w8123/EnterpriseAgentFramework

ReachAI企业级智能体开发平台：快速、安全完成已有业务系统智能化改造，让 AI 在 OA、ERP、CRM 等原系统中查数据、填表单、办业务。ReachAI: Quickly and securely bring AI to existing enterprise systems, enabling AI to query data, fill out forms, and execute business tasks directly within OA, ERP, CRM, and other business applications.

ReachAI – Enterprise‑Agent Framework (w8123/EnterpriseAgentFramework)

What it is – A Java‑centric platform that lets companies retrofit existing enterprise applications (OA, ERP, CRM, HR, procurement, etc.) with AI‑driven assistants. The system automatically scans a legacy web app, registers its pages, APIs and business‑logic as Capabilities, then generates and deploys AI Agents and Workflows that can answer natural‑language queries, fill forms and trigger transactions on‑the‑fly – all while preserving the original UI, authentication and audit trails.

Core pieces

Component Role Key tech
Capability SDK / Spring‑Boot starter Developers annotate Java service methods with @ReachCapability to expose them to the AI runtime. Java 8‑compatible, Spring Boot 2/3
Page Embed & Bridge Injects a chat widget into the existing web UI and registers explicit page actions (read, filter, write) that the agent may invoke. Vue 3, TypeScript, eafPageBridge.ts
AI Coding pipeline Scans the target system, suggests up‑to‑three concrete refactor tasks, and uses LLM‑based code generators (Codex, Claude‑Code, etc.) to implement them in the real codebase. LLM‑driven code generation, task protocol with acceptance criteria
Workflow Studio Visual editor for GraphSpec workflows that combine AI‑generated steps with verified business APIs. LangGraph4j, GraphSpec, Vue‑Flow
Agent Runtime Executes agents, selects an authorized workflow, enforces ACL/guard policies, records a full trace for audit. AgentScope, Runtime Service, MySQL + Redis + Milvus
Model & Knowledge services Centralised gateway for LLM chat, embeddings, RAG and vector search. Spring AI, Milvus, multiple LLM providers
Ops consoles Dashboards for operations, automation, evaluation (EvalOps) and run‑time governance (RunOps). Vue UI, trace replay, metrics aggregation

How a typical upgrade works

  1. Connect a Java‑based business system to ReachAI.
  2. Scan the UI and backend APIs → a capability map is produced.
  3. Recommend up to three low‑risk refactor items (read‑only analysis).
  4. AI Coding turns the selected item into a concrete engineering task; an LLM writes the code, runs tests and pushes it back.
  5. Publish the new Capability and associated Workflow; the agent can now invoke it.
  6. Run – users type natural‑language commands in the embedded chat; the agent selects a vetted workflow, calls the capability, and the UI updates automatically.
  7. Govern – every call is logged, requires user confirmation for writes, and can be replayed from the RunOps console.

Who uses which part?

Role Interaction
Developer Reviews scan results, approves AI‑suggested refactor, monitors AI Coding task, may add @ReachCapability annotations.
Business user Stays in the familiar OA/CRM UI, types requests like “Create a purchase request for 10 monitors”.
Manager / Ops Views dashboards showing agent usage, token consumption, risk tickets, and overall ROI.

What makes ReachAI different from similar tools?

  • Enterprise focus – built for multi‑tenant Java SaaS, respects existing auth, roles and transaction boundaries.
  • Evidence‑based refactor – every suggested change is backed by code references and acceptance criteria before any code is written.
  • Governance‑first – explicit ACLs, user confirmations for writes, immutable traces, and a separate evaluation loop (EvalOps).
  • Hybrid workflow – combines LLM‑generated reasoning with pre‑approved, versioned business workflows for reliability.

Getting started (local dev)

  1. Install JDK 17, Maven, Node ≥ 20 and Docker‑Compose.
  2. docker compose … up -d to start MySQL, Redis, Milvus.
  3. Run mvn clean install -DskipTests then start the five services (model, knowledge, control, runtime, capability) – a provided IntelliJ run configuration can launch them together.
  4. Launch the Vue admin UI (npm ci && npm run dev). Default dev credentials are admin / admin123.
  5. Follow the docs to register a sample Java app with the SDK and embed the chat widget.

License

Open‑source under the MIT License.

Bottom line – ReachAI is a full‑stack, production‑oriented framework for turning legacy Java enterprise applications into AI‑augmented assistants, with a strong emphasis on safety, auditability and incremental, AI‑assisted code modernization.

Related

  • Project
    nageoffer/ragent

    Ragent AI is a production‑ready Java platform for building Agentic RAG applications. It provides hybrid retrieval (vector, keyword, graph, web), advanced query understanding, model tiering with fault‑tolerance, multi‑level session memory, Redis‑based traffic protection, and a full ingestion‑to‑answer loop. The code is split into seven Maven modules (infra‑ai, system, rag, agent, bootstrap, framework, mcp‑server) that isolate AI vendor, storage, and retrieval concerns, making components interchangeable. v2 uses a ReAct‑style agent loop with MCP tool calls, Skills, and human‑in‑the‑loop confirmations. Production features include circuit‑breaker, transactional RocketMQ, idempotency, tracing (Langfuse optional), SSE streaming, and Sa‑Token security. A React UI offers both chat with citation/agent view and an admin console for knowledge‑base and intent‑tree management. Targeted at Java backend engineers and interview‑preparation candidates who need a real‑world, extensible AI stack rather than a toy demo.

  • Project
    clawdotnet/openclaw.net

    OpenClaw.NET is a .NET‑native, self‑hosted AI‑agent runtime with a desktop Avalonia UI (AgentQi Companion). It runs a gateway that talks to any OpenAI‑compatible model, offers streaming, tool execution, session memory, and many integrations (REST, WebSocket, MCP, 80+ tools, channel adapters). The gateway can be compiled with NativeAOT, making it suitable for edge or on‑prem deployments. Users can start via a ready‑made desktop bundle or from source, configure the assistant through chat commands, and extend functionality with .NET plugins or reusable MetaSkill DAG workflows.

  • Project
    openai/openai-java

    Official Java client library for OpenAI’s REST API, offering sync/async, streaming, workload‑identity auth, and optional Bedrock support.

  • Project
    LnYo-Cly/ai4j

    A Java AI Agentic SDK for JDK 8+ that provides unified access to multiple LLM platforms and built-in capabilities for agent orchestration, RAG, and durable long-term task management.

  • Project
    itwanger/PaiAgent

    PaiAgent is an open‑source, enterprise‑grade platform that lets users visually compose AI workflows with drag‑and‑drop nodes. It supports multiple large‑model providers via Spring AI, offers two execution engines (a classic DAG engine and a LangGraph4j state‑graph engine), includes a declarative “Skills” system for LLM‑driven function calls, and provides real‑time SSE debugging. The stack is React + TypeScript on the front‑end and Spring Boot 3 + Java 21 on the back‑end, with MySQL (and optional MinIO) for persistence.