feat(augment): add augment AI assistant workspace#2555
feat(augment): add augment AI assistant workspace#2555rrbanda wants to merge 1 commit intoredhat-developer:mainfrom
Conversation
|
Important This PR includes changes that affect public-facing API. Please ensure you are adding/updating documentation for new features or behavior. Changed Packages
|
|
Review Summary by Qodofeat(augment): Add augment AI assistant workspace with multi-agent orchestration, streaming chat, RAG, and comprehensive test coverage
WalkthroughsDescription• Introduces the **augment** plugin workspace with three packages: frontend chat interface, backend plugin with Llama Stack integration, and shared types/APIs • **Multi-agent architecture** with router dispatching to specialist agents (Cluster Engineer via MCP, Product Knowledge Specialist via RHOKP) • **Streaming chat** with SSE-based real-time responses, tool call visualization, and human-in-the-loop approval workflow • **MCP tool integration** connecting to external MCP servers with optional approval gates and SSRF protection • **RAG capabilities** with vector store-backed document retrieval, hybrid search, and multiple document sources (GitHub, directory, URL) • **Security modes** (none, plugin-only via RBAC, full OAuth chain) with Keycloak integration and three-layer permission enforcement • **Safety guardrails** via Llama Guard/Prompt Guard shield integration through Llama Stack safety API • **Comprehensive test coverage** including admin routes (1698 lines), multi-agent orchestration E2E (1451 lines), router endpoints (977 lines), ResponsesApiProvider (993 lines), stream event normalization (560 lines), and session/config management tests • **Configuration system** with YAML baseline, database overrides, runtime resolution, and validation with sensible defaults • **Frontend API client** with endpoints for chat, conversations, admin config, model management, RAG testing, and vector store operations • **Dynamic plugin support** for RHDH with scalprum config and OCI-based plugin loading Diagramflowchart LR
User["User Chat Interface"]
Frontend["Frontend Plugin<br/>Chat UI & Admin Panel"]
API["API Client<br/>AugmentApi"]
Backend["Backend Plugin<br/>ResponsesApiProvider"]
LlamaStack["Llama Stack<br/>Multi-Agent Orchestrator"]
MCP["MCP Servers<br/>Tool Execution"]
RAG["RAG System<br/>Vector Store & Documents"]
Config["Config System<br/>YAML + DB Overrides"]
User -->|Chat Messages| Frontend
Frontend -->|API Calls| API
API -->|HTTP Requests| Backend
Backend -->|Orchestration| LlamaStack
LlamaStack -->|Tool Calls| MCP
LlamaStack -->|Document Search| RAG
Config -->|Runtime Config| Backend
Backend -->|Streaming Responses| Frontend
Frontend -->|Display Results| User
File Changes1. workspaces/augment/plugins/augment-backend/src/routes/adminRoutes.test.ts
|
Code Review by Qodo
1. SSRF via redirects
|
![rhdh-qodo-merge[bot]](https://avatars.githubusercontent.com/in/1961108?s=60&v=4){kind=small}
workspaces/augment/plugins/augment-backend/src/services/DocumentIngestionService.ts
Show resolved
Hide resolved
workspaces/augment/plugins/augment-backend/src/services/utils/mcpClient.ts
Show resolved
Hide resolved
rrbanda
force-pushed
the
feat/augment-workspace-v2
branch
from
c0f979f to
5a65089
Compare
|
SonarCloud Hotspot AnalysisThe 119 security hotspots break down as:
The 87 Qodo bugs addressed in latest push:
|
rrbanda
force-pushed
the
feat/augment-workspace-v2
branch
9 times, most recently
from
b9c2c4d to
74150fe
Compare
Introduces the augment plugin family (frontend, backend, common) under workspaces/augment. Provides a chat interface backed by Llama Stack with multi-agent orchestration, RAG, MCP tool calling, and configurable safety guardrails. Key capabilities: - Multi-agent architecture with router, cluster-engineer, and product-knowledge specialist agents - Streaming chat via SSE with conversation history - MCP server integration for live OpenShift cluster operations and Red Hat knowledge platform queries - Human-in-the-loop tool approval - RHDH dynamic plugin support (frontend scalprum + backend) - Configurable security modes (none, plugin-only, full) - Response evaluation via Llama Stack Scoring API Signed-off-by: Raghu Banda <rbanda@redhat.com>
rrbanda
force-pushed
the
feat/augment-workspace-v2
branch
from
74150fe to
6814749
Compare
|
|
please take a look @hopehadfield @alizard0 @durandom |
2 participants
Summary
Introduces the augment plugin workspace (
workspaces/augment/) with three packages:@red-hat-developer-hub/backstage-plugin-augment— Frontend chat interface with streaming responses, conversation history, multi-agent handoff visualization, and admin settings panel@red-hat-developer-hub/backstage-plugin-augment-backend— Backend plugin providing the Responses API integration with Llama Stack, MCP server proxy, RAG document ingestion, and multi-agent orchestration@red-hat-developer-hub/backstage-plugin-augment-common— Shared types, API refs, and permission definitionsKey capabilities
Test plan
yarn buildsucceeds for all three packages (common, frontend, backend)yarn testpasses for backend unit testsdist-dynamic/for both frontend (withdist-scalprum/plugin-manifest.json) and backend