This repository is one part of a five-project AI engineering portfolio. The matrix is meant to make each project's role clear without repeating the full table in every README front page.
| Repo | Role | Core scenario | Engineering proof |
|---|---|---|---|
| knowledgeops-agent | Enterprise Spring AI RAG platform | Governed enterprise knowledge Q&A | Spring AI, RAG, JWT/RBAC, async ingestion, observability, regression evaluation |
| tianji-ai-agent | Business Agent engineering case | Course consulting, recommendation, and pre-order flow | Java, Spring AI, multi-agent routing, Tool Calling, MCP, SSE, multimodal entry points |
| nebula-kb | Local AI Knowledge Platform | Knowledge lifecycle + RAG engine (DeepDoc) + AI chat (Open WebUI) | Django, PostgreSQL, Redis, RAGFlow, Open WebUI, lifecycle workflow |
| forgepilot-studio | AI engineering execution workspace | Auditable AI coding task execution for teams | Python, FastAPI, React, runtime sandbox, MCP governance, audit replay |
| however-microservices-lab | Cloud-native microservices and AI lab | Multi-language microservices with AI assistant integration | Go, Python, Java, Node.js, C#, Kubernetes, gRPC, Ollama/Gemini |
knowledgeops-agent is the enterprise backend slice. It proves that RAG can be treated as a governed platform with tenant boundaries, asynchronous ingestion, auditability, observability, and repeatable quality checks.
The matrix link between KnowledgeOps and tianji is verified by:
- Code path:
KnowledgeOpsClientin tianji calls KnowledgeOps Agent's REST API (/ai/rag/search,/ai/memory/query,/ai/graph/search) - Fallback strategy: When platform is unavailable, tianji agents fall back to local VectorStore + Advisor
- CI evidence: Both repositories have green CI on
mainbranch - Docker compose: See
docs/evidence/README.mdfor a 2-service + 3-env-var cross-repo Docker Compose - Interactive architecture page: however-yir.github.io/matrix — click any node to jump to its Evidence Pack