经 AI Skill Hub 精选评估,深度翻译工作室 获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 8.0 分,适合有一定技术背景的用户使用。
深度翻译工作室 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
深度翻译工作室 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:npm 全局安装 npm install -g deeptrans-studio # 方式二:npx 直接运行(无需安装) npx deeptrans-studio --help # 方式三:项目依赖安装 npm install deeptrans-studio # 方式四:从源码运行 git clone https://github.com/hint-lab/deeptrans-studio cd deeptrans-studio npm install npm start
# 命令行使用
deeptrans-studio --help
# 基本用法
deeptrans-studio [options] <input>
# Node.js 代码中使用
const deeptrans_studio = require('deeptrans-studio');
const result = await deeptrans_studio.run(options);
console.log(result);
# deeptrans-studio 配置说明 # 查看配置选项 deeptrans-studio --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export DEEPTRANS_STUDIO_CONFIG="/path/to/config.yml"
---
DeepTrans Studio is an enterprise-grade translation platform that combines AI-powered translation, localization engineering, and team collaboration capabilities. Built for professional translators and localization teams, it provides comprehensive end-to-end translation workflow management.
corepack when you want the repository lockfile)npm run worker
```bash
corepack enable corepack prepare yarn@1.22.22 --activate yarn install
npm run local:check
```bash
docker compose -p deeptrans-studio --env-file "$DEPLOY_ENV_FILE" -f docker-compose-prod.yml \ build db migrate app app_worker
npm run local:setup
npm install ```
npm run local:secret
npm run dev:all ```
For a first install, run local:setup before local:check: setup starts the owned dependencies itself, applies migrations, creates the demo account, and creates the owned local storage bucket. local:check is read-only and verifies that PostgreSQL, Valkey, and both MinIO ports belong to the deeptrans-local Compose project and map to the exact loopback ports, then checks local credentials, storage-bucket readiness, and migration status. Use local:up only to start already-initialized dependencies or while diagnosing startup; it never replaces setup. The guarded app and worker receive only the validated local profile, so shell variables or a production-style .env cannot re-enable remote services. SMTP, COS, and remote AI are disabled by default. The translation-memory vector contract is fixed at 2048 dimensions. Translation-memory imports, vector backfills, and other queued workflows require npm run worker; when it is unavailable, queued work remains durable and resumes after the Worker becomes ready. To use remote AI deliberately, set LOCAL_ALLOW_REMOTE_AI=yes and provide development-only credentials in .env.local.
The fixed test@example.com / 123456 demo credential is active only when IS_DEMO=yes; a normal runtime follows the regular mailbox-verification path and never creates or accepts that demo account.
npm run dev also refuses to start when port 3000 is already occupied on either loopback family. Do not rely on Next.js choosing 3001: the local profile deliberately keeps authentication and internal API URLs fixed at port 3000.
After the Web process starts, GET http://localhost:3000/api/health returns { "status": "ok", "scope": "web" } only when that process can serve a route. It deliberately does not claim that PostgreSQL, storage, Valkey, or the Worker is ready; use npm run local:check and the relevant workflow status for those checks.
Additional UIs:
npm run local:check. Direct Prisma CLI commands are intentionally not part of the safe local-start path.cp .env.example .env.production
#
```
memory-import-upgrade-audit.ts refuses all database/Redis connections unless --live is supplied in this stopped deployment window. With --live, it is read-only by default and exits non-zero while legacy jobs still need gates. It cannot reconstruct jobs that Redis has already pruned, so --apply additionally requires a credential-free snapshot reference and its SHA-256; both are printed in the audit record and must be retained with the deployment. The generated gates do not declare old work successful; the current memory owner must review and release each one. An old job remains write-blocked if it is processed by the upgraded worker even if an administrator manually retries it. The barrier cannot retroactively control an old image or external queue consumer, so verify every old worker is stopped before the migration window.
Production service set:
db: PostgreSQL 18 with pgvector and PGroongavalkey: Redis-protocol cache and BullMQ runtimemigrate: one-shot Prisma migration gate, required before app startupapp: DeepTrans Studio web applicationapp_worker: background workertraefik: HTTPS reverse proxynpm run dev
| Component | Technology | Purpose |
|---|---|---|
| **Studio** | Next.js 15, React 19, TypeScript | Frontend UI, Server Actions, Authentication |
| **Worker** | Node.js, BullMQ | Background job processing, batch operations |
| **Database** | PostgreSQL 18, pgvector, PGroonga, Prisma 6 | Relational data, vector search, CJK keyword search, and ORM |
| **Cache** | Valkey | Redis-protocol cache, session state, task queues |
| **Storage** | StorageService interface, MinIO, Tencent COS | Document and asset storage |
| **Parser** | DOCX XML parser, MinerU online PDF parser, text parser | Document parsing for DOCX, PDF, TXT, and Markdown |
| **Gateway** | Traefik | Reverse proxy, SSL/TLS termination |
1. Fork & Clone: Fork the repository and clone your fork 2. Create Branch: Create a feature branch from main 3. Code Changes: Make your changes following our coding standards 4. Quality Checks: Run linting and type checking
yarn lint
yarn type-check
5. Commit: Use Conventional Commits format feat: add translation memory import
fix: resolve authentication bug
docs: update installation guide
6. Pull Request: Submit PR with clear description
DeepTrans Studio 是一款企业级的 AI 驱动翻译平台,集成了 AI 翻译、本地化工程(Localization Engineering)以及团队协作功能。该平台专为专业翻译人员和本地化团队打造,旨在提供从文档解析到翻译管理的全流程端到端工作流解决方案,助力实现高效、专业的翻译任务管理。
DeepTrans Studio 提供了一系列强大的核心功能,包括基于 AI 的智能化翻译、完��的本地化工程工具链以及支持团队协作的管理系统。通过集成先进的 AI 模型,平台能够处理复杂的翻译任务,并确保翻译质量符合专业标准,满足企业级应用场景的需求。
在开始使用前,请确保您的开发环境已安装以下必要组件:Node.js 版本需 ≥ 18.18(建议通过 corepack 管理 Yarn 1.22.22);需安装 Git 进行版本控制;此外,必须安装 Docker 及 Docker Compose,用于运行数据库、Valkey 和 MinIO 等核心服务。
首先,通过 `corepack enable` 启用 corepack 并配置 Yarn 1.22.22 环境。接着,使用 `yarn install` 安装项目依赖。随后,利用 `docker compose up -d` 命令启动必要的后端服务(包括 db、valkey 和 minio),以确保数据库和存储环境准备就绪。
项目提供了快速启动指南,帮助开发者快速上手。您可以参考提供的示例(如 Tencent COS 配置示例)来配置存储服务,并按照文档指引完成从环境初始化到首次翻译任务运行的全过程,实现从零到一的快速部署与体验。
项目通过 `.env.local` 文件进行环境配置。开发者需要配置 AI 服务相关的参数,包括 OPENAI_API_KEY、OPENAI_BASE_URL 以及指定的 OPENAI_API_MODEL��系统支持为聊天或翻译任务配置专用的 LLM 配置,若未指定,则会默认回退至 OPENAI 配置。
本项目集成了多种 AI 能力接口,支持通过配置 LLM_API_KEY、LLM_BASE_URL 以及 EMBEDDING_API_KEY 来调用大语言模型及向量化模型,确保翻译与语义理解功能的稳定输出。
系统采用基于 BullMQ 的队列驱动异步处理机制,支持批量翻译、评估及质量检查等自动化操作。在文档处理方面,支持 DOCX、PDF、TXT 和 Markdown 格式,并集成 MinerU 在线 PDF 解析能力。通过完整的状态追踪功能,用户可以实时监控翻译任务的全生命周期。
高质量开源AI工作流平台,实现人机协作翻译
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
AI Skill Hub 点评:深度翻译工作室 的核心功能完整,质量优秀。对于自动化工程师和运维人员来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | deeptrans-studio |
| 原始描述 | 开源AI工作流:DeepTrans Studio is a multi-agent intelligent translation platform for professio。⭐214 · TypeScript |
| Topics | 机器翻译人机协作多智能体系统TypeScript |
| GitHub | https://github.com/hint-lab/deeptrans-studio |
| License | MIT |
| 语言 | TypeScript |
收录时间:2026-06-28 · 更新时间:2026-07-03 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端