LOCAL PREVIEW View on GitHub

9. Data and Integrations — What Systems MangaAssist Connects To

Integration Map

graph TD
    subgraph "MangaAssist Orchestrator"
        O[Orchestrator]
    end

    subgraph "Product Data"
        A[Product Catalog] -->|ASIN, title, format,<br>language, page count| O
        B[Pricing Service] -->|current price, list price,<br>Prime pricing| O
        C[Inventory Service] -->|in stock, back-ordered,<br>restock date| O
        D[Reviews & Ratings] -->|avg rating, review count,<br>top review snippets| O
    end

    subgraph "Personalization"
        E[Recommendation Engine] -->|similar ASINs,<br>personalized picks| O
        F[Customer Profile] -->|preferences, locale,<br>Prime status| O
        G[Browsing History] -->|recent views, searches| O
    end

    subgraph "Order & Fulfillment"
        H[Order Service] -->|order status, items,<br>order date| O
        I[Shipping & Delivery] -->|tracking, ETA,<br>carrier info| O
        J[Returns Service] -->|eligibility, return window,<br>refund status| O
    end

    subgraph "Commerce"
        K[Promotions Service] -->|active deals, coupons,<br>bundle offers| O
        L[Cart Service] -->|cart contents,<br>add-to-cart API| O
    end

    subgraph "Knowledge"
        M[Support Knowledge Base] -->|FAQ articles,<br>policy docs| O
    end

Integration Details

1. Product Catalog

Aspect Detail
What data ASIN, title, author/artist, publisher, format (paperback/hardcover/Kindle), language, page count, dimensions, ISBN, age rating, series name, volume number, cover image URL, product description
Why needed Every product question requires catalog data. "Is this in English?" "How many volumes are there?" "Who is the author?"
Access pattern Lookup by ASIN (single or batch), search by keyword/filters
Protocol Internal gRPC or REST API
Latency < 100ms for batch of 10 ASINs
Update frequency Near real-time (catalog changes propagate within minutes)

2. Pricing Service

Aspect Detail
What data Current selling price, list price, discount percentage, Prime-exclusive price, used/renewed prices, price history (last 30 days)
Why needed "How much is this?" "Is this on sale?" "Is the box set a better deal?"
Access pattern By ASIN, batch supported
Critical rule Chatbot must NEVER cache prices. Always fetch real-time. Showing a wrong price is a legal and trust issue.
Latency < 50ms

3. Inventory Service

Aspect Detail
What data Stock status (in stock, limited, out of stock, pre-order), estimated restock date, warehouse location (for delivery estimation)
Why needed "Is Volume 12 available?" "When will this be back in stock?"
Access pattern By ASIN + fulfillment region
Latency < 50ms

4. Reviews & Ratings

Aspect Detail
What data Average star rating, total review count, top 3 review snippets (pre-generated), review sentiment summary
Why needed "Is this manga any good?" "What do people say about the translation quality?"
Access pattern By ASIN
Privacy note Never expose reviewer names or full review text — only pre-approved snippets
Latency < 100ms

5. Recommendation Engine

Aspect Detail
What data Ranked list of recommended ASINs based on: collaborative filtering ("customers who bought X also bought Y"), content-based similarity (genre, author, themes), personalized signals (user history)
Why needed Core to the "recommend something like X" use case and proactive suggestions
Access pattern By seed ASIN(s) + optional user ID + optional filters (genre, format, price range)
Protocol Amazon Personalize API or internal recommendation service
Latency < 200ms

6. Customer Profile

Aspect Detail
What data Customer ID, display name (first name only), Prime status, default locale, preferred language, default shipping address (city/state only — never full address)
Why needed Personalization ("Welcome back, Kenji!"), Prime-aware answers, locale-aware shipping estimates
Access pattern By customer ID (authenticated sessions only)
Privacy rule Minimal access. Never fetch or display full address, payment info, or email.
Latency < 50ms

7. Browsing History (Session)

Aspect Detail
What data List of ASINs viewed in the current session, search queries issued, time spent per page
Why needed Context injection: if a user just viewed Chainsaw Man Vol 1, the chatbot can reference it without the user having to say it
Access pattern Passed from frontend in page_context (not fetched from a backend service)
Privacy rule Session-scoped only. Not stored beyond the chat session.

8. Order Service

Aspect Detail
What data Order ID, order date, items (ASIN, title, quantity), order status (placed, shipped, delivered, cancelled), payment status
Why needed "Where is my order?" "What did I order last month?"
Access pattern By customer ID, optionally filtered by date range or category (manga)
Auth requirement Authenticated users only. Guest users cannot access orders.
Latency < 200ms

9. Shipping & Delivery

Aspect Detail
What data Carrier name, tracking number, estimated delivery date, delivery status (in transit, out for delivery, delivered), shipping speed selected
Why needed "When will my package arrive?" "Can I get this by Friday?"
Access pattern By order ID
Latency < 200ms

10. Returns Service

Aspect Detail
What data Return eligibility (yes/no + reason), return window end date, return label generation URL, refund status (pending, processed, credited), replacement availability
Why needed "Can I return this?" "My book arrived damaged." "When will I get my refund?"
Access pattern By order ID + item ASIN
Latency < 200ms

11. Promotions Service

Aspect Detail
What data Active promotions for manga category (Buy 2 Get 1, percentage discounts, Lightning Deals), coupon availability per ASIN, bundle offers, Subscribe & Save eligibility
Why needed "Any deals on manga right now?" "Is there a coupon for this?"
Access pattern By category (manga) or by ASIN
Latency < 100ms

12. Support Knowledge Base

Aspect Detail
What data FAQ articles, return policy documents, shipping policy, payment help articles, manga-specific guides (edition comparison, reading order guides)
Why needed RAG pipeline retrieves chunks from this KB to ground FAQ answers
Access pattern Vector search (embedding similarity) via OpenSearch
Update frequency Weekly refresh of embeddings. Content updates by operations team.

Integration Summary Table

System Read/Write Auth Required Real-Time? Cacheable?
Product Catalog Read No Yes 5-min TTL
Pricing Read No Yes Never cache
Inventory Read No Yes 1-min TTL
Reviews Read No No 1-hour TTL
Recommendations Read Optional Yes Session-level
Customer Profile Read Yes Yes Session-level
Browsing History Read No Yes Not stored
Order Service Read Yes Yes Not cached
Shipping Read Yes Yes 5-min TTL
Returns Read/Write Yes Yes Not cached
Promotions Read No Yes 15-min TTL
Knowledge Base Read No No Pre-indexed
Cart Read/Write Yes Yes Not cached