Chat Interface

The chat interface is where you have conversations with AI. Here's how to use it effectively.

Starting a Chat

1. Click New Chat in the sidebar (or the + button on mobile)
2. Type your message and press Enter (or click Send)

Your chat is saved automatically as you go.

The Plus Menu

Click the + button next to the message input to access all chat options:

• Project — Assign the chat to a project before sending (the chat will be created in that project)
• Load composition — Insert a saved composition into your message
• Model — Switch between AI models
• Assistant — Choose a custom AI personality
• Style — Adjust the AI's communication tone
• Attach file — Upload files from your device
• File library — Select from your saved files
• Online mode — Toggle web search capability
• Thinking mode — Toggle extended reasoning (Pro)

On desktop: Hover over items to reveal flyout submenus. On mobile: Tap items to slide into the submenu, tap the back arrow to return.

When you select a project, a pill appears next to the input showing the project name. Click the pill to remove the project selection.

Choosing an AI Model

PromptBlocks gives you access to multiple AI models. Pick the one that fits your task.

Premium Models

These are the most capable models—great for complex tasks, detailed writing, and nuanced reasoning. We include the latest from OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini), and Grok.

Premium models are marked with a star in the dropdown.

Basic Models

These are faster and work well for quick questions, simple tasks, and everyday use. They're efficient without sacrificing quality for straightforward needs.

Which Should You Use?

Using Prompts in Chat

Type @ to insert a saved prompt into your message.

1. Type @ in the message box
2. Start typing the prompt name to filter
3. Use arrow keys to navigate, Enter to select
4. The prompt appears as a colored badge
5. Click the badge to fill in any variables

This is much faster than retyping prompts, and it keeps your messages consistent.

Loading Compositions

Click the Compositions button (or use the + menu) to load a saved composition.

1. Select a composition from the list
2. It loads into your message box
3. Fill in any variables by clicking the prompt badges
4. Send when ready

Learn more about compositions →

Special Modes

Online Mode

Click the globe icon to let the AI search the web. This is useful when you need current information that the AI might not have in its training data.

When online mode is active, you'll see an indicator in the input area.

Thinking Mode

Click the brain icon to enable extended reasoning. The AI will "think through" complex problems step by step, showing you its reasoning process.

This is especially helpful for:

• Math and logic problems
• Complex code debugging
• Multi-step analysis
• Strategic planning

Auto-enabled for budget models: When using budget models (Gemini Flash, GPT-5 Mini, DeepSeek, Llama Maverick), thinking mode is automatically turned on. This gives you premium-quality reasoning at budget prices. The Settings panel shows an "Auto" badge when this is active.

Availability:

• Budget models: Available to all users (auto-enabled by default)
• Premium models: Requires Pro subscription or BYOK

You can toggle thinking off via the Settings panel if you prefer faster responses without the reasoning step.

Assistants and Styles

Choosing an Assistant

Select an assistant from the dropdown to change how the AI responds. Assistants have custom personalities defined by system prompts you create.

Example: Select "Technical Writer" for documentation tasks, "Creative Partner" for brainstorming.

Choosing a Style

Select a communication style to adjust the AI's tone:

• Concise — Brief, focused responses
• Detailed — Comprehensive explanations
• Friendly — Casual and approachable

You can combine any assistant with any style.

Set up assistants and styles →

Voice Input

Click the microphone icon to dictate your message instead of typing.

1. Click the mic button
2. Speak your message
3. Your words appear as text in the message box
4. Edit if needed, then send

Voice input uses your browser's speech recognition. It works best in Chrome.

Attaching Files

Drag and drop files onto the chat, or click the attachment button.

Supported files:

• Images (PNG, JPG, GIF, WebP)
• PDFs
• Text files (.txt, .md)
• Documents (.docx)

Files are processed and included in your message context, so the AI can reference them.

Chat History

Your previous conversations appear in the sidebar.

• Click any chat to continue it
• Right-click for options (rename, move to project, delete)
• Use the search on the All Chats page to find old conversations

Chats save automatically—you never lose your work.

Archiving Chats

Archive chats to declutter your sidebar and project views while keeping them accessible:

How to Archive

• From any context menu: Click the three-dots menu on a chat (sidebar, header, /chats page, or project page) and select "Archive"
• To unarchive: Open an archived chat and click "Unarchive" in the context menu

Where Archived Chats Appear

Archive Filter (/chats page)

Use the dropdown on the /chats page to filter by archive status:

• Active (default): Shows only non-archived chats
• All: Shows both active and archived chats
• Archived: Shows only archived chats

Your filter preference is saved automatically. Archived chats are fully functional—you can send messages, rename, export, and more.

Bookmarkable Chat Links

Each chat has a unique URL you can bookmark:

• URL format: /chat?id=abc-123
• Copy the URL from your browser's address bar while viewing any chat
• The URL updates automatically when you switch chats

Export to Markdown

Export any chat to a Markdown file for sharing or archiving:

1. Click the three-dots menu on any chat
2. Select "Export to Markdown"

What's included:

• Chat title and URL back to the original
• Creation date and project name (if any)
• Full conversation with speaker info preserved

Stop Streaming

Stop an AI response at any time:

• While the AI is responding, the Send button becomes a Stop button
• Click it to stop the response
• Partial content is saved with a note that it was stopped

Draft Recovery

Your unsent messages are automatically saved as you type:

• Every 2 seconds, your current message is saved to your browser
• Return to a chat and your draft is restored
• Each chat has its own independent draft
• Drafts expire after 7 days

What's protected:

• Page refresh
• Navigation away and back
• Layout switches
• Tab close and reopen

Reliable Streaming

AI responses continue even if you leave:

• Responses continue processing on the server if you refresh or navigate away
• Return mid-response and it automatically reconnects
• You won't lose a response just because you switched tabs

Header Bar

The chat header shows quick action buttons for fast access:

• Settings — Open the settings panel (model, assistant, style)
• Memory — View and manage project memories (only for project chats)
• Prompts — Browse and insert prompts from your library
• Compositions — Browse and load saved compositions
• Chat Menu (⋮) — Rename, Make a copy, Move to project, Delete

On mobile, all options are combined into a single menu.

Quick Actions Panel

The Quick Actions panel (icons on the right side of the screen) gives you fast access to:

• Settings — Change model, assistant, or style for this chat
• Prompts — Browse and insert prompts
• Compositions — Load compositions
• Project Info — Edit project details inline (on project pages)

On mobile, access Quick Actions through the "..." menu in the chat header.

Project Chats

When you chat inside a project, the AI automatically knows about:

• The project's description
• The names of attached files

This gives the AI context without you having to explain everything. The project name appears in the chat header with a link back to the project.

Learn more about projects →

Message Headers

Each AI message shows contextual information in its header:

Format: PromptBlocks · ProjectName (Assistant / Style)

Per-message tracking: When you switch a chat between projects, old messages retain their original project context while new messages show the current project. This makes it easy to see exactly what context the AI had for each response.

Keyboard Shortcuts

Tips for Better Chats

1. Be specific — Clear instructions get better results
2. Use prompts — Don't retype; use your library
3. Try different models — Some tasks work better with certain models
4. Load compositions — Use saved workflows for structured tasks
5. Enable thinking mode — For complex problems that need step-by-step reasoning

Related

• Build your prompt library →
• Create compositions →
• Organize with projects →
• Understanding plans and limits →