🎨 Flow Studio User Guide

📖 Introduction

Flow Studio is a powerful visual node-based editor that lets you create sophisticated AI processing pipelines. Connect different AI models, combine their outputs, and automate your workflow.

📋 How to Use the Default Best-of-5 Flow

The default flow is designed to analyze your code and generate the best possible response by consulting 5 different AI models and synthesizing their answers.

1. Step 1: Write Your Question
   In the main AICodePrep window, type your question or task in the "Prompt / Question" text box
2. Step 2: Generate Context
   Click the "GENERATE CONTEXT!" button
   → This creates fullcode.txt with your code + question
3. Step 3: Configure API Key
   In Flow Studio, click "🔑 Manage API Keys" and add your OpenRouter API key
   → Get a free key at openrouter.ai/keys
4. Step 4: Run the Flow
   Click the "▶ Run Flow" button in Flow Studio toolbar
   → Watch as 5 AI models analyze your code and synthesize the best answer
5. Step 5: Get Results
   Find your answer in:
   • 📋 Clipboard (automatically copied)
   • 📄 best_of_all1.txt (saved to disk)
   • 👁️ Output Display panel (visible in Flow Studio)

🔑 Setting Up API Keys

Before you can use AI nodes, you need to configure your API keys. You have two options:

Option 1: In-App API Key Manager (Easiest! ⭐)

1. In Flow Studio toolbar, click "🔑 Manage API Keys" button
2. Enter your API keys for each provider you want to use
3. Click "Save"
4. Done! Your keys are securely saved to the config file

Option 2: Manual Configuration File

Step 1: Locate your config file at:

Windows: C:\Users\[YourUsername]\.aicodeprep-gui\api-keys.toml
Mac/Linux: ~/.aicodeprep-gui/api-keys.toml

Step 2: Open the file in a text editor and add your keys:

[openrouter]
api_key = "sk-or-v1-your-openrouter-key-here"
base_url = "https://openrouter.ai/api/v1"

[openai]
api_key = "sk-your-openai-key-here"
base_url = "https://api.openai.com/v1"

[gemini]
api_key = "your-gemini-api-key-here"

🎯 Node Types

🤖 LLM Provider Nodes

These nodes send your input to AI models and return the response:

• OpenRouter Access to 100+ models from various providers
• OpenAI GPT-4.1, GPT-5, o3, and other OpenAI models
• Gemini Google's Gemini models
• OpenAI Compatible Any OpenAI-compatible API endpoint

📁 Input/Output Nodes

• Context Output Provides your selected files and prompt as input
• File Write Saves output to a file
• Clipboard Copies output to clipboard
• Output Display Shows output in the UI

🔧 Utility Nodes

• Best of N Takes multiple AI responses and synthesizes the best answer using another AI model

🎮 Using Flow Studio

Navigation Controls

• Pan: Hold Spacebar and drag, or use the hand tool (✋)
• Pan with Arrows: Use arrow keys (←↑↓→) to pan
• Zoom: Mouse wheel or 🔍+/🔍- buttons
• Fit View: Click ⊡ button to fit all nodes

Working with Nodes

• Add Node: Click "➕ Add Node" or press Tab
• Connect Nodes: Drag from an output port (right side) to an input port (left side)
• Select Node: Click on a node to select it
• Move Node: Click and drag a node
• Delete Node: Select node and press Delete or Backspace
• Configure Node: Select a node to see its properties in the Properties Panel on the right

Configuring LLM Nodes

1. Click on an LLM node to select it
2. In the Properties Panel, you'll see:
   • model_mode: Choose "choose" to specify a model, or "random"/"random_free" for automatic selection
   • model: Enter the model ID (e.g., "openai/gpt-4", "anthropic/claude-3-opus")
   • output_file: Where to save the model's response (e.g., "LLM1.md")
   • temperature: Controls randomness (0.0-2.0, typically 0.7)

🚀 Running Your Flow

1. Make sure all nodes are connected properly (no red error indicators)
2. Ensure your API keys are configured
3. Click "Run Flow" button in the toolbar
4. Watch the execution progress in the status bar
5. Check the output files or clipboard for results

📝 Built-in Flow Templates

Best-of-5 (Blank)

A template with 5 LLM nodes and a Best-of-N synthesizer. Models are not pre-configured, so you need to set them up yourself.

Best-of-5 (Configured)

A ready-to-use flow with pre-configured models from OpenRouter:

• GPT-5 Codex
• Claude Sonnet 4.5
• GLM-4.6
• Qwen3 Next 80B
• O4 Mini
• Synthesized by Gemini 2.5 Pro

Just add your OpenRouter API key and you're ready to go!

💾 Saving and Loading Flows

• Export: Flow → Export Flow JSON… saves your current flow to a .json file
• Import: Flow → Import Flow JSON… loads a previously saved flow
• Reset: Flow → Reset to Default Flow clears the canvas

❓ Common Issues

Node won't execute / Shows error

• Check that your API key is configured in api-keys.toml
• Verify the model ID is correct
• Ensure input ports are connected
• Check your internet connection

Can't connect nodes

• Make sure you're connecting output (right) to input (left)
• Ports must be compatible types (text → text)

Flow runs but no output

• Check that output nodes (Clipboard, File Write) are connected
• Verify file paths in File Write nodes are valid
• Look for error messages in the log

🎓 Example Use Cases

Best-of-N Analysis

Send the same prompt to multiple different AI models and synthesize their responses into one superior answer.

Multi-Stage Processing

Chain multiple AI calls: first model generates ideas, second model refines them, third model writes code.

Parallel Processing

Run multiple independent AI tasks simultaneously and collect all results.

📚 Additional Resources

• OpenRouter Documentation
• OpenAI API Documentation
• Google AI Documentation