Interactive Mode

Feature Overview: iFlow CLI provides multiple flexible interaction methods, supporting text input, image processing, file references, and intelligent multimodal processing.

Learning Time: 10-15 minutes

Prerequisites: iFlow CLI installed and configured, understanding basic command line operations

What is Interactive Mode

Interactive mode is the core functionality of iFlow CLI, allowing users to engage in natural conversations and collaboration with AI through multiple methods. The system supports various input forms including text, images, and file references, and provides intelligent multimodal processing capabilities that enable any model to "understand" image content.

Core Features

Feature	Description	Platform Support
Multiple Input Methods	Various input forms including text, images, file references	All platforms
Intelligent Multimodal Processing	Enables any model to "understand" image content	All platforms
Automatic Content Detection	Intelligently recognizes and processes different types of input content	All platforms
Large Text Optimization	Automatically handles long text with optimized interface display	All platforms
Real-time Response	Processes user input in real-time without waiting	All platforms

How It Works

Input Processing Flow

User Input → Content Type Detection → Preprocessing → Model Adaptation → AI Response
    ↓
[Text/Image/File] → [Auto Recognition] → [Format Optimization] → [Multimodal Processing] → [Generate Reply]

Intelligent Adaptation Mechanism

• Text Input: Directly passed to AI model for processing
• Image Input: Automatically detects model capabilities, generates image descriptions when necessary
• File References: Reads file content and integrates into conversation context
• Mixed Input: Intelligently combines different types of input content

Detailed Feature Description

Text Input

Single Line Text

Directly input your questions or instructions in the command line interface:

> Help me optimize the performance of this React component

Multi-line Text Input

Supports multiple methods for multi-line text input:

Method	Operation	Description
Backslash newline	\ + Enter	Quick multi-line input creation
Shift + Enter	Shift + Enter	Available after terminal configuration

Example:

> Please help me implement a user management system, including:\
  1. User registration and login functionality
  2. User information CRUD operations
  3. Permission management
  4. Data persistence

Image Processing

Supported Image Formats

Format	Extensions	Description
PNG	.png	High-quality images with transparency support
JPEG	.jpg, .jpeg	Compressed image format
GIF	.gif	Supports animated images
WebP	.webp	Modern image format
BMP	.bmp	Bitmap format

Image Input Methods

Screenshot Pasting

Platform	Screenshot Shortcut	Paste Shortcut
macOS	Cmd + Shift + 4	Cmd + V
Linux	PrtScn or others	Ctrl + V

Windows currently does not support screenshot pasting.

Operation steps:

1. Use system screenshot tool to capture the screen area you want to analyze
2. Press paste shortcut in iFlow CLI
3. System automatically generates image placeholder and processes it

File Pasting

• Copy image files in file manager
• Use paste shortcut in CLI
• All platforms support image file pasting

Display effect after pasting:

> [Pasted image #1] What's wrong with this interface?

Note that native terminals, iTerm terminals, and IDE built-in terminals will forcibly filter out image paste events, causing image pasting to fail when using platform native paste shortcuts (for example, using Cmd + V to paste images on macOS is ineffective). In this case, you can use Ctrl + V or Shift + Ctrl + V to bypass this filter and paste images.

Image Processing Examples

> [Pasted image #1] Please analyze the design issues of this user interface
> [Pasted image #2] Which of these two interface layouts is better?
> Help me write the corresponding CSS code based on [Pasted image #1] this design draft

Large Text Processing

Automatic Detection Rules

Detection Item	Threshold	Processing Method
Long text paste	>800 characters or >10 lines	Generate text placeholder
Long content display	>900 characters	Truncated display optimization
Placeholder format	-	[Pasted text #X +Y lines]

Interface Optimization Features

• Visual Simplicity: Avoids long text causing scrolling chaos in terminal
• Content Integrity: Although display is optimized, model still receives complete original text
• Smart Truncation: Shows first 5 lines and last 5 lines (when line count doesn't exceed but character count exceeds threshold, only shows first 400 characters and last 400 characters), with middle content hidden
• Clear Identification: Placeholder clearly identifies text block line count for easy content scale recognition

Usage Examples

> Please help me refactor this code: [Pasted text #1 +45 lines]
> Analyze the errors in this log file: [Pasted text #2 +120 lines]

Operation Flow

1. Copy large amounts of text content to clipboard
2. Press Ctrl/Cmd + V in CLI
3. System automatically generates placeholder and saves original content
4. Continue inputting your questions or instructions

Paste Performance Optimization Configuration

In some Windows systems, clipboard operations have performance issues, manifesting as slow response when pasting text or image files (some taking over 1000ms, while macOS and Linux systems typically complete within 100ms). This is mainly due to performance limitations of reading clipboard content through PowerShell on Windows systems.

Performance Comparison

Operating System	Average Paste Time	Extreme Paste Time	Performance
Windows	300-600ms	>1000ms	Some machines may have performance issues
macOS	<100ms	<100ms	Good performance
Linux	<100ms	<100ms	Good performance

Solution

For affected Windows users, the system provides a configuration option to skip clipboard reading:

• pasteFromClipboard (boolean):
  • Description: Controls whether to handle paste operations through clipboard reading. When enabled, provides multi-line text paste optimization, image file pasting, and screenshot pasting functionality
  • Default: true (uses clipboard reading mechanism)
  • Example:

    "pasteFromClipboard": false

Usage Notes

• Applicable Scenario: Only recommended for Windows users affected by paste performance issues
• Feature Limitations: Setting to false will lose multi-line text optimization, image file pasting, and screenshot pasting functionality
• Terminal Compatibility: After setting to false, modern terminals (such as Windows Terminal or IDE built-in terminals) will significantly improve normal text paste speed (<100ms); however, in traditional Windows consoles (such as cmd.exe, PowerShell), setting to false will cause inability to paste any content because these terminals do not support the bracketed paste mechanism.

File References

Reference Methods

Type	Syntax	Example
Single file	@filepath	@src/App.tsx
Directory reference	@directorypath	@src/components
Multiple files	Space separated	@file1.ts @file2.ts

Usage Examples

Single File Reference

> Help me optimize the @src/components/UserProfile.tsx component
> Please explain the API design philosophy in @docs/api.md

Directory Reference

> Analyze the utility functions in the @src/utils directory
> Refactor the component structure in the @src/components directory

Multiple File References

> Compare the differences between @src/old-component.tsx and @src/new-component.tsx
> Optimize project structure based on @package.json and @tsconfig.json configurations

Intelligent Multimodal Processing

Core Features

The unique aspect of iFlow CLI lies in its intelligent multimodal processing mechanism, enabling any model to "see" images.

How It Works

Automatic Image Description Generation

Step	Operation	Description
1	Model capability detection	Checks if main model supports multimodal
2	Multimodal call	Automatically calls multimodal model when not supported
3	Description generation	Generates detailed image description information
4	Content integration	Passes description to main model for processing

Description Content Includes

• Overall layout and composition
• Main objects and positional relationships
• Color information and visual features
• Text content (complete transcription)
• Background environment and detail features

Registration Configuration

iFlow AI Login Method

Item	Description
Recommended Choice	System-provided models are all optimized
Automatic Processing	Automatically uses qwen-vl-max to generate image descriptions
Configuration Needs	No manual configuration needed, fully automated

OpenAI Compatible Login Method

Configuration Item	Description	Required
Main model name	Specify the primary model to use	Yes
Multimodal model	Used for image description generation	No
Skip configuration	Send directly to main model	No

Performance Notes

Item	Description
Advantage	Enables all models to have image understanding capabilities
Note	Non-multimodal models process images slightly slower (because image descriptions need to be generated first, then input to the main model for understanding)
Recommendation	For frequent multimodal needs, recommend using multimodal models directly

Keyboard Shortcuts

Basic Shortcuts

Function	Shortcut	Description
Cancel operation	Ctrl/Cmd + C	Cancel current input or generation
Exit program	Ctrl/Cmd + D	Exit CLI session
Clear screen	Ctrl/Cmd + L	Clear terminal screen
Command history	↑/↓	Browse command history

Input Related

Function	Shortcut	Description
Paste content	Ctrl/Cmd + V	Automatically detects images, text
Multi-line input	\ + Enter	Backslash newline
Multi-line input	Shift/Option + Enter	Based on terminal configuration

Vim Mode

After enabling with /vim command, supports:

Operation	Key	Description
Cursor movement	h/j/k/l	Left/Down/Up/Right
Delete line	dd	Delete current line
Delete character	x	Delete current character

Usage Examples

Common Scenarios

UI Design Analysis

> [Pasted image #1] Please analyze the user experience issues of this login page

Code Debugging

> [Pasted image #1] How to solve the problem shown in this error screenshot?

File Analysis

> Please analyze the performance issues of @src/components/Header.tsx and provide optimization suggestions

Complex Problems

> Please help me implement a user management system, including:\
  1. User registration and login functionality
  2. User information CRUD operations
  3. Permission management
  4. Data persistence

Best Practices

Image Usage Tips

• UI Design Analysis: Screenshot interfaces then ask for design improvement suggestions
• Code Debugging: Paste error screenshots to quickly locate problems
• Document Understanding: Upload charts, flowcharts and other complex visual content
• Comparative Analysis: Paste multiple images simultaneously for comparison

Text Processing Optimization

• Large File Handling: Use automatic placeholder feature to handle long code or logs
• Structured Questioning: Use multi-line input to organize complex questions
• Context References: Combine @ file references to provide complete context

Model Selection Recommendations

• Pure Text Tasks: Text models are sufficient and faster
• Occasional Image Needs: Continue using text models, automatically handle images
• Frequent Multimodal: Choose multimodal models for best experience

Troubleshooting

Common Issues and Solutions

Issue	Possible Cause	Solution
Image paste failure	Unsupported format or no clipboard data	Check image format, re-screenshot or copy
Invalid file reference	Incorrect path or file doesn't exist	Confirm file path is correct and file is accessible
Multi-line input abnormal	Terminal configuration issue	Try different newline shortcut combinations
Multimodal processing error	Incorrect model configuration	Check model support and configuration
Large text upload failure	Content too large or network issues	Upload in segments or check network connection

Diagnostic Steps

1. Permission Check

   • Confirm system allows CLI access to clipboard
   • Check file read permissions

2. Network Connection

   • Confirm network connection is normal (multimodal processing requires API calls)
   • Check firewall settings

3. Configuration Verification

   • Review error message information
   • Adjust operations based on prompts
   • Restart CLI to ensure configuration takes effect

Platform Compatibility

Platform	Support Status	Notes
Windows	Full support	Note path separators
macOS	Full support	System permissions may need authorization
Linux	Full support	Depends on system clipboard tools