zhaohai666 opened a new pull request, #426:
URL: https://github.com/apache/rocketmq-dashboard/pull/426
# PR Description
## PR Title
[ISSUE #10600][RIP-1 + RIP-3] Control Plane 5.0 Unified Implementation: Core
Architecture Refactor + LLM Native Integration (CLI / MCP / Embedded Chat)
---
## Summary
This PR implements the core delivery of RocketMQ 5.0 Studio proposed in
community recruitment issue #10600, delivering integrated capabilities of
**RIP-1 (Unified Control Plane Architecture)** and **RIP-3 (LLM Native
Integration)** within `apache/rocketmq-dashboard`.
Per the three-track construction plan of RocketMQ Studio, RIP-1 corresponds
to Track 1 AI-Native unified control platform and Track 2 Proxy Admin
capability upgrade, adopting a standardized three-layer abstract architecture
to realize unified O&M for 4.0 standalone clusters, 5.0 Local/Cluster Proxy and
cloud-managed instances.
All AI-Native capabilities of RIP-3 belong to Track 3 LLM native
integration, fully built on RIP-1 underlying abstractions, providing three
programmable interaction modes: `rmqctl` CLI, standard MCP Server and built-in
LLM chat panel. All portals share unified tool schema, metadata, authentication
and observability capabilities, without any independent redundant backend
interfaces.
This project has been selected into the Taiyi Open Source Challenge, all
code submissions under this branch will be recorded as official Apache RocketMQ
contributions.
---
## 1. Background & Motivation
RocketMQ 5.0 has completed the iteration of cloud-native elastic
architecture, gRPC multi-language access and LiteTopic, and the community
launched RocketMQ Studio co-construction plan (#10600) to build a new AI-Native
control plane, split into three orthogonal development tracks.
The original `rocketmq-dashboard v2.1.0` cannot match the new architecture
requirements, with core pain points as follows:
1. **Coupled architecture**: Mixed logic of 4.0 and 5.0 versions lacks
standardized SPI layers, repeatedly triggering compatibility bugs (#390, #401,
#402, #403).
2. **Single operation portal**: Only Web UI is provided, lacking
standardized CLI and programmable interfaces for automation and third-party
system docking.
3. **No AI agent invocation standard**: No general protocol exposed for LLM
and IDE Copilot to call O&M capabilities, requiring enterprises to develop
extra adaptation layers.
4. **Decentralized authentication and monitoring**: ACL 1.0/2.0
multi-architecture adaptation incomplete; no unified Prometheus query proxy,
inconsistent audit logic.
This PR solves all above problems at once: RIP-1 builds the unified
underlying control plane covering Track1 & Track2 demands, RIP-3 supplements
Track3 AI native interaction capabilities, and all modules reuse the underlying
layer without duplicate development.
---
## 2. Core Architecture & Design Principles
### 2.1 Three-Layer Abstraction (RIP-1 Foundation, Support Track1 & Track2)
```
┌─────────────────────────────────────────────────────────────┐
│ Business Layer (RIP-1 Web UI + RIP-3 CLI / MCP / LLM Chat) │
│ Capability-driven dynamic rendering & multi-modal access │
└───────────────────────────┬─────────────────────────────────┘
│
┌───────────────────────────▼─────────────────────────────────┐
│ MetadataProvider Domain Layer (META-01) │
│ Namespace / Topic / Group / ACL / Client / LiteTopic │
│ Cloud vendor SPI extension point │
└───────────────────────────┬─────────────────────────────────┘
│
┌───────────────────────────▼─────────────────────────────────┐
│ AdminClient Protocol Layer (ARCH-01, Track2 Core) │
│ Remoting / gRPC / Cloud OpenAPI multi-protocol support │
│ Standard auth hook injection (AUTH-01) │
└───────────────────────────┬─────────────────────────────────┘
│
┌─────────────────┼─────────────────┐
▼ ▼ ▼
4.0 NameSrv+Broker 5.0 Proxy Cluster Cloud Mgmt API
```
### 2.2 Key Design Principles Aligned with Studio Track Goals
1. **Single source of truth**: Web, CLI, MCP all reuse identical
`MetadataProvider`, `AdminClient` and tool schema to avoid capability
inconsistency.
2. **Capability-driven UI rendering**: Pages and operation entries
dynamically display based on cluster version features, hide unsupported
functions automatically.
3. **Unified security control**: Three-level risk classification L1/L2/L3
takes effect on all portals uniformly; all write operations adopt two-stage
dry-run confirmation.
