MXMinglang Xie
Active projects
AI SaaSRAGFull-stack

Agarwood AI

A bilingual RAG knowledge assistant for agarwood education, fragrance guidance, cultural knowledge, and buyer consultation.

External link Progress board

Knowledge + retrieval iteration

Public progress

55%

Updated
2026-07-02

Next.jsSupabasePostgresOpenAIRAGBilingual content

Current focus

  • • Structuring sourced agarwood knowledge for retrieval-grounded explanation.
  • • Balancing educational guidance with practical buyer consultation flows.
  • • Keeping the deployed assistant stable under the /agarwood path.

Next steps

  • • Improve ingestion and source-review workflows.
  • • Add lightweight answer-quality checks for core knowledge categories.
  • • Refine fallback states when external AI services are unavailable.

A specialist knowledge assistant

Agarwood AI explores how retrieval-augmented generation can make a specialised cultural and commercial domain easier to understand without flattening its nuance.

The product direction includes an AI fragrance mentor, an agarwood encyclopedia, guided product explanations, and a recommendation agent grounded in curated source material.

Engineering focus

  • Clear separation between sourced knowledge and generated explanation.
  • Bilingual content for English and Chinese audiences.
  • Practical ingestion and retrieval workflows that can evolve with the collection.
  • A graceful fallback experience while external AI services are unavailable.

Current progress

The assistant is deployed under the /agarwood path on this domain. Current work focuses on making the knowledge base more useful, improving retrieval behavior, and shaping the assistant around the questions a real learner or buyer would ask.

The public portfolio root and this assistant intentionally share the same domain without sharing application state: the portfolio is served statically, while Agarwood AI remains isolated behind its existing /agarwood routing.