LangChain in visual form
In February 2023, Rodrigo Nader and Gabriel Almeida from the Brazilian company Logspace released LangFlow, an open source visual builder for constructing LangChain applications through a node-based graphical interface. The project was born a few weeks after the popularity explosion of LangChain and rapidly became one of the first low-code tools for that ecosystem.
Stack and architecture
LangFlow is a Python application on the backend, with a React frontend. Unlike Flowise (TypeScript on LangChain.js), LangFlow operates on the Python version of LangChain, making it consistent with the dominant language in the AI/ML community and allowing direct integration with scientific and data-processing libraries in the Python ecosystem.
The user builds the flow by dragging nodes onto the canvas — LLM, prompt, chain, agent, vector store, tool — and connecting them. LangFlow supports custom components written in Python: it is possible to insert custom classes that extend LangChain and use them as visual nodes inside flows.
Each flow can be exported as a REST API or as a serialised LangChain object, usable in Python applications outside the canvas.
Licence and acquisition
LangFlow is distributed under the MIT licence, among the most permissive in the open source ecosystem. In April 2024, DataStax — the company known for the managed Cassandra database and the Astra DB platform — acquired Logspace and took over LangFlow’s development, keeping its open source nature and adding native integration with Astra DB as the reference vector store.
Positioning
LangFlow and Flowise represent the two main open source options for visual construction of LangChain applications. LangFlow, with its Python stack and the corporate backing of DataStax, is the natural choice for teams already oriented towards Python data engineering; Flowise is preferable in JavaScript/TypeScript contexts.
Link: langflow.org