An ultra-lightweight, high-performance autonomous coding agent for VS Code powered by Google AI Studio models (Gemini 2.5 Flash, 2.5 Pro, 2.0 Flash Thinking).
Unlike traditional agents that consume dozens of API calls and thousands of tokens per single edit, Gemini Lean Chat follows the "1 Request = 1 Response" philosophy: it modifies files directly, computes concise diffs, and manages version control in a single interaction.
✨ Key Features
⚡ Zero-Loop / Token-Saver: No infinite agentic loops. Makes a single API call using only the exact tokens required.
🛠️ Autonomous File Modifications: Writes, creates, moves, or deletes files automatically in your workspace using an XML tag protocol.
👁️ Compact Diff Viewer: Cleanly displays removed lines (-) in red and added lines (+) in green with smooth preview containers.
🕒 Git-like Time Machine:
Every prompt generates a version snapshot (v1.0.0, v1.0.1, ...).
Pin important milestones with Restore Point badges (🚩).
Instant workspace rollback: orphaned files created in subsequent versions are cleaned up automatically.
Rolled-back subsequent messages are automatically deactivated and skipped in future API calls to save context tokens.
🖼️ Multimodal Image Support: Attach UI mockups, screenshots, or diagrams via drag & drop, file explorer, or clipboard paste (Ctrl+V).
📎 File & Workspace Context: Easily reference specific files and folders with absolute directory paths to prevent ambiguity.
⚖️ Instant Model Switching: Switch seamlessly between available Gemini models directly from the status bar pill.
🚀 Getting Started
Open the Gemini Lean Chat side bar view.
Click the settings icon (⚙️) in the top right header.
Enter your Google AI Studio API Key (obtainable for free from Google AI Studio).
Select your preferred model and start building!
🔒 Security & Privacy
Your API Key is stored securely inside VS Code's native Secret Storage.
Terminal commands (bash/cmd) are never executed automatically—they feature a dedicated ▶️ Run button giving you full manual control.
