports one-click cloud-native deployment with Docker + Kubernetes.

• Package management: Uses pnpm workspace for multi-module dependency management.

• Documentation generation: Automatically generates API documentation through Swagger UI, making it convenient for developers to call and integrate APIs.

Use Cases

• Intelligent customer service system: Connects to an enterprise knowledge base to automatically answer product questions.

• Contract review assistant: Parses legal documents and automatically generates risk reports.

• Scientific literature analysis: Batch-processes papers to extract key research findings.

• Intelligent recruitment system: Automatically parses resumes and generates candidate evaluations.

Github:https://github.com/FlowiseAI/Flowise