AI Chat Interface

The AI Chat interface is the heart of TalkCody, providing an interactive conversational experience with AI models. Chat with AI to get coding help, review code, generate tests, and more.

Overview

TalkCody's chat interface offers:

• Streaming Responses: Real-time text generation as the AI formulates answers
• Multi-Turn Conversations: Context-aware dialogues that remember previous messages
• Code Highlighting: Syntax-highlighted code blocks with language detection
• Tool Execution: AI can execute file operations, searches, and web crawling
• Attachments: Upload images or files for AI analysis
• Voice Input: Hands-free input with voice-to-text
• Message Compression: Automatic context management for long conversations

Starting a Conversation

Chat Features

Streaming Responses

TalkCody uses streaming to display AI responses in real-time:

User: Explain async/await in JavaScript

AI: Async/await is a modern JavaScript syntax for handling
asynchronous operations. It makes asynchronous code look
and behave like synchronous code...
[Response continues to stream in real-time]

Code Blocks

The AI automatically formats code with syntax highlighting:

User: Write a React component for a counter

AI: Here's a simple counter component:

```jsx
import { useState } from 'react';

export function Counter() {
  const [count, setCount] = useState(0);

  return (
    <div>
      <p>Count: {count}</p>
      <button onClick={() => setCount(count + 1)}>
        Increment
      </button>
    </div>
  );
}
```

Code block features:

• Automatic language detection
• Syntax highlighting for 100+ languages
• Copy button for quick copying
• Line numbers (optional)

File Attachments

Attach files or images to your conversations:

1. Click the 📎 Attach button
2. Select a file from your computer
3. The AI can analyze:
   • Images (screenshots, diagrams, UI mockups)
   • Code files
   • Text documents
   • Configuration files

Example use cases:

[Attach screenshot of error]
User: What's causing this error?

[Attach diagram]
User: Generate code based on this architecture diagram

[Attach config file]
User: Review this configuration and suggest improvements

Tool Execution

AI agents can execute tools to assist you:

Available Tools:

• File Operations: Read, write, edit files in your project
• Code Search: Search through your codebase
• Bash Commands: Execute terminal commands
• Web Search: Search the internet for information
• Web Crawling: Fetch and analyze web pages

Example:

User: Find all TODO comments in my project

AI: I'll search for TODO comments in your project.
[Executes code search tool]
[Displays results from the search]
Found 15 TODO items across 8 files:
- src/app.tsx:45 - TODO: Implement error handling
- src/utils/api.ts:120 - TODO: Add retry logic
...

Voice Input

Use voice-to-text for hands-free input:

1. Click the 🎤 Microphone icon
2. Speak your message
3. The text appears in the input field
4. Review and send

Tips for better voice recognition:

• Speak clearly and at a normal pace
• Use a quiet environment
• Pronounce technical terms carefully
• Review transcription before sending

Context Management

Message History

TalkCody maintains conversation context automatically:

• Previous messages are included in context
• AI remembers earlier parts of the conversation
• Code and file references are preserved

Example of context awareness:

User: Write a function to calculate factorial

AI: [Provides factorial function]

User: Now add error handling to it

AI: [Adds error handling to the previously written function]

Message Compression

For long conversations, TalkCody automatically compresses older messages:

• Summarizes earlier messages
• Preserves important context
• Manages token limits effectively
• Keeps recent messages in full detail

Conversation Limits

Different models have different context windows:

Token counts are displayed in the chat interface (enable in Settings).

Advanced Chat Features

Regenerate Responses

Not satisfied with a response?

• Click the ↻ Regenerate button
• The AI generates a new response
• Previous response is preserved in history

Edit Messages

Edit your previous messages:

• Hover over your message
• Click the Edit button
• Modify the text
• The conversation branches from that point

Copy and Share

• Copy Message: Click the copy button on any message
• Copy Code: Copy code blocks separately
• Export Conversation: Export the full conversation as markdown

Keyboard Shortcuts

Working with Projects

When you have a project open, the AI has enhanced capabilities:

File Context

The AI can:

• Read files from your project
• Search through your codebase
• Understand project structure
• Reference multiple files

Example:

User: Refactor the authentication logic in auth.ts

AI: [Reads auth.ts from your project]
I've reviewed your authentication code. Here are my suggestions:
[Provides specific refactoring advice based on actual code]

File Editing

AI agents can propose file edits:

1. AI suggests changes to a file
2. Review dialog shows the proposed changes
3. You can accept, reject, or modify the changes
4. Changes are applied to the actual file

Multi-File Operations

The AI can work across multiple files:

User: Create a new React component with tests and styles

AI: I'll create three files for you:
1. components/Button.tsx - The component
2. components/Button.test.tsx - Unit tests
3. components/Button.module.css - Styles

[Shows preview of all three files]
[Awaits your approval]

Best Practices

Writing Effective Prompts

Be specific:

❌ "Fix my code"
✅ "Fix the TypeError in the fetchData function in api.ts"

Provide context:

❌ "Write a component"
✅ "Write a React component for a user profile card that displays
    name, avatar, and bio. Use TypeScript and Tailwind CSS."

Break down complex tasks:

Instead of: "Build a complete authentication system"

Try:
1. "Design the database schema for user authentication"
2. "Implement user registration with email validation"
3. "Create login endpoint with JWT tokens"
4. "Add password reset functionality"

Managing Long Conversations

• Start new conversations for unrelated topics
• Use message compression for extended discussions
• Export important conversations for reference
• Periodically summarize progress with the AI

Choosing the Right Model

Different models excel at different tasks:

• Code Generation: Qwen 3 Coder, Codestral, GPT-4.1
• Explanations: Claude 4.5 Sonnet, GPT-4.1
• Quick Questions: Claude Haiku, GPT-4.1 Turbo
• Complex Reasoning: Claude 4.5 Opus, GPT-5
• Budget-Friendly: DeepSeek Chat, GLM 4.5 Air

Troubleshooting

AI Not Responding

• Check your internet connection
• Verify API key is valid
• Check provider status page
• Try a different model

Responses Are Cut Off

• Increase max tokens in Settings
• Use a model with larger context window
• Break request into smaller parts

AI Can't Access My Files

• Verify project is opened in TalkCody
• Check file permissions
• Ensure file tools are enabled for the agent
• Try explicitly providing the file path

Code Formatting Issues

• Specify the programming language in your prompt
• Ask the AI to format code in markdown code blocks
• Copy code separately using the code block copy button

Next Steps

• AI Agents - Create specialized AI assistants
• AI Completions - Use inline code suggestions
• Model Settings - Configure AI models
• Agent Configuration - Customize agent behavior