GenAI app architecture cheat sheet (TypeScript + Node)

1. 1The request flowWhat happens between a user pressing Send and the answer appearing, and which part of the system owns each step.5 concepts

   1. 1Request flow end to endFrontend sends the message to your backend API.
   2. 2AI orchestratorThe code that turns one user message into model calls: it loads memory, rewrites the query, retrieves, builds the prompt, runs the tool loop and applies guards.
   3. 3Model client wrapperOne small module that every model call goes through.
   4. 4Data stores in a GenAI appUsually four: the main database (users, chats, messages), a vector index (often inside Postgres), object storage for uploads, and a cache such as Redis.
   5. 5Request path vs background jobsChat replies run in the request.

   5 in this chapterOpen this chapter
2. 2Frontend and backend layersWhat the browser owns (chat, sign-in, uploads, streaming, history) and what the server owns before any model is called: identity, permissions, validation, business rules, limits and logs.10 concepts

   1. 1Chat UIMessage list, input box, stop button, retry, copy, citations shown as links and a thumbs up or down on each answer.
   2. 2Streaming to the browserThe backend sends tokens as they arrive, usually over server-sent events, and the UI appends them to the last message.
   3. 3File uploadsThe browser uploads to object storage (often with a short-lived signed URL), then tells the backend the file id so a worker can parse and index it.
   4. 4Auth and chat history in the UIThe frontend holds a session or token and loads past conversations from the backend, never from local storage alone.
   5. 5No model keys in the frontendThe browser never calls the model provider directly with your API key; every call goes through your backend.
   6. 6Authentication and authorisationAuthentication proves who the user is; authorisation decides which chats, documents and tools they may use.
   7. 7Request validationCheck message length, file ids, conversation ownership and allowed options with a schema before doing any work.
   8. 8Business logic outside the promptPlans, quotas, pricing, which documents a team owns and what an action may change are decided in normal code and data, then passed to the model as facts.
   9. 9Rate limits and quotasLimits per user and per IP on requests per minute, plus a daily token or cost budget per user or team.
   10. 10Structured loggingEvery request logs a request id, user id, route, status, latency, model, tokens and cost as structured fields, not free text.

   10 in this chapterOpen this chapter
3. The RAG pipelineWhere retrieval sits in the system. The RAG cheat sheet covers each technique in depth; this chapter is about the moving parts and who runs them.5 conceptsOpens with credits
4. Tools and the agent loopHow the orchestrator lets a model act: plan, choose a tool, run it, check the result and decide again, inside limits your code sets.6 conceptsOpens with credits
5. Memory, caching, rewriting and routingThe parts that make a chat feel continuous, cheap and fast: what the app remembers, what it reuses and which model handles which request.6 conceptsOpens with credits
6. Three guard layersChecks around the model in three places: on what the user sends, on the content the model reads, and on what the model returns.5 conceptsOpens with credits
7. The evaluation folderEvals live in the repo next to the prompts they test. The AI evals cheat sheet covers metrics and tools; this is how they fit the project.4 conceptsOpens with credits
8. ObservabilitySeeing what happened inside one request, what users thought of it and what it cost.4 conceptsOpens with credits
9. A reference project layoutOne way to lay out the repo so each layer has a home. Rename freely; the point is that prompts, evals and agent instructions are files in Git.6 conceptsOpens with credits
10. Production checklistWhat must be true before real users arrive: packaging, delivery, secrets, API security, monitoring and cost limits.7 conceptsOpens with credits
11. 11Glossary, A to ZEvery concept in the sheet and the smaller terms it uses, in one list.64 terms

How a production GenAI app fits together: request flow, layers, RAG, agent loop, memory, caching, guards, evals, tracing, folder layout and launch checklist.