Chat Interface

The ai.KMITL chat interface is designed to be intuitive, powerful, and feature-rich. This guide covers everything you can do with the interface.

Interface Overview

Main Components

Message Input

Text Input

The message box at the bottom is where you type:

1. Click in the input area
2. Type your message
3. Press Enter to send (or Shift+Enter for new line) Or click the Send button

Text Formatting

Markdown Support

ai.KMITL supports full markdown formatting:

Basic Formatting:

**bold text**
_italic text_
~~strikethrough~~
`inline code`

Lists:

- Bullet point 1
- Bullet point 2
  - Nested point

1. Numbered item
2. Another item

Headings:

# Large Heading

## Medium Heading

### Small Heading

Links:

[Link text](https://example.com)

Code Blocks:

```python
def hello():
    print("Hello, World!")
```

LaTeX Math Support

For mathematical equations:

Inline Math:

The formula is $$E = mc^2$$ for energy.

Block Math:

$$
\int_0^\infty e^{-x^2} dx = \frac{\sqrt{\pi}}{2}
$$

Complex Equations:

$$
\begin{align}
f(x) &= x^2 + 2x + 1 \\
     &= (x + 1)^2
\end{align}
$$

Math Rendering

Math is rendered beautifully using KaTeX. Both inline and block math are supported!

Tutor Mode

Turn on Tutor mode from the Tools and modes menu (the sliders button next to web search) when you want to learn rather than get the answer straight away. The same menu holds optional tools such as Image generation. The menu button shows a badge while something in it is on. In tutor mode the AI:

• Asks what you already know and where you are stuck
• Breaks problems into small steps and gives hints instead of full solutions
• Checks your attempts and explains mistakes without simply fixing them
• Checks your understanding with short questions and practice quizzes that you answer right in the chat
• Walks through a full solution only when you are still stuck after several hints, or when you ask for it

Tutor mode is saved per chat, so you can use it for homework in one chat and keep normal answers in another.

Previews

Code blocks for HTML, React, Mermaid, and Markdown have a Preview tab. Use Expand to view a preview larger over the chat, or Open in side panel to keep it next to the conversation while you keep chatting. Previews run in an isolated sandbox, so code inside them cannot access your account or browser storage.

File Attachments

Supported File Types

• Images: JPG, PNG, GIF, WebP, SVG, BMP
• Documents: PDF, Word (.docx), Excel (.xlsx), TXT, MD
• Code: Most programming languages (JS/TS, Python, Java, C/C++, Go, Rust, SQL, shell, and more)
• Data: CSV, TSV, JSON, XML, YAML, TOML

Word & Excel

Word and Excel files are converted to text before being sent to the AI. Very large documents are automatically summarized to fit the context window — you'll see a note in the response when this happens. Legacy formats (.doc, .xls) are not supported; please re-save as .docx/.xlsx.

How PDFs Reach the AI

PDFs are converted to text when you upload them, so the AI reads text instead of re-processing every page image on each message:

• Short PDFs (roughly up to 8,000 tokens of text, often 10–20 pages) are included in full as text.
• Long PDFs are included as an outline (the first line of each page). The AI reads the exact pages it needs, or asks a fast reader model about the whole document, then answers with page references. You can ask about any part of the file as usual.
• Scanned PDFs (no selectable text) of up to 30 pages are transcribed automatically right after upload. Longer scans are read page by page by the reader model when you ask about them.
• $ models read the original PDF directly, because it costs them little.

Thai text from older PDF fonts (broken tone marks and vowels) is repaired during conversion.

Send Full Document

When a question depends on layout, diagrams, or charts across the whole file, open the PDF chip in the message box and turn on Send full document. The original file then goes to the model with every message while it stays in the chat, which uses much more quota; the chip shows the estimated difference before you send. It is available for PDFs up to 100 pages on models that can read PDFs.

How to Attach Files

1. Click the 📎 attachment button next to the message input
2. Select one or more files from your device
3. Wait for upload to complete (you'll see a preview)
4. Add a message describing what you want to know
5. Send

File Limits

• Maximum file size: 5 MB per file
• Maximum files per message: 10 files
• PDFs: up to 300 pages; damaged or password-protected PDFs cannot be uploaded

Workspace Files

Every chat has its own workspace: each file you attach is added to it, and the AI can read those files, edit text files, and create new ones while you chat.

• Ask for changes directly: "Fix the complexity section in my notes" edits the file itself instead of pasting a copy into the chat
• Ask for deliverables: "Write a Python script for this" or "Make this an HTML page" creates a file you can keep
• Versions: every edit saves a new version (v1, v2, …), and older versions stay available
• Download: when the AI finishes, it presents the files it made or changed as cards with View and Download

Open the Files button in the chat header to see every file in the chat. Selecting a file opens it in the side panel, where you can switch versions, preview Markdown, HTML, and CSV, and download any version.

What the AI can edit

The AI can create and edit text formats such as Markdown, plain text, CSV, HTML, JSON, SVG, and code. Word, Excel, PDF, and image files are read-only: the AI can read them, and can write the result as a new text file (for example Markdown or CSV) when you ask for changes.

Workspace files are private to you. Anyone you share a chat with sees the conversation, but cannot open or download your workspace files.

What You Can Do with Files

Images:

• "What's in this image?"
• "Transcribe the text from this screenshot"
• "Analyze this diagram"
• "Fix the code in this image"

PDFs:

• "Summarize this document"
• "Find information about X in this PDF"
• "Translate this document to English"
• "Extract data from this form"

Code Files:

• "Review this code"
• "Find bugs in this file"
• "Explain what this code does"
• "Optimize this function"

Privacy Note

Files are processed securely and not stored permanently. They're used only for the current conversation.

Voice Input

Using Speech-to-Text

Voice Input is not available

Voice Input is not currently avaiable due to security reasons

1. Click the 🎤 microphone button
2. Allow microphone access (first time only)
3. Speak your message clearly
4. Click Stop when finished
5. Review the transcribed text
6. Edit if needed and send

Voice Input Tips

• Speak clearly at normal pace
• Minimize background noise for accuracy
• Use punctuation words: "period", "comma", "question mark"
• Edit the transcript before sending if needed

Best Use Cases

• Quick questions while multitasking
• Long messages that are faster to speak
• Accessibility needs
• Mobile device use

Message Display

Viewing Responses

AI responses appear in the chat with:

• Markdown rendering: Formatted text
• Code highlighting: Syntax-colored code blocks
• Math rendering: Beautiful equations
• Clickable links: Direct navigation

Code Blocks

Code in responses includes:

• Syntax highlighting for many languages
• Copy button in the top-right corner
• Line numbers (optional)
• Language label

Example:

def fibonacci(n):
    if n <= 1:
        return n
    return fibonacci(n-1) + fibonacci(n-2)

Copying Content

• Copy button appears on hover over code blocks
• Select text to copy any part manually
• Right-click for context menu options

Message Actions

Edit Your Message

1. Hover over your sent message
2. Click the Edit button (✏️)
3. Modify the text
4. Press Enter to resend

Edit Effect

Editing regenerates the AI's response based on your new message. The conversation continues from that point.

Retry Generation

If you're not satisfied with a response:

1. Hover over the AI's message
2. Click Retry (🔄)
3. The AI will generate a new response

Retry Another Models

Users can retry using a different model by selecting a new model in the model selector before pressing the Retry button. The new response will use the currently selected model.

Pro Tip

When you want to compare how different models answer the same question, simply using branch-off button to duplicate the conversation and retry using a different model.

Continue an Unfinished Response

Sometimes a long, tool-heavy response stops before the AI writes its final answer, for example when a provider error interrupts it or the output gets cut off. When that happens, the last response shows Continue with the note "The response stopped before it finished."

Click Continue to resume that same response from where it stopped. The AI keeps everything it already did (searches, file edits, tool results) and picks up from there instead of starting over, so the earlier work is not wasted. The continued part counts toward your quota like any other response.

Conversation Management

Viewing Chat History

• All conversations are saved automatically
• Access Recent Threads in the left sidebar
• Threads show first message as title, grouped by last conversation date.
• Click any thread to open it

Thread Names

Threads are automatically named based on:

• Your first message
• AI-generated title
• You can rename threads manually

Searching Conversations

Use the search function to find:

• Specific conversations titles
• Keywords in messages

Keyboard Shortcuts

Essential Shortcuts

Action	Windows/Linux	Mac
Send message	Enter	Enter
New line	Shift+Enter	Shift+Enter
Search	Ctrl+K	Cmd+K

Conversation Context

How Context Works

The AI remembers your conversation:

• Previous messages provide context
• Uploaded files stay in context
• Switched models receive full history
• Context limit depends on model

When you exceed the limit, older messages are automatically removed from the conversation context.

Managing Context

Keep context relevant:

• Start new chat for new topics
• Summarize long conversations
• Use folders to organize by topic

Special Features

Streaming Responses

Responses stream in real-time:

• See the AI "thinking"
• Get faster perceived results
• Can stop generation early

Stop Generation

If a response is taking too long:

1. Click the Stop button
2. The AI stops immediately
3. You keep the partial response

Regenerate

Don't like a response?

1. Click Regenerate
2. Get a different answer
3. Compare versions

Mobile Experience

Mobile Interface

On mobile devices:

• Sidebar collapses to hamburger menu
• Swipe to navigate threads
• Touch to interact with messages
• Responsive design adjusts automatically

Mobile Tips

• Use voice input for easier typing
• Landscape mode for more space
• Share feature to export conversations
• Works on all modern mobile browsers

Tips for Best Experience

General Tips

1. Be specific in your questions
2. Use formatting to structure complex questions
3. Attach files when relevant
4. Enable web search for current info
5. Try different models for different tasks

Performance Tips

1. Close unused threads to free memory
2. Start new chats for new topics
3. Limit file sizes for faster uploads
4. Use faster models for simple questions

Organization Tips

1. Use folders to group related chats
2. Name threads descriptively
3. Delete old conversations you don't need
4. Export important conversations

---

Master the Interface

The more you use ai.KMITL, the more efficient you'll become. Experiment with features and find your workflow!

Next Steps

• Learn about Web Search & Tools
• Organize with Project Folders
• Improve results with Great Prompts