Reviewer Guide¶
Campus RAG Assistant is a production-style enterprise RAG platform for governed campus knowledge. It is intended to be reviewed through source code, architecture, screenshots, evaluation results, and operational artifacts — not as a hosted public product.
90-second read¶
- Full-stack RAG platform built around governed institutional knowledge.
- Multicloud provider boundaries across AWS Bedrock KB, Azure AI Search / OpenAI, and mock CI/local providers.
- Vue 3 product UI with sessions, streaming chat, cited sources, feedback, OAuth handoff, and opt-in web research.
- LangGraph orchestration, RAGAS evaluation baseline, LangSmith traces, Prometheus metrics, k6 load tests, CI/CD, and security docs.
- Mock providers allow local and CI execution without cloud credentials.
- Not presented as a hosted public product.
Senior signals¶
| Signal | Evidence |
|---|---|
| AI platform architecture | Provider registry, AWS / Azure / mock separation, tenant config — ADR-001 |
| RAG engineering | LangGraph stages, multi-query retrieval, rerank hooks, explicit source contracts — LangGraph roadmap |
| Evaluation discipline | RAGAS golden set, documented baseline, release-oriented gates — Evaluation |
| Observability | LangSmith traces, request IDs, Prometheus metrics — Operations |
| Product judgment | KB-first answers, opt-in web research, source transparency, feedback — Web Research roadmap |
| Production thinking | CI/CD, gitleaks, dependency review, rate limits, load testing, hardening backlog — CI/CD, Security, Load Testing, Production Hardening |
What this repository implements¶
This project builds from the public ets-berkeley-edu/chabot codebase and extends it into a source-reviewable AI platform. Implemented surface:
- Vue 3 product UI (sessions, streaming, sources, feedback, OAuth handoff)
- Provider registry for AWS, Azure, and mock execution modes
- LangGraph RAG pipeline (
condense→multi_query→retrieve→rerank→generate→format) - Tenant-hydrated prompt and config model in Postgres
- RAGAS evaluation harness with documented baseline
- LangSmith trace capture for KB and web research paths
- Opt-in web research path with disclaimer UI and WEB-labeled sources
- CI/CD, gitleaks, dependency review, k6 load testing, release docs, and a production hardening backlog
Suggested review paths¶
| Reviewer | Start here |
|---|---|
| Hiring manager | Case Study |
| Staff / principal engineer | Architecture, Design Notes, ADRs |
| AI engineer | Evaluation, LangGraph roadmap, Provider registry ADR |
| Platform / DevOps reviewer | CI/CD, Operations, Security, Load Testing |
| Product reviewer | Screenshots, Case Study, Product Roadmap |
What this is not¶
- Not an official UC Berkeley or UC product.
- Not a public hosted SaaS product.
- Not a production deployment claim.
- Not a generic chatbot demo.
- Not a claim that the current evaluation set is production-sufficient.
See Notice and License for attribution and licensing details.