Aevi — AI Coding Assistant

Aevi is an AI-powered coding assistant built into VS Code. It supports local LLMs via Ollama, vLLM as well as cloud APIs (OpenAI, Anthropic, Groq, Gemini), giving you full control over your model and your data.

Features

• Chat — ask questions about your code with full file context
• Agent — autonomously make changes to your codebase with accept/reject diffs
• Inline completions — AI suggestions as you type (Beta)
• RAG — semantic search over your workspace for accurate context
• Multi-provider — Ollama, OpenAI, Anthropic, Groq, Gemini, LM Studio, llama.cpp, vLLM
• Model picker — switch models instantly from the sidebar
• Privacy-first — run entirely locally with Ollama, no data leaves your machine

Requirements

Aevi requires python3.11 or above.

The server runs on http://127.0.0.1:8765 by default.

Setup

Local models (Ollama)

1. Install Ollama
2. Pull a model: ollama pull qwen2.5-coder:7b
3. Open Aevi settings in the sidebar and set the Ollama URL (default: http://localhost:11434)

Cloud APIs

Open the ⚙ settings panel in the Aevi sidebar and enter your API keys:

Keys are stored securely using VS Code's built-in secret storage.

Usage

Chat mode

Ask questions about your code. The currently open file is automatically added as context. Use + Add file to pin additional files.

Agent mode

Describe a task and Aevi will plan and execute it step by step, showing you a diff before applying any change. You can accept or reject each edit.

Inline completions

Enable via the ⚙ settings toggle. Works best with qwen2.5-coder or codellama models.

Index workspace

Run Aevi: Index Workspace from the command palette to enable semantic search (RAG) across all your project files.

Extension Settings

Supported Models

Local: Ollama, LM Studio, llama.cpp, vLLM

Cloud: OpenAI, Anthropic, Groq, Gemini

Model lists are fetched live from each provider — you see exactly what your API key has access to.

Troubleshooting

If you encounter a "Could not reach Aevi backend" error, or if the Python environment failed to install correctly the first time, you can force the extension to run the setup process again.

How to Re-trigger the Setup Process

If the backend isn't running, the best first step is to completely wipe the corrupted environment and start fresh.

1. Run the Reinstall Command:

   • Open the VS Code Command Palette (Cmd+Shift+P on Mac, Ctrl+Shift+P on Windows/Linux).
   • Type and select Aevi: Reinstall Backend.
   • Note: This will safely delete the hidden ~/.aevi folder and the installation marker.

2. Restart VS Code:

   • Close and reopen VS Code entirely, or run the Developer: Reload Window command from the Command Palette.

3. Run the Setup:

   • Click the Aevi icon in your sidebar.
   • Because the environment was cleared, you will see the initial "Setup Aevi" screen again.
   • Click the button to download the dependencies and rebuild the Python backend.

Still having issues?

• Check Python Version: Ensure you have Python 3.11 or higher installed on your system and accessible in your system's PATH.
• Port Conflicts: Aevi uses port 8765 by default. If another app is using this port, you can change it in your VS Code settings by searching for aevi.backendUrl and updating it to a different port (e.g., http://127.0.0.1:8888).

License

MIT