OpenCode In-Memoria Plugin

A persistent intelligence plugin for OpenCode that captures, learns, and applies codebase patterns across AI-assisted development sessions.

What This Plugin Does

This plugin integrates with In-Memoria to create an organizational memory system for your codebase. It:

• Automatically learns your codebase patterns, conventions, and architecture when a project opens
• Captures insights from successful tasks, code decisions, and bug fixes
• Provides contextual recommendations based on your project's specific patterns
• Builds persistent intelligence that improves AI responses over time

Quick Start

Install the plugin globally:

npm install -g opencode-plugin-in-memoria@latest

Then configure your .opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "description": "My Project with In-Memoria",
  "mcp": {
    "in-memoria": {
      "type": "local",
      "command": ["npx", "in-memoria", "server"],
      "description": "In-Memoria persistent intelligence server"
    }
  },
  "plugin": [
    "opencode-in-memoria-plugin"
  ]
}

That's it! The plugin works out-of-the-box with sensible defaults and will automatically learn your codebase.

Optional Configuration

To customize behavior, create .opencode/inmemoria.json:

{
  "autoLearn": true,
  "contextInjection": true,
  "insightCapture": true,
  "logTools": false
}

Lifecycle Hooks

The plugin hooks into OpenCode's lifecycle to automatically capture knowledge:

| Hook | What It Captures | |------|------------------| | project.open | Auto-learns codebase patterns and architecture | | ai.response.before | Injects project context and pattern recommendations | | task.complete | Records best practices and decisions from successful work | | ai.error | Captures bug patterns for future prevention | | file.change / file.save | Triggers incremental learning on code changes | | conversation.end | Summarizes session insights | | tools.list | Provides file suggestions based on task context |

Installation

npm install
npm run build

Configuration

Create .opencode/inmemoria.json in your project:

{
  "autoLearn": true,
  "contextInjection": true
}

Usage

Available Tools

inmemoria_quick_setup

Initialize and learn project:

await opencode.execute("inmemoria_quick_setup")

inmemoria_get_context

Get project context:

const ctx = await opencode.execute("inmemoria_get_context")

inmemoria_suggest_files

Get file suggestions:

await opencode.execute("inmemoria_suggest_files", {
  task: "implement auth"
})

inmemoria_search

Search codebase:

await opencode.execute("inmemoria_search", {
  query: "database connection",
  type: "semantic"
})

inmemoria_learn

Force re-learning:

await opencode.execute("inmemoria_learn")

inmemoria_status

Check status:

await opencode.execute("inmemoria_status")

How It Works

1. Pattern Learning: The plugin analyzes your codebase structure, naming conventions, architectural patterns, and common code structures

2. Insight Capture: As you work, it captures:

   • Best practices from successful implementations
   • Bug patterns and their fixes
   • Code organization decisions
   • Performance optimizations

3. Context Injection: Before AI generates responses, the plugin injects relevant project context and pattern recommendations specific to your codebase

4. Persistent Memory: All captured knowledge persists across sessions, building up an organizational memory that improves AI assistance quality over time

Documentation

• Architecture
• API Reference
• Configuration

Development

npm run dev       # Watch mode
npm run build     # Build once
npm run lint      # Run ESLint
npm run typecheck # Type check

License

MIT