Skip to content
Clawdemy

Summary: Project walkthrough

The bootcamp’s worked example, askFSDL, is a chat-style Q&A app over the FSDL course materials. Its overall shape is the lesson-1 five-component pipeline with the lesson-4 RAG pipeline wedged into the application code. The value of the walkthrough is in the production decisions at each step: scope (a narrow, well-retrievable corpus); chunking for the content’s natural unit, with metadata; retrieval that carries the source label through to the end; a scope-honest, citation-asking system prompt; streaming generation with citations; and logging enough to debug (question, retrieval IDs, prompt version, model + parameters, response, user feedback). What it deliberately defers: sophisticated UX (lesson 6), production observability/eval pipelines (lesson 7), agentic multi-tool flow (lesson 10). The reframing worth carrying: a real LLM app of this shape is a few hundred lines plus prompts plus the corpus plus a hosted model; the complexity lives in the decisions, not the line count. This is the scan version; the lesson reads the decisions.

Core ideas

Section titled “Core ideas”
  • askFSDL = the worked example. Q&A over the FSDL course corpus, the shape of a large class of production LLM apps. Quality is the sum of the decisions at each stage.
  • Scope the knowledge source narrowly. Narrow + retrievable beats broad + inconsistent; lets the system prompt promise scope and refuse out-of-scope honestly.
  • Chunk for the content’s natural unit, with metadata (source, section, URL) for citations and filtering.
  • Carry the source label through retrieval and into the prompt; citations are not bolted on at the end.
  • Scope-honest system prompt: answer in-scope from the provided context, cite sources, refuse out-of-scope plainly. End-place critical instructions; cap max_tokens.
  • Streaming + citations are call-site decisions rendered in the UI; not afterthoughts.
  • Logging seeds LLMOps: question, retrieved chunk IDs, prompt version, model identifier, parameters, response, user feedback signal. Trivial up front; near-impossible to backfill.
  • What the walkthrough defers (and where it goes): sophisticated UX -> lesson 6; production observability/eval pipelines -> lesson 7; agentic multi-tool flow -> lesson 10.
  • The “five hours, not five weeks” reframing: real apps of this shape are small in code; the complexity is in the decisions.

What changes for you

Section titled “What changes for you”

Reading a real application is the cheapest way to develop the production-decision eye that distinguishes builders who ship from builders who tinker. Most LLM applications you will build look like askFSDL in shape: a scoped knowledge source, deliberate chunking, source-carrying retrieval, a small disciplined system prompt, streaming generation with citations, and logs that seed an evaluation practice. Once you can see those decisions in someone else’s application, you can make them deliberately in your own, and the gap from “I built a prototype” to “I shipped a useful app” closes quickly. The next two lessons fill in what the walkthrough deliberately deferred: the UX layer (lesson 6) and the LLMOps layer (lesson 7).

A real LLM application is mostly the parts you have learned, wired together with care. The walkthrough is not asking you to copy askFSDL; it is asking you to develop the eye that sees the same decisions in any LLM app, including the one you build next.