Editor Code Assistant (ECA) provides AI pair programming capabilities that are agnostic to the editor, aiming to simplify the integration of Large Language Models (LLMs) with various development environments. It offers a unified configuration and chat interface for features like code review, question answering, and agentic task execution, benefiting developers seeking a consistent AI coding experience across different editors.

How It Works

ECA utilizes a protocol inspired by the Language Server Protocol (LSP) for editor communication, with the server implemented in Clojure. Editors communicate with the ECA server via stdin/stdout, enabling features like tool call management, multi-LLM interaction, and telemetry. This approach allows editors to focus on UX while ECA handles LLM integration and configuration, facilitating rapid feature additions and a consistent user experience.

Quick Start & Requirements

• Install editor plugin (Emacs, VSCode, Vim supported; IntelliJ planned).
• Configure at least one LLM model via .eca/config.json with API keys (e.g., OpenAI, Anthropic, Ollama).
• Start chatting in the editor; use /init to create an AGENT.md for project context.
• Documentation: Models, Custom Providers

Highlighted Details

• Editor-agnostic design via a defined protocol.
• Unified configuration for multiple editors.
• Support for multiple LLM providers including local Ollama models.
• Agentic capabilities allowing LLMs to act with configurable tools.
• Context-aware features for providing detailed code information to LLMs.

Maintenance & Community

• Roadmap available at planned work.
• Contributions welcome via issues or pull requests.
• Project sponsorship is encouraged.

Licensing & Compatibility

• The README does not explicitly state the license.

Limitations & Caveats

IntelliJ support is planned but not yet implemented. The project does not specify its license, which may impact commercial or closed-source usage.