⚡WindyFlo Quickstart

WindyFlo is a visual platform for building and testing AI pipelines — from prompt chaining to model orchestration.

This guide will walk you through the basics. You'll be up and running with your first pipeline in minutes.

---

1. What is a Pipeline?

A pipeline in WindyFlo is a visual, node-based automation flow that connects inputs, logic, and outputs. It can power tasks such as conversational AI, document search, or external API calls.

How it works:

• Drag and drop nodes onto a visual canvas

• Connect them to create data flow from input → processing → output

• Control each node’s behavior using parameters (LLM prompts, API settings, etc.)

---

2. Creating a New Pipeline Workspace

WindyFlo uses workspaces to design and manage your AI pipelines.

Steps:

1. Go to My Pipeline

   In the top menu, click My Pipeline to view your existing pipelines.

2. Click + Create Pipeline

   This takes you to a form for creating a new workspace.

3. Enter Basic Info:

   • Pipeline name: e.g. sample_chat, mail_assistant

   • Description: Short description (max 100 characters)

   • Tags: Add keywords to categorize your pipeline

   • Visibility: Choose Public (shared) or Private (for Pro users)

   • Allow others to run: Toggle access permission for other users

4. Click Create AI Model Pipeline

   This creates the pipeline and opens the visual editor.

---

3. Connecting Basic Nodes

WindyFlo pipelines follow a basic structure:

Input → Processing → Output

Key Concepts:

• Each node has:

  • Inputs on the left

  • Outputs on the right

• Common starting/ending nodes:

  • Agents and Chains nodes handle user input and generate the final response

• Extend functionality using:

  • Document Loaders: For document-based Q&A

  • Tools/API nodes: For web searches, API calls, etc.

Advanced Flow Control Nodes:

• Multi Agents:

  Combine Supervisor and Worker nodes to create agent teams with distinct roles

• Sequential Agents:

  Use seqStart, seqLoop, seqCondition, seqEnd nodes for loops, conditions, and multi-step logic

This no-code interface lets you visually compose powerful AI workflows.

---

4. Configuring Node Parameters

Once nodes are connected, configure each one’s parameters for it to function properly.

Node Structure:

Each node includes:

• Inputs: Data from external sources or previous nodes (e.g., documents, chat models)

• Parameters: Controls logic (e.g., prompt text, model settings, loop count)

• Outputs: Passes results to the next node (e.g., summaries, search results)

Tips:

• Fields marked with * are required

• Use the expand icon next to prompt fields to open a larger editor

• Some nodes offer advanced configuration (e.g., Format Prompt Values)

• For external APIs, set Credentials (API keys, tokens, etc.)

If a node is misconfigured, a red warning icon will appear.

---

5. Running and Testing the Pipeline

After setup, test your pipeline with the built-in chat interface.

How to Run:

1. Click Save

   You must save the pipeline before running.

2. Click ▶ Run

   This opens the chatbot interface.

3. Enter a Message

   Type a question like:

   • “Summarize the contents of this PDF”

   • “Tell me the latest news”

   This message is passed to the starting Agent/Chain node.

4. View Responses

   You’ll see output like plain text, summaries, JSON, or triggered tool actions.

5. Iterate

   Modify nodes, click Save, then Run again to test updated logic.

Warnings:

• Check all required parameters

• Agents/Chains must connect to a Chat Model

• Tool/API nodes require valid credentials

• For complex pipelines (Multi Agents, Sequential), verify each step individually

---

6. Saving and Deploying Pipelines

Pipelines can be saved, shared, and deployed as APIs or embeddable components.

Edit Pipeline Info:

Click the ✏️ Edit icon in the left sidebar to update:

• Name

• Description

• Tags

• Visibility (Public = shared in Pipeline Hub, Private = user-only)

• Allow others to run

Add a README:

Click the 📄 icon to open the README editor

Add markdown content to explain use cases, warnings, and how to run it.

This helps others understand your pipeline when you publish it.

Deploy as API:

Click </> Embed as API (top-right) to open deployment options:

• Embed: Get HTML/JS code to add chatbot to your website

• Python / JS / curl: Auto-generated code to call your pipeline from external apps

• Share Chatbot: Generate a public URL for others to interact with your chatbot

You can customize title, welcome message, font, background color, and access permissions.

---

Next Step: Try the Tutorial

Ready to build your own?

Follow the official tutorial and build a complete chatbot pipeline step-by-step.

👉 Try the WindyFlo Tutorial

---

Let me now export this into a Word and PDF format for you.

✅ Here is the complete, developer-friendly English version of the WindyFlo Quickstart Guide in Word format:

📄 Download WindyFlo_Quickstart_EN_Full.docx

Let me know if you’d like a PDF version as well or want help embedding this into GitBook or your website.