flowise-builder

name	flowise-builder
description	Flowise visual LLM workflow builder — drag-drop chatflows, API endpoints, document loaders, tools. Use when working with flowise builder.
domain	automation
author	oyi77
license	Apache-2.0
subdomain	workflow-automation
tags	["api","automation","builder","flowise","productivity","workflow"]
version	1.0.0

Overview

Flowise is an open-source visual tool for building LLM workflows. It provides a drag-drop interface to connect LLMs, document loaders, vector stores, tools, and chains — then deploy as API endpoints.

Capabilities

• Build chatflows visually with drag-drop nodes
• Connect to OpenAI, Anthropic, Ollama, and local models
• Add document loaders (PDF, web, CSV, Notion)
• Integrate vector stores (Pinecone, FAISS, Chroma, Qdrant)
• Add tools (web search, calculator, API calls)
• Deploy as REST API with streaming support
• Embed chatbot widget in websites

When to Use

Trigger phrases:

• "flowise builder"

• "Flowise visual LLM workflow builder — drag-drop chatflows, API endpoints, docume"

• Building LLM apps without writing code

• Prototyping RAG chatbots quickly

• Needing visual workflow design for AI pipelines

• Deploying AI chatbots as APIs or website widgets

• Self-hosting AI infrastructure

When NOT to Use

• Task requires custom AI model training (use ML tools)
• You need complex AI agent logic (use LangChain directly)
• Task is about data processing, not AI app building
• You don't have Flowise instance running
• Task requires real-time AI inference (use dedicated AI services)
• You need to build a custom AI application (use development tools)

Pseudo Code

Implementation patterns for common use cases with this skill.

Installation

# npm
npm install -g flowise
npx flowise start

# Docker
docker run -d -p 3000:3000 flowiseai/flowise

# Access at http://localhost:3000

Chatflow Architecture

Document Loader → Text Splitter → Embedding → Vector Store
                                                    ↓
User Question → Embedding → Vector Store Retriever → LLM Chain → Response

API Usage

# Prediction
curl -X POST http://localhost:3000/api/v1/prediction/{chatflow-id} \
  -H "Content-Type: application/json" \
  -d '{"question": "What is the return policy?", "overrideConfig": {}}'

# Streaming
curl -X POST http://localhost:3000/api/v1/prediction/{chatflow-id} \
  -H "Content-Type: application/json" \
  -d '{"question": "Hello", "streaming": true}'

Embed Widget

<script type="module">
  import Chatbot from "https://cdn.jsdelivr.net/npm/flowise-embed/dist/web.js"
  Chatbot.init({
    chatflowid: "your-chatflow-id",
    apiHost: "http://localhost:3000",
  })
</script>

Node Configuration

Node	Config
ChatOpenAI	model, temperature, maxTokens, apiKey
OpenAIEmbeddings	modelName, apiKey
VectorStoreRetriever	topK, filter
TextSplitter	chunkSize, chunkOverlap
Calculator	—
RequestsGet	url, headers
CustomJS	code

Common Patterns

Pattern	When to Use
Document Loader → Vector Store	Index knowledge base
Retriever → LLM Chain	RAG chatbot
Agent + Tools	Autonomous assistant
Conditional Branches	Different paths based on input
Memory	Multi-turn conversations

Error Handling

Error	Cause	Fix
API key not set	Missing env var	Set OPENAI_API_KEY in .env
Vector store empty	Documents not indexed	Re-upload and process documents
Node connection error	Invalid node config	Check node settings in UI
Streaming not working	Missing streaming flag	Add streaming: true in API call

Red Flags

• Not testing flows before deployment
• Ignoring error handling in flows
• Missing logging and monitoring
• Not documenting flow logic
• Ignoring rate limits and quotas

Verification

• Flows are tested end-to-end
• Error handling is in place
• Logging and monitoring are configured
• Flow logic is documented
• Rate limits are respected

Process

1. Analyze the task requirements
2. Apply domain expertise
3. Verify output quality

Anti-Rationalization Table

Rationalization	Reality
"Manual is faster for one-off tasks"	One-off tasks become recurring. Automate early, save time later.
"I will add error handling later"	You never do. Handle errors from day one.
"Automation is overkill"	If you do it twice, automate it. If you do it daily, it is critical infrastructure.