autorag.generate

LLM response generation over retrieved context.

Defines the Generator interface. Concrete backends (Ollama, hosted APIs) override Generator.generate() to call the model of their choice with the prompt assembled by Generator._build_prompt().

class autorag.generate.Generator(model, api_key=None)

Bases: object

Base interface for RAG response generation.

Subclasses must implement generate(). The base class provides a default [idx] chunk prompt assembly via _build_prompt() that subