HN 표시: Multibot – Cloudflare Workers의 오픈 소스 다중 봇 AI 플랫폼
hackernews
|
|
📦 오픈소스
#ai 딜
#ai 플랫폼
#anthropic
#cloudflare workers
#openai
#멀티봇
#서버리스
#오픈 소스
원문 출처: hackernews · Genesis Park에서 요약 및 분석
요약
클라우드플레어 엣지를 기반으로 서버 관리 없이 여러 AI 봇을 생성 및 배포할 수 있는 오픈소스 플랫폼 '멀티봇(Multibot)'이 공개되었습니다. 이 플랫폼은 웹 대시보드를 통해 텔레그램, 디스코드, 슬랙 등과의 연동을 지원하며, 사용자가 직접 LLM 공급자의 API 키를 등록하여 활용할 수 있는 'BYOK(Bring Your Own Key)' 방식을 채택했습니다. 또한 1개 채팅방당 최저 $5 수준의 저렴한 비용으로 봇 샌드박스, 하위 에이전트 생성, 음성 인식 및 합성 등 다양한 기능을 제공합니다.
본문
A fully serverless, multi-bot AI platform that runs on Cloudflare edge. Create, configure, and manage multiple AI bots through a web dashboard -- connect them to Telegram, Discord, or Slack with one click. - Fully Serverless: Runs on Cloudflare edge (Workers + Durable Objects), scales automatically, one-click deploy - Low Cost: Starting at $5/month with pay-as-you-go pricing, no DevOps overhead - Multi-Bot Group Chat: Multiple AI bots collaborate in group conversations, coordinated by an orchestrator - Memory: Two-layer architecture (MEMORY + HISTORY) with LLM-driven consolidation across conversations - Skills: Markdown-driven skill system with progressive loading, extend bot capabilities without code changes - Voice: Speech-to-Text (Whisper) and Text-to-Speech (OpenAI TTS), talk to your bots naturally - Scheduling: One-shot, interval, and cron expressions with timezone support | Bot Management | Bot Configuration | Group Chat | Settings | |---|---|---|---| - Node.js 20+ - Cloudflare account with Workers Paid plan ($5/mo for Durable Objects & Cron Triggers) - Fly.io Sprites API token (for bot sandbox -- shell & file tools) git clone https://github.com/codance-ai/multibot.git cd multibot ./scripts/setup.sh # installs deps, logs in to CF, creates all resources npm run deploy # builds dashboard + deploys worker setup.sh handles everything: installs dependencies, logs in to Cloudflare (opens browser), creates D1/R2 resources, auto-detects your account email and workers.dev subdomain, and generates wrangler.toml (gitignored). Open https://multibot..workers.dev : - Add LLM API keys -- Settings page, supports multiple providers - Create a bot -- name, system prompt, model, skills - Bind a channel -- connect Telegram, Discord, or Slack Set DASHBOARD_PASSWORD and OWNER_ID as Worker secrets to protect the dashboard with password authentication: echo "your-password" | npx wrangler secret put DASHBOARD_PASSWORD echo "your-owner-id" | npx wrangler secret put OWNER_ID npm run dev # local dev server npm test # run all tests npm run deploy # build dashboard + deploy worker Telegram ──┐ Discord ──┼── Gateway (Worker) ──► Durable Object (Agent) ──► LLM Provider Slack ──┘ token → botId per conversation OpenAI / Anthropic / Google / ... chatId → agent memory, tools, skills │ Sandbox (Sprites) shell, filesystem | Concept | Description | |---|---| | User | Platform user, owns bots, manages LLM API keys | | Bot | AI assistant with its own prompt, model, skills config | | Channel | Chat platform binding (Telegram / Discord / Slack) | | Agent | Runtime instance (DO), one per {botId}-{channel}-{chatId} | | Sandbox | Linux sandbox, one per bot, for shell/file tools (Fly.io Sprites) | MIT
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유