@anygpt/cody

Sourcegraph Cody connector for AnyGPT with flexible API and CLI modes.

Features

• 🚀 Dual-mode - API (fast) or CLI (feature-rich) with auto-fallback
• 🔄 Multiple models - Claude, GPT-4, Gemini, and more
• 🎯 Enterprise ready - Works with any Sourcegraph instance
• 🔧 Auto-config - Reads from ~/.config/Cody-nodejs/config.json

Installation

npm install @anygpt/cody

Prerequisites:

• Access to a Sourcegraph instance with API token
• API mode (default): No CLI installation required
• CLI mode: Requires @sourcegraph/cody CLI installed

Quick Start

import { createCodyConnector } from '@anygpt/cody';

// Auto-reads from ~/.config/Cody-nodejs/config.json
const connector = await createCodyConnector();

const response = await connector.chatCompletion({
  model: 'anthropic::2024-10-22::claude-3-5-sonnet-latest',
  messages: [{ role: 'user', content: 'Hello, Cody!' }]
});

With Configuration

const connector = await createCodyConnector({
  connectionMode: 'api',  // 'api' | 'cli' | 'auto'
  endpoint: 'https://sourcegraph.example.com/',
  accessToken: process.env.SRC_ACCESS_TOKEN
});

Connection Modes

Choose between three modes:

| Mode | Speed | Best For | |------|-------|----------| | api (default) | ⚡ Fast | Production, CI/CD | | cli | 🐢 Slower | Local dev, debugging | | auto | ⚡ Fast* | Resilient setups |

// API mode (default)
const connector = await createCodyConnector({
  connectionMode: 'api'
});

// CLI mode (requires Cody CLI installed)
const connector = await createCodyConnector({
  connectionMode: 'cli',
  showContext: true
});

// Auto mode (tries API, falls back to CLI)
const connector = await createCodyConnector({
  connectionMode: 'auto'
});

📖 Detailed Mode Documentation

Configuration

| Option | Type | Default | |--------|------|---------| | connectionMode | 'api' \| 'cli' \| 'auto' | 'api' | | endpoint | string | 'https://sourcegraph.com/' | | accessToken | string | Auto-detected | | timeout | number | 60000 | | model | string | - |

CLI-specific (for cli and auto modes):

• cliPath, workingDirectory, showContext, debug

💡 Auto-reads from ~/.config/Cody-nodejs/config.json

Models

Models are dynamically fetched from your Sourcegraph instance:

const models = await connector.listModels();

Common models: Claude Sonnet 4, Claude 3.5 Sonnet, GPT-4o, GPT-5, Gemini 2.5 Pro, DeepSeek v3

Authentication

Via environment variables:

export SRC_ACCESS_TOKEN="sgp_your-token"
export SRC_ENDPOINT="https://sourcegraph.example.com/"

Or in config:

const connector = await createCodyConnector({
  accessToken: 'sgp_your-token',
  endpoint: 'https://sourcegraph.example.com/'
});

Examples

See examples/connection-modes.ts for working examples of all modes.

Config examples: examples/cody.config.ts

Troubleshooting

CLI mode fails:

• Install CLI: npm install -g @sourcegraph/cody
• Verify: which cody
• Auth: cody auth

API mode fails:

• Check token: curl -H "Authorization: token TOKEN" https://instance/.api/llm/models
• Verify endpoint ends with /

Slow responses:

• Use API mode (fastest)
• Check network latency

📖 Detailed Troubleshooting

Documentation

• Connection Modes - Detailed mode comparison and configuration
• Architecture - Technical implementation details
• Examples - Working code examples
• Changelog - Version history

License

MIT