Chat Templates

Chat templates seed new conversations with consistent system prompts and a recognizable emoji avatar. Each template stores its name, emoji, system prompt, and whether it inherits the app-level prompt. Templates do not capture tool selections, starter messages, or model preferences.

Create or Import

  • Generate from an active conversation: Open the conversation menu and choose Generate Chat Template. FlowDown uses the current chat model to extract a name, emoji, and prompt from recent exchanges (needs at least one user and one assistant message) and saves it to Settings → Models → Chat Template.
  • Create from scratch: Go to Settings → Models → Chat Template and tap ＋ → Create Template to start with a blank template.
  • Rewrite an existing template: In the template editor, tap the sparkle button to have the default chat model rewrite the prompt based on your instructions.
  • Import: In the template list, choose ＋ → Import from File to load .fdtemplate files (supports multi-select). You can also drag and drop .fdtemplate files into the list or open them from Finder/Share to import; templates with the same identifier are replaced.

Chat template quick actions

AI rewrite workflow

Edit and Manage

  • Inside a template, change the emoji avatar, name, prompt, and whether it inherits the app-level prompt. Save with the checkmark.
  • Use Create Copy to duplicate a template, Export Template to share a .fdtemplate, and Delete Template to remove it. Templates are stored locally and can be reordered by dragging in the list; swipe or context menus also delete entries.

Template configuration and lifecycle

Use Templates

  • Start a conversation via ＋ → Choose Template in the new chat menu (including the floating New Chat button). On macOS, the app menu also offers “New Chat With Template.”
  • Applying a template sets the conversation icon and name, writes the template prompt as a system message, and—if you choose Ignore App Prompt—first clears the default system prompt and notes which template supplied it.

Choose a template when starting a chat

Tips

  • Keep prompts concise so they stay reusable across models.
  • Pair templates with Inference Configuration to pre-allocate reasoning budgets when needed.