AppPlatform: AI Application Engineering Platform
Build, orchestrate, and deploy AI applications with a low-code visual interface and declarative programming.
- Source repo
- ModelEngine-Group/app-platform
- Stars
- ★ 1.5k
- Last updated
- 20d ago
- License
- MIT
- Primary language
- Java
- FA score
- 23/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A software engineer uses the visual editor to rapidly prototype an AI application with multiple models and custom logic without writing extensive code.
- Not a fit if
- Teams wanting to avoid ModelEngine-specific frameworks (FIT, Waterflow, Elsa)
- Teams without PostgreSQL 14+ and Docker available
- Teams needing Linux source builds or complete documentation
- Source review
- 23/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
AppPlatform is an open-source, low-code platform for building AI applications, targeting software engineers and product managers. It covers the full lifecycle from concept to deployment. The backend is developed on the FIT framework with a plugin architecture, and process orchestration uses the Waterflow framework, enabling multi-model collaboration and operator extensibility. The frontend is built with React, offering visual orchestration, an app marketplace, intelligent forms, and plugin management, with Elsa graphics engine support. Deployment options include Docker Compose and source compilation. It supports Java and Python operators and includes a Store template for app reuse and sharing. However, it is tightly integrated with ModelEngine-specific frameworks and requires PostgreSQL and Docker for deployment.
AppPlatform manages the entire lifecycle of AI applications. Users create blank apps or select templates from the marketplace, then visually orchestrate models, operators, and plugins into workflows executed by the Waterflow framework. Model configuration requires adding API keys and endpoints via the 'model configuration app'. Developers can build custom operators and plugins in Java or Python. The frontend supports JSON Schema-based forms and plugin management. Published apps can be stored in the Store for reuse across projects.
- A software engineer uses the visual editor to rapidly prototype an AI application with multiple models and custom logic without writing extensive code.
- A product manager creates a customer support chatbot by selecting a template, configuring model parameters, and testing through the chat interface.
- An algorithm developer implements a custom Java or Python operator and integrates it into the orchestration to handle domain-specific processing.
- A team member publishes a completed application to the Store, making it reusable by other projects or community members.
- A frontend developer uses the intelligent forms module to build interactive forms that integrate with AI models for real-time inference.
- An IT operator deploys the platform on a server using Docker Compose and scales it according to organizational needs.
How do you install or deploy this agent?
Requirements: CPU 2+ cores, x86 architecture, 4+ GB RAM, Docker and Docker Compose installed. Clone the repo, then in the root: cp docker/.env.example docker/.env and bash docker/deploy.sh. After containers are running, open http://localhost:8001. For source compilation, install Java 17, Maven 3.8.8+, Node.js 20+, PostgreSQL ≥14, and follow FIT framework build steps. Configure backend by editing fitframework.yml with database and form paths; configure frontend proxy in proxy.config.json pointing to the backend.
How do you use this agent?
After startup, open the 'model configuration app' from the marketplace, add a model with name, API key, and endpoint. Return to the home page, click 'create blank app', fill in name and description, create it, then in the app configuration page select the configured model under 'large model'. You can then chat. For development, compile backend plugins with mvn clean install or build the frontend with npm run build:prod, and deploy using the provided scripts.
What are this agent's strengths and limitations?
- Visual low-code orchestration dramatically reduces development effort, enabling non-programmers to build AI apps.
- Extensible backend based on FIT and Waterflow, supporting Java and Python operators for high performance and scalability.
- Built-in Store templates and plugin marketplace encourage reuse and collaboration.
- Support for multiple AI models in a single workflow, allowing flexible integration based on business needs.
- Tight coupling to ModelEngine frameworks (FIT, Waterflow, Elsa) makes migration to other stacks challenging.
- Deployment requires PostgreSQL 14+ and Docker, and the database initialization script currently only supports Windows, not Linux.
- Documentation is marked as 'to be improved'; some sections like Linux source build are incomplete.
- Targets Java and JavaScript/React ecosystem; teams unfamiliar with these may face a learning curve.
How does this agent compare with similar options?
Competitors like Dify or Flowise offer similar low-code AI app platforms, but AppPlatform is deeply tied to the ModelEngine ecosystem and primarily fits within that context.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AppPlatform: AI Application Engineering Platform This agent | 23 · Major gaps | Self-hosted serviceFree + model costs | ★ 1.5k | 20d ago | Java | — |
| Spring AI Alibaba | 53 · Major gaps | FrameworkFree + model costs | ★ 11k | 11d ago | Java | — |
| LazyLLM | 37 · Major gaps | Library / SDKFree + model costs | ★ 3.9k | 6d ago | Python | OpenAI API |
| Okto Pulse | 80 · Good | CLIFree | ★ 51 | 2d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows SECURITY.md and MIT license, but no implementation of permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. The anti-spam-filter.yml workflow uses eval on secret code, posing security risk, and has overly broad permissions (contents: write). Therefore all trust criteria score 0.
The architecture described in README is mostly consistent with code structure, but inconsistencies exist: README mentions React frontend, but example code uses Ant Design components, and build scripts use npm link, potentially affecting reproducibility. Dependency availability: README specifies JDK 17 and Node 20, but no lock files are provided, potentially affecting build stability. Failure messages: no error handling or user-friendly error messages found.
README clearly identifies target users (software engineers and product managers) and use cases (AI application development), but capability boundaries are vague, not specifying supported models or feature limitations. Trigger precision: no clear trigger conditions or input/output specifications defined. Environment fit: Docker Compose and source compilation provided, but only x86 architecture supported, and depends on specific versions of external frameworks (FIT, Waterflow, Elsa).
Information architecture is clear, README provides quick start, development testing sections. Install notes are detailed, including Docker Compose and source compilation steps. Naming stability: project and module names are consistent, but no version history or changelog. Examples and FAQ: quick start example provided, but FAQ missing. Known limitations not mentioned. License is MIT, but copyright held by Huawei. Versioning/changelog missing. Maintenance responsibility: SECURITY.md provides security update support, but no clear maintainers or contribution guidelines.
Output usability: Docker Compose one-click deployment provided, but requires user to configure environment variables, and no detailed output instructions. Marginal value: low-code graphical interface and plugin system are innovative, but no comparison with similar tools. Cost-benefit: deployment requires 2+ cores and 4GB RAM, and depends on multiple external frameworks, potentially increasing deployment cost, but no performance or resource consumption data provided.
Claims in README (e.g., support for multi-model collaboration) lack specific implementation or test evidence. Cross-source corroboration: no other sources (e.g., docs, test reports) support these claims. Fact-inference separation: README mixes product description with architecture explanation, not clearly distinguishing facts from inferences.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- The anti-spam-filter.yml workflow uses eval on secret code, posing code injection risk, and has overly broad permissions (contents: write); restrict permissions and avoid eval.
- README claims support for multi-model collaboration, but no specific implementation or test evidence is provided; further verification needed.
- Depends on specific versions of external frameworks (FIT, Waterflow, Elsa) and lacks dependency lock files, potentially affecting build reproducibility.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
🎉 App Platform v1.3.2 发布说明 感谢您使用 App Platform！我们很高兴地发布 v1.3.2 版本。本次更新带来了多项功能增强与问题修复，旨在进一步提升平台的稳定性与用户体验。 ⚙️ 功能增强 - 优化重名应用创建报错提示 by @loveTsong in #580 - 删除http插件中url的校验，保证更加灵活的配置方式 by @loveTsong in #576 🐛 问题修复 - 删除工具流在工具中不合理的预览跳转 by @loveTsong in #579 - 修复分域功能场景考虑不全的问题 by @loveTsong in #577 - 修复接口文档中字段不匹配的部分 by @loveTsong in #578 - 删除失效的在线帮助跳转 by @loveTsong in #575 - 修复创建应用后修改提示词保存后读取有可能导致页面上数据覆盖的问题 by @RonnyChan96 in #575 #566 #567 - 修复访客模式的配置在发布后没有继承的问题 by @desAweis in #553 👏 贡献者 特别感谢以下开发者为本次发布做出的杰出贡献： @loveTsong, @RonnyChan96, @desAweis …View on GitHub ↗ -
Release notes excerpt
🎉 App Platform v1.3.1 发布说明 感谢您使用 App Platform！我们很高兴地发布 v1.3.1 版本。本次更新带来了多项功能增强与问题修复，旨在进一步提升平台的稳定性与用户体验。 ✨ 更新亮点 - 应用编排增强：优化了大模型节点提示词的配置体验。 - 问题修复：解决了问题分类节点调用工具报错的问题。 - 开发效率提升：新增前端快速开发脚本，让本地开发前端更顺畅。 ⚙️ 功能增强 - 应用编排： - 新增节点配置项更新功能 by @wuayee in #516 - 调整提示词内容的大小限制 by @wuayee in #529 - 移除知识库节点的可行性检查 by @wuayee in #531 - 支持配置输入长度 by @loveTsong in #534 - 应用市场： - 取消内置应用在市场中默认置顶的规则 by @ZJUPJM in #526 - 为应用市场新增短链属性与应用分类功能 by @wuayee in #528 🐛 问题修复 - 工具调用：修复了问题分类节点调用工具时出错的问题 by @loveTsong in #518 🚀 构建部署 …View on GitHub ↗ -
Release notes excerpt
🎉 App Platform v1.3.0 发布说明 感谢您使用 App Platform！我们很高兴推出 v1.3.0 版本，本次更新带来了多项功能优化与问题修复，为您提供更稳定、更高效的使用体验。 --- 🚀 新增与优化 - 智能体优化： - 调用大模型生成智能体超时后，优化了提示信息，提升用户体验。 @wuayee in #504 - 修复创建智能体时无工具列表的问题，提升创建流程完整性。 @wuayee in #488 - 增加工具调用过程展示，使调用逻辑更加直观可见。 @wuayee in #492 - 应用编排： - 新增变量更新节点，增强流程灵活性。 @desAweis in #452 - 新增文本拼接节点，用于拼接多段文本内容。 @reeeborn33 in #444 - 支持通过弹窗新增或编辑开始节点的自定义变量。 @reeeborn33 in #458 - 新增直接回复节点。 @reeeborn33 in #471 - 联网搜索：新增联网搜索插件并支持知识库溯源优化。 @surpercodehang in #474, #472 - 首页加载：优化首页体积，提升页面加载速度。 @loveTsong in #502 …View on GitHub ↗