4. **Full backward compatibility**: 100% retain all v2.1.0 original
functions, zero migration cost for existing clusters.
5. **Graceful degradation**: 5.0 Proxy and LLM exclusive capabilities will
be hidden automatically on 4.0 clusters or environments without LLM
configuration.
---
## 3. Detailed Changes
### Part A: RIP-1 — Control Plane 5.0 Core Modules (Track1 & Track2)
6 core modules fully delivered to meet unified governance and Proxy Admin
standardization requirements:
| Module ID | Module Name | Key Deliverables |
|-----------|-------------|------------------|
| **BASE-01** | Baseline Parity | Complete migration of 11 categories of
v2.1.0 capabilities; fix 7 known compatibility defects (#380, #381, #390, #401,
#402, #403, #407) |
| **ARCH-01** | Multi-Architecture Support | Realize SPI abstraction of
`ClusterProvider`, `AdminClient`, `MetadataProvider`, compatible with
v4-namesrv / v5-proxy-local / v5-proxy-cluster / cloud deployment modes,
realize runtime dynamic switching, complete Proxy Admin gRPC protocol
standardization required by Track2 |
| **META-01** | Metadata Management | Namespace resource tree management;
strong-type governance for 5 Topic types; LiteTopic prefix aggregation
lifecycle management; reserve cloud vendor extension SPI |
| **AUTH-01** | Full-Chain ACL | Dual compatible ACL 1.0(Remoting) / ACL
2.0(gRPC), automatic hybrid mode identification, cluster-level encrypted
credentials, configuration hot-reload, full-link standardized audit log |
| **CLIENT-01** | Dual-Protocol Clients | Unified Remoting/gRPC client view,
realize subscription, heartbeat and consumption progress diagnosis capabilities
defined by Track2 Proxy Admin, compatible with current telemetry metrics and
subsequent RIP-2 official view |
| **METRICS-01** | Observability | Access Prometheus data source, build
authenticated PromQL proxy, built-in 10+ monitoring dashboards and 20+ alert
templates, support one-click Grafana export |
### Part B: RIP-3 — LLM Native Integration (Track3, 100% Reuse RIP-1
Abstractions)
> No independent new backend interfaces added. All LLM operation
capabilities directly reuse RIP-1 metadata, client and authentication chain
logic.
#### Phase 1: `rmqctl` CLI
- Cover 9 resource categories with 30 standardized O&M tools
- kubectl-style multi-cluster context management, persistent configuration
file `~/.rmqctl/config.yaml`
- Support table/json/yaml multi-format output, unified error structure
`{code, message, hint}`
- Three-level risk interception consistent with Web console:
- L1 read-only: execute directly without confirmation
- L2 controllable modification: default dry-run preview, manual
confirmation required for submission
- L3 high-risk operation: blocked by default, must carry `--yes --force`
to execute
- Local audit log storage and encrypted credential management
#### Phase 2: Standard MCP Server
- One-to-one mapping with CLI commands, consistent tool schema without
deviation
- Dual transmission protocol: stdio for local AI tools (Claude Desktop,
Cursor), SSE for web front-end
- Built-in `rmq.capabilities` detection tool to help LLM identify cluster
feature boundaries and avoid functional hallucinations
- Built-in SecurityGate interceptor, uniformly execute three-level risk
rules and dry-run logic
- Reuse RIP-1 AUTH-01 full-link authentication for all tool calls
#### Phase 3: Console Embedded LLM Chat
- Three quick entry ways: homepage command bar, global `Cmd/Ctrl+K` shortcut
pop-up, resource sidebar chat with automatic context injection
- Backend MCP forwarding proxy, model API keys will not be exposed to
front-end
- Natural language execution results rendered with native page components,
support one-click jump to resource details
- Consistent security interception rules with CLI/MCP
- Automatic downgrade to full-text search when no LLM provider configured
- Support OpenAI, Azure OpenAI, AWS Bedrock, DeepSeek, Tongyi Qwen, Ollama
six mainstream model vendors
---
## 4. Cross-RIP & Cross-Track Synergy Highlights
1. **Unified authentication system**: Web, CLI and MCP share the same
credential storage and permission judgment logic from AUTH-01.
2. **Unified capability model**: All portals identify cluster feature set
uniformly, automatically hide 5.0 and LLM exclusive functions for 4.0 clusters.
3. **Unified observability pipeline**: CLI/MCP monitoring query and Web
dashboard all reuse METRICS-01 PromQL proxy module.
4. **Unified client diagnosis capability**: Client subscription and
heartbeat query of CLI/MCP directly reuse CLIENT-01 aggregation logic defined
by Track2.
5. **Unified audit standard**: All write operations triggered by three
portals generate audit logs with operation source tags.
6. **Optional modular deployment**: CLI, MCP and LLM modules can be closed
through configuration to adapt to compliance-sensitive environments.
---
## 5. Repository Structure (rocketmq-dashboard rip-1-control-plane-5.0
branch)
```
rocketmq-dashboard/
├── src/main/java/.../architecture/ # RIP-1 ARCH-01 SPI abstraction
(Track2 core)
│ └── impl/ # V4/V5/Cloud
multi-architecture implementation
├── src/main/java/.../service/ # RIP-1 six core business
modules (Track1)
├── src/main/java/.../controller/ # Web REST interface
├── rocketmq-dashboard-app/ # Studio Web main startup
service
├── rocketmq-dashboard-cli/ # RIP-3 Phase1 rmqctl CLI
(Track3)
├── rocketmq-dashboard-mcp/ # RIP-3 Phase2 MCP Server
(Track3)
├── rocketmq-dashboard-llm/ # RIP-3 Phase3 LLM adapter &
MCP forwarding bridge
├── frontend-new/
│ ├── src/components/llm/ # Front-end LLM chat components
│ └── src/pages/LlmSettings/ # LLM vendor configuration page
```
---
## 6. Testing & Verification
| Module | Test Coverage | Status |
|--------|---------------|--------|
| RIP-1 Core (ARCH/META/AUTH/CLIENT/METRICS) | 40 test files, full coverage
of core interface logic | ✅ All passed |
| RIP-3 CLI (rmqctl) | 161 test cases, ~80% command coverage | ✅ All passed |
| RIP-3 MCP Server | 101 test cases, ~81% tool schema coverage | ✅ All
passed |
| RIP-3 Console LLM Backend | 98 test cases covering interfaces,
interceptors and model vendors | ✅ All passed |
| End-to-end integration | Full process verification on 4.0 / 5.0-Local /
5.0-Cluster | ✅ All passed |
---
## 7. Compatibility & Upgrade Notes
1. **Full backward compatibility**: All v2.1.0 REST APIs and page functions
remain intact, existing users upgrade without business transformation.
2. **Automatic multi-architecture adaptation**: Identify cluster deployment
mode automatically, dynamically display or hide version-exclusive functions.
3. **Configurable optional modules**: CLI, MCP and LLM chat can be disabled
via configuration for compliance requirements.
4. **Smooth docking with RIP-2**: CLIENT-01 module compatible with current
telemetry metrics, seamlessly switch to official Proxy Admin view after RIP-2
release.
---
## 8. Acceptance Checklist
### RIP-1 (Track1 Unified Control + Track2 Proxy Admin Upgrade)
- [x] ARCH-01: Three-layer SPI abstraction + multi-architecture
implementation + runtime switching + standardized Proxy Admin gRPC protocol
- [x] BASE-01: 100% v2.1.0 feature equivalence + repair of 7 historical
compatibility defects
- [x] META-01: Namespace, 5 types of Topic, LiteTopic management, cloud
vendor SPI extension
- [x] AUTH-01: Dual ACL adaptation, cluster encrypted credentials,
configuration hot-reload, full-link audit
- [x] CLIENT-01: Unified Remoting/gRPC client view, complete Proxy Admin
runtime diagnosis capability
- [x] METRICS-01: Unified Prometheus integration, built-in monitoring
dashboards and alert templates
### RIP-3 (Track3 LLM Native Integration)
- [x] CLI: 30 standardized O&M tools, multi-cluster context, structured
output, three-level safety dry-run mechanism
- [x] MCP Server: stdio/SSE dual transport, cluster capability detection,
unified security interceptor
- [x] Console LLM: multi-entry chat panel, back-end MCP proxy, structured UI
rendering, consistent safety rules
- [x] 100% reuse RIP-1 underlying logic, no redundant independent interfaces
- [x] Automatic downgrade processing for 4.0 clusters and empty LLM
configuration
---
### Related Work
- **RIP-2 (Proxy Admin API)**: Corresponding specification implementation in
the main repository `apache/rocketmq`. The CLIENT-01 module of this PR is
compatible with both the current transition diagnosis view and the final
standard view after RIP-2 release.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
