How does a declarative LLM framework differ from an imperative one in what the developer actually writes?

Attempt it

Declarative: the developer writes a signature (inputs → outputs + semantic descriptions) and a metric; the framework compiles a prompt, and often few-shot exemplars, that scores well on the metric. The final prompt is generated, not authored. Imperative: the developer writes the prompt and the chain steps directly; the framework executes them. Declarative trades step-level visibility for automatic prompt optimization; imperative trades optimization for debuggability and explicit control.

Your answer

Quick reference

TL;DR

Declarative tools like DSPy and BAML compile prompts from a signature plus a metric; imperative tools like LangChain and LlamaIndex execute the prompts and steps the developer wrote.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Picture two cooks asked to make the perfect omelet. The first cook writes the recipe themselves, three eggs, this much butter, sixty seconds on each side, and follows it exactly. If the omelet is bad, they edit the recipe and try again. The second cook only writes down what counts as a great omelet and hands a robot a basket of eggs. The robot tries hundreds of recipes, keeps the ones the judge liked best, and serves the winner. Cook one knows every detail of how the omelet was made. Cook two never sees the recipe but the omelet usually scores higher. Different tradeoffs, both valid.

Key concepts

LLM Frameworks Prompt engineering LLM evaluation

Quick scan · 2–4 min

Everything important, quickly.

4 minutes: the two families, who writes the prompt, why DSPy compiles, why BAML codegens, when teams mix the two.

Side-by-side

Axis	Imperative (LangChain, LlamaIndex)	Declarative (DSPy, BAML)
What developer writes	Prompts plus chain code	Signature or schema plus metric
What model receives	What the developer wrote	What the compiler generated
Optimization	Manual prompt iteration	Automatic search against a metric or codegen from schema
Debuggability	Every step visible	Final prompt is a compiler artifact
Best fit	Moving target, no labels, deep custom control	Stable task, labeled data, want prompt to keep improving

Real-world examples

Real products, models, and research that use this idea.

DSPy 2.6 ships MIPROv2 as the default optimizer, used in production by Databricks and several Anthropic-customer pipelines.
BAML is used by Vercel and Notion-style teams that want schema-first LLM calls with TypeScript codegen.
LangChain 0.3 LCEL pipelines remain the most common imperative composition layer in 2026.
LlamaIndex 0.13 keeps imperative QueryEngines as the default, with Workflows adding event-driven control flow on top.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QWalk through how DSPy's MIPROv2 optimizer actually searches the prompt space.

Sign in or go Pro to see the approach for this follow-up.

QWhen would you choose BAML over DSPy for a structured-output use case?

Sign in or go Pro to see the approach for this follow-up.

QHow would you A/B test a DSPy-compiled prompt against a hand-written prompt in production?

Sign in or go Pro to see the approach for this follow-up.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.