opensync

# OpenSync - AI Coding Session Dashboard

> Sync, search, and share your AI coding sessions from OpenCode and Claude Code.

## Overview

OpenSync is a cloud-based dashboard that syncs, stores, and searches AI coding sessions. Built on Convex (real-time database) and WorkOS (authentication).

## Ecosystem

| Project | Description | Links |
|---------|-------------|-------|
| OpenSync | Dashboard for AI coding sessions | [Website](https://www.opensync.dev) / [GitHub](https://github.com/waynesutton/opensync) |
| opencode-sync-plugin | Sync OpenCode sessions | [GitHub](https://github.com/waynesutton/opencode-sync-plugin) / [npm](https://www.npmjs.com/package/opencode-sync-plugin) |
| claude-code-sync | Sync Claude Code sessions | [GitHub](https://github.com/waynesutton/claude-code-sync) / [npm](https://www.npmjs.com/package/claude-code-sync) |

## Quick Start

### Install Plugin

For OpenCode:
```bash
npm install -g opencode-sync-plugin
opencode-sync login
```

For Claude Code:
```bash
npm install -g claude-code-sync
claude-code-sync login
```

### Self-Host

```bash
git clone https://github.com/waynesutton/opensync.git
cd opensync
npm install
npx convex dev
```

## Features

- Auto sync: Sessions sync in real-time as you work
- Full-text search: Search by keywords across all sessions
- Semantic search: Search by meaning using vector embeddings
- Hybrid search: Combines full-text and semantic for best results
- Public sharing: Share sessions with a single click
- Markdown export: Download sessions as Markdown files
- API access: Secure API for external integrations
- Usage stats: Track tokens, cost, time per session
- RAG support: Built-in retrieval for context engineering
- Eval export: Export datasets for DeepEval, OpenAI Evals

## API Endpoints

All endpoints require Bearer token authentication (API key starting with `osk_` or JWT).

| Endpoint | Method | Description |
|----------|--------|-------------|
| /api/sessions | GET | List all sessions |
| /api/sessions/get | GET | Get session with messages |
| /api/search | GET | Search sessions (fulltext/semantic/hybrid) |
| /api/context | GET | Get relevant context for LLM |
| /api/export | GET | Export session (json/markdown/jsonl) |
| /api/stats | GET | Get usage statistics |
| /health | GET | Health check (no auth) |

### Search Types

- `?type=fulltext` - Keyword matching, fast and exact
- `?type=semantic` - Meaning-based search using embeddings
- `?type=hybrid` - Combines both using Reciprocal Rank Fusion

## Authentication

### Web Dashboard
Uses WorkOS AuthKit with SSO, magic link, and OAuth support.

### Plugins and API
Use API keys (osk_*) generated in Dashboard Settings. No browser required.

## Data Schema

### Sessions
- userId, externalId, title, projectPath, projectName
- model, provider, source (opencode/claude-code)
- promptTokens, completionTokens, totalTokens, cost
- durationMs, isPublic, publicSlug
- evalReady, evalNotes, evalTags

### Messages
- sessionId, externalId, role (user/assistant/system)
- textContent, model, promptTokens, completionTokens
- durationMs, parts (tool calls, code blocks)

## Tech Stack

- Backend: Convex (real-time database, serverless functions, vector search)
- Auth: WorkOS AuthKit (SSO, magic link, OAuth)
- Frontend: React + Vite + Tailwind CSS
- Embeddings: OpenAI text-embedding-3-small
- Hosting: Netlify (frontend), Convex Cloud (backend)

## Environment Variables

### Frontend (Vite)
- VITE_CONVEX_URL - Convex deployment URL
- VITE_WORKOS_CLIENT_ID - WorkOS client ID

### Backend (Convex)
- OPENAI_API_KEY - For semantic search embeddings
- WORKOS_API_KEY - For user management

## Privacy

- All data goes to YOUR Convex deployment
- Sensitive fields (passwords, tokens, keys) are redacted before sync
- Full file contents are not synced, only paths and lengths
- Thinking traces are off by default
- You control what gets synced via configuration

## Resources

- Convex Docs: https://docs.convex.dev
- WorkOS Docs: https://workos.com/docs
- Convex + WorkOS: https://docs.convex.dev/auth/authkit
- Convex Vector Search: https://docs.convex.dev/search/vector-search

## License

MIT