​

The Coding Studio Workbench

​

Studio Interface Overview

+-----------------------------------------------------------------------------------+
| 💻 {repoOwner}/{repoName} ⌄   GitHub Workspace                Press ⌘ + Enter to run|
|-----------------------------------------------------------------------------------|
|                                                                                   |
|  Plan, Build, / for skills, @ for context...                                      |
|                                                                                   |
|  (+) [ Default Engine ⌄ ]   [ Interactive plan ⌄ ]                       (🎤) [Start]|
+-----------------------------------------------------------------------------------+

​

1. Top Context Bar & Repository Selector

• Repository Selector: Click the repository dropdown in the top-left corner to switch between linked repositories. You can search repositories or link a new one via + Add Repo.
• Instant Switching: Switching repositories is instantaneous and smoothly updates your session list with zero page reloads.
• Keyboard Shortcut: Press ⌘ + Enter (or Ctrl + Enter) anywhere inside the prompt box to launch your session immediately.

​

2. Prompt Editor & Attachments

• Multi-Image Attachments (+ button): Upload up to 3 images per prompt (screenshots of UI bugs, Figma mockups, wireframes, or error logs). The AI analyzes the visuals to match your exact UI requirements.
• Voice Dictation (🎤 button): Click the microphone icon to dictate prompts hands-free using real-time speech recognition.
• Action Row Starters: Quick-start buttons below the input box to launch common tasks with one click:
  • 📋 Start with a plan: “Align on implementation before writing code”
  • 🐛 Debug an issue: “Find root causes and fix tricky bugs”
  • 🎨 Create a dashboard: “Visualize information in a custom interface”

​

3. Slash Commands

​

4. The 3 Execution Modes

​

1. ✦ Interactive Plan (interactive_plan — Default)

• Best for: Complex features, multi-file architecture changes, or when you want to brainstorm.
• How it works: Your character holds a conversational chat with you first. It explains its proposed architecture, asks clarifying questions if anything is ambiguous, and refines the plan based on your feedback before proceeding to execution.

​

2. 📄 Review (plan_review)

• Best for: Structured tasks where you want to verify the planned steps before any code is generated.
• How it works: The character analyzes the codebase and generates a complete, numbered implementation plan. Execution halts and awaits your one-click approval (Approve Plan or Request Changes).

​

3. 🚀 Direct (direct)

• Best for: Quick bug fixes, small utility functions, documentation, or routine edits.
• How it works: The AI skips the approval gate and immediately begins analyzing, writing code, and testing.

​

5. Model Switcher (EngineSwitcher)

• By default, it is set to Default Engine (from Settings), which automatically uses the engine configured in your character settings.
• You can override the engine for a specific session by choosing from specialized models (e.g. GLM 5.2, NVIDIA Nemotron, Meta Llama 3.3, MiniMax M3, OpenAI GPT-OSS 120B).
• Note: Switching the engine in the studio toolbar only affects code generation for that session and never alters your character’s global settings.

​

6. Session History & Management

• Filter Tabs: Quickly view All Sessions, Active (currently running), Review (awaiting plan approval), or Completed sessions.
• Search Bar: Search past sessions by prompt keywords, branch name, or commit SHA.
• Three-Dots Context Menu (···):
  • 📌 Pin / Unpin: Keep critical sessions pinned to the top of your list.
  • ✏️ Rename: Edit the session title or prompt description.
  • 🗑️ Delete: Permanently delete old session transcripts (with a safety confirmation if the session is currently executing).