LLM Applications

Basic LLM Application

The simplest production LLM application: a UI calls an API which calls the model. Everything else is built on top of this shape.

clientservicemodeldatabaseagentexternal

Explanation

A client sends a request to your API. The API constructs a prompt, calls the LLM provider, and returns the response to the UI. This is the foundation — but it has no memory, no retrieval, no tools and no safety.

Components

Web UIAPI serverLLM provider SDKStreaming response layer

When to use

Prototypes and demos
Stateless utilities (summarise this, translate this)
Internal developer tools with trusted users

When NOT to use

Anything touching private data (use RAG)
Anything requiring action (use Agents)
Customer-facing products without guardrails

Failure modes

Prompt injection via user input
Token cost blowouts from unbounded input
No streaming — high time-to-first-token
No retries — single provider failure breaks the app

Production checklist

Input validation and length limits
Streaming responses
Retry with exponential backoff
Rate limiting per user
Prompt injection defence
Cost monitoring and budgets
Logging of prompts and responses (PII-aware)