Your First Chat
This page will guide you step-by-step through your first AI chat from scratch, including configuring a provider, selecting a model, sending a message, and experiencing core features like branches and regeneration.
Prerequisites
- Elftia has been successfully installed and started (see Download and Install)
- You have at least one LLM provider API Key (such as OpenAI, Anthropic, Google Gemini, DeepSeek, etc.)
If you don't have an API Key yet, you can register an account with each provider and create one:
- OpenAI: platform.openai.com
- Anthropic: console.anthropic.com
- Google Gemini: aistudio.google.com
- DeepSeek: platform.deepseek.com
Step 1: Open Settings
After starting Elftia, click the gear icon at the bottom of the left Workspace Rail to open the Settings page.
You can also use a keyboard shortcut to quickly open Settings (the specific shortcut depends on your system configuration).
Step 2: Add an LLM Provider
- In the Settings page, find the Provider Settings tab.
- In the list on the left, you will see all supported LLM providers (50+). Select the provider you want to use, such as "OpenAI".
- In the details area on the right:
- API Key input field: Paste your API Key.
- If the provider supports custom endpoints (such as third-party services compatible with OpenAI format), you can modify the Base URL.
- Click the Test Connection button.
- Success: A green "Connection successful" message appears, indicating your API Key is valid.
- Failure: A red error message appears. Please check if your API Key is correct, or refer to Connection Errors for troubleshooting.
- After confirming the test passes, toggle the Enabled switch to activate the provider.
Expected Result: In the provider list, enabled providers will display an enabled status indicator.
You can enable multiple providers at the same time and freely switch between different models during chat.
Step 3: Start a New Chat
- Click the chat icon (chat bubble) on the left Workspace Rail to switch to the chat view.
- Click the New Chat button (+ icon) at the top of the sidebar to create a new chat session.
- In the content area, you will see a blank chat interface, ready to receive your messages.
Step 4: Select a Model
In the toolbar above the input box, find the Model Selection dropdown:
- Click the model dropdown to expand the list of available models.
- Models are displayed grouped by provider. Select a model, such as
gpt-4o(OpenAI) orclaude-sonnet-4-20250514(Anthropic). - The selected model name will be displayed in the toolbar.
Expected Result: The toolbar displays the currently selected model name, indicating that the model is ready.
Unsure which model to choose? For everyday conversations, it's recommended to use mid-tier models from each provider (such as GPT-4o, Claude Sonnet, Gemini 2.0 Flash), which balance quality and speed.
Step 5: Send Your First Message
- In the input box at the bottom, type your message, for example: "Hello, please introduce yourself."
- Press Enter to send the message (with the default setting, Enter sends directly; if you prefer to use Ctrl+Enter, you can adjust this in Settings → General).
- After the message is sent, you will see the AI's reply displayed character by character in a streaming manner.
Expected Result: The AI reply appears smoothly and stops updating after the content is complete. If the reply contains code or Markdown formatting, it will automatically render as rich text.
Step 6: Experience Regeneration
If you are not satisfied with the AI's reply, you can regenerate it:
- Hover your mouse over the AI's reply message, and action buttons will appear at the bottom of the message.
- Click the Regenerate button (refresh icon).
- The AI will generate a new reply.
Key Concept — Branches: Regenerating does not overwrite the original reply, but creates a new "branch". You can see the branch navigator (left and right arrows) in the message area to switch between different reply versions.
Expected Result: The message area displays the new reply, the branch navigator appears (e.g., 1/2), and you can click the arrows to switch between different versions.
Step 7: Experience Editing Messages
You can also edit messages you've already sent:
- Hover your mouse over your sent message and click the Edit button (pencil icon).
- The message enters edit mode. After modifying the content, click Save or press Enter.
- Elftia will regenerate the AI reply based on the edited message.
Like regeneration, editing a message also creates a new branch, and the original conversation history remains unchanged.
Explore More Features
After completing your first chat, you can continue exploring the following features:
Branch Navigation
- On any message with multiple versions, use the left and right arrows of the branch navigator to switch between different versions.
- Branches allow you to explore different conversation directions without losing history.
Multi-Tab Conversations
- In the title bar, click the + button to open a new conversation tab.
- You can conduct multiple conversations simultaneously and quickly switch between tabs.
Attachments
In the input box, you can add multiple types of attachments:
- Images: Drag and drop or click the attachment button to upload images. The AI will recognize the image content (requires the model to support vision capabilities).
- Files: Upload documents for the AI to analyze.
- Code Snippets: Paste code for analysis and suggestions.
Web Search
In the toolbar, find the Web Search toggle. When enabled, the AI will search the internet for the latest information before replying, which is useful for asking about current events and the latest technology news.
Conversation Management
- Rename: Right-click a conversation in the sidebar and select "Rename".
- Organize with Folders: Create folders and drag conversations into them for categorization.
- Search: Use the search box at the top of the sidebar to quickly find past conversations.
- Tags: Add tags to conversations for easy filtering.
Congratulations on completing your first chat! Next, you can read Interface Navigation to get a comprehensive overview of Elftia's interface layout, or jump directly to Feature Documentation to explore the various features in depth.