@ai-ide-bridge/core

The core HTTP server, session management, and shared infrastructure for AI IDE Bridge.

AI IDE Bridge is a local HTTP server that translates OpenAI-compatible API requests into provider-specific calls (Cursor SDK, GitHub Copilot, Windsurf), enabling any OpenAI-format client to use any AI IDE's model catalog.

Overview

@ai-ide-bridge/core provides the foundational building blocks for creating AI IDE bridge plugins:

• HTTP Server: Native Node.js server handling /health, /v1/models, and /v1/chat/completions.
• OpenAI Parser & Formatter: Zod-based request validation and SSE stream formatting.
• Plugin Registry & Model Router: Manages active plugins and routes requests by model ID prefixes.
• Session Management: In-memory session tracking with TTL cleanup.
• Daemon Manager: Infrastructure for managing local daemon binaries (stdio/JSON-RPC communication).

Installation

npm install @ai-ide-bridge/core

Creating a Plugin

Plugins implement the BridgePlugin interface:

import type {
  BridgePlugin,
  BridgeSession,
  ModelInfo,
  Message,
  ToolDefinition,
  StreamChunk,
} from '@ai-ide-bridge/core';

export class MyBridgePlugin implements BridgePlugin {
  name = 'my-provider';
  version = '1.0.0';

  async authenticate(config: Record<string, string>): Promise<boolean> {
    // Validate credentials
    return true;
  }

  async listModels(config: Record<string, string>): Promise<ModelInfo[]> {
    // Return available models
    return [];
  }

  async createSession(config: Record<string, string>, model: string): Promise<BridgeSession> {
    // Return a streaming session
    return new MyBridgeSession();
  }
}

Documentation

For full documentation, plugin development guides, and architecture details, please visit the main repository: https://github.com/aeswibon/llm-bridge.

License

MIT