Skip to content

Create j_cq_ai_idea.md#128

Open
jeongmossland wants to merge 1 commit intomainfrom
j_cq_ai_idea
Open

Create j_cq_ai_idea.md#128
jeongmossland wants to merge 1 commit intomainfrom
j_cq_ai_idea

Conversation

@jeongmossland
Copy link
Contributor

No description provided.

@mossland
Copy link
Contributor

📝 Review (EN)

1. Idea Clarity

  • I understand that the service will generate custom PC part lists based on a user’s stated use-case (e.g., gaming, video editing) and budget.
  • You plan to:
    • Let users describe their needs in free text.
    • Parse the text with an LLM.
    • Query up-to-date parts/price data you crawl.
    • Show several build options and (later) allow one-click ordering with affiliate/commission revenue.
  • Good starting point, but a few areas remain vague (see questions below).

2. Execution Considerations

  • Service form factor – Web, mobile app, or both? Any preference on cross-platform tech?
  • Core workflow – Natural-language → spec selection → compatibility check → price lookup → output builds. We will need explicit rules/ML logic for:
    • Compatibility (socket, RAM type, PSU wattage, case size, etc.).
    • Regional availability & currency.
  • Data layer – How often and from which sources will we crawl? Existing price-comparison APIs might shorten time-to-market.
  • AI layer – Will we rely on an external LLM (OpenAI, Anthropic, etc.) or host an open-source model? Fine-tuning vs prompt-engineering?
  • Scalability – Real-time price updates can be heavy if we self-crawl; consider on-demand API calls.
  • Team load – Looks feasible within one squad (backend + full-stack + AI) if scope is limited to MVP.

3. UI / UX Detail

We will eventually need:

  • Input page: use-case text box, optional sliders (budget, noise level, RGB preference).
  • Output: at least three builds (budget / balanced / premium) with part thumbnails, price per store, “replace part” dropdowns, and estimated FPS / benchmark.
  • “Order” CTA wired to affiliate links or internal cart.

4. Cost & Monetization

  • Main cost drivers: GPU-heavy server if we self-host inference, continuous crawling, and price DB storage.
  • Monetization: affiliate commission, potential B2B white-label for local PC shops, premium subscription for advanced analytic features (e.g., power-draw, noise).
  • Please confirm target market (KR only? Global?) to validate CPL and affiliate rates.

5. Competitive Landscape

  • PCPartPicker, Pangoly, Logical Increments, Build Redux, and multiple regional “견적 게시판” sites.
  • Differentiators we must highlight: natural-language input in Korean, automatically updated KR price DB, end-to-end ordering, maybe energy-efficiency optimization.

6. Key Open Questions

  1. Platform choice: Web first? Any mobile requirement for v1?
  2. Region & currency scope at launch.
  3. Data sources and refresh cycle (hourly? daily?).
  4. Compatibility engine plan (rule-based vs ML).
  5. LLM hosting preference & budget ceiling per inference.
  6. Monetization timeline (immediately via affiliate vs later).
  7. MVP feature cut: Do we postpone FPS estimation or require it day-1?
  8. Expected monthly traffic in year-1 (affects infra decisions).

Could you clarify the above so we can break down tasks and prepare sprint estimates?


📝 리뷰 (KR)

1. 아이디어 명확성

  • 사용자가 용도와 예산을 자유롭게 입력하면 AI가 부품 호환성과 가격을 고려한 견적을 추천한다는 점은 명확합니다.
  • 최신 가격 데이터를 수집하고, 향후 주문 연동 및 수수료 모델까지 계획하고 계신 점도 확인했습니다.
  • 다만 서비스 형태와 기술적 구현 방식이 구체적으로 드러나지 않아 추가 정보가 필요합니다.

2. 구현 고려사항

  • 서비스 형태 – 웹인지 모바일 앱인지, 둘 다인지 명확히 해주세요. 초기에는 웹으로 가볍게 MVP를 만들 수도 있습니다.
  • 핵심 로직 –
    • 자연어 입력 → 부품 스펙 도출 → 호환성 확인 → 가격 조회 → 다중 견적 제시.
    • 호환성(소켓, RAM, 전원, 케이스 규격 등) 체크 규칙을 어떻게 구성할지 결정해야 합니다.
  • 데이터 레이어 – 어떤 사이트/API에서 가격·재고를 가져올지, 크롤 주기는 어느 정도로 잡을지 결정이 필요합니다.
  • AI 레이어 – 외부 LLM API를 쓸지, 자체 호스팅할지에 따라 비용·성능이 달라집니다.
  • 스케일 – 실시간 가격 반영이 중요한데, 직접 크롤링이면 서버 부하가 클 수 있습니다.
  • 팀 이용 가능 인력으로는 백엔드 1~2명, 프론트 1명, AI 1명 정도로 MVP는 가능할 것으로 보입니다.

3. UI / UX 구체성

  • 입력: 용도 자유 입력, 예산 슬라이더, 추가 옵션(소음/LED 여부 등).
  • 출력: 최소 3가지 빌드(가성비/중간/고급) + 각 부품 이미지, 상점별 가격, 교체 옵션, 예상 성능 그래프.
  • 주문: 제휴 링크 혹은 장바구니 연동 버튼.

4. 비용·수익 모델

  • 비용: AI 호출비+서버, 가격 DB 구축 유지비, 크롤링 인프라.
  • 수익: 제휴 수수료, PC 샵 대상 B2B 솔루션, 프리미엄 구독(고급 견적·소음/전력 분석).
  • 목표 시장(국내만인지, 글로벌인지)에 따라 수익 구조가 크게 달라지므로 반드시 확정이 필요합니다.

5. 유사 서비스

  • 국내외로 PCPartPicker, Pangoly, 밑짚기 견적 사이트들이 이미 존재합니다.
  • 차별점: 한국어 자연어 입력, 실시간 국내 최저가, 원클릭 주문, 친환경/전력 최적화 등.

6. 핵심 오픈 질문

  1. 플랫폼: 1차 출시를 웹으로 할지, 모바일을 포함할지?
  2. 출시 지역·통화 범위.
  3. 데이터 소스 및 갱신 주기(시간/일 단위?).
  4. 호환성 엔진 방식(규칙 기반 vs ML).
  5. LLM 호스팅 방안과 호출 예산.
  6. 수익화 시작 시점(바로 제휴 vs 후속).
  7. MVP에서 FPS·벤치 표시를 포함할지.
  8. 예상 월간 방문자 수(인프라 규모 산정).

위 내용을 구체적으로 알려주시면 스프린트 일정과 세부 작업 항목을 산출할 수 있습니다!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants