Flowise is an open-source, low-code visual builder for LLM apps and agent workflows — drag boxes onto a canvas, connect them with arrows, and you have a working chatbot or agent without writing code. Built on LangChain.js and ReactFlow. Acquired by Workday in late 2025 to become the foundation of their enterprise AI agent platform. This section captures my hands-on observations as I build with it.
Most posts about agent platforms either evangelize from the marketing side or critique from the gripe side. I wanted something in the middle — actually using the platform, reading the codebase, building real flows, and recording what works, what surprises me, and what I'd evolve. These posts are that.
Each entry focuses on one concrete build (Hello LLM, RAG over docs, agentic patterns) with screenshots from my own running instance, code-level observations from the cloned repo, and a short section on what I'd change at the architectural level. The goal: a working knowledge resource that future-me (or anyone landing here from a search) can use as a credible first-hand reference.
The series grows as I build. Live posts are linked; coming-soon posts are scoped and on the roadmap.
Cloning the codebase, getting it running on a non-default port (Docker, after four install attempts), building a Hello LLM chatflow, and inspecting the SQLite database in real time to see what gets written when. Six codebase observations + six things I'd evolve.
A fully-local AI assistant that searches PDFs, Word docs, Markdown, and text files on your own machine — no cloud, no API calls, no data leaving the laptop. Built with Flowise patterns + Ollama + Chroma. The privacy-first angle.
The 8-node pattern that powers most enterprise document Q&A: document loader → text splitter → embeddings → vector store → retriever → conversational retrieval QA chain. With state-specific metadata filtering.
The LangGraph-style runtime in Flowise — multi-step reasoning agents that decide which tool to call instead of following a fixed pipeline. Where chatflow ends and agentflow begins.
Standardized tool surfaces via Model Context Protocol. How MCP solves the N×M integration problem for agent platforms, and where the MCP-vs-native trade-off lands inside Flowise.
What enterprise Flowise at Workday scale looks like — Org/Workspace/User hierarchy, AsyncLocalStorage-based tenant scoping, envelope encryption with KMS, eval-as-promotion-gate.
From zero to first chatflow is the entry point. It defines what Flowise is, who it's for (regular users vs. enterprise customers), walks through a real Hello LLM build, and inspects the database to show what gets written when. After reading it, the rest of the series builds on that foundation.
LangChain, CrewAI, AutoGen, LlamaIndex, Pydantic AI — how they compare, when to pick each. Flowise is built on LangChain.js, so this is the context for what Flowise abstracts.
The 15-step production RAG anatomy that the Flowise RAG flows in this series implement. Vertical case studies for HR, customer transactions, healthcare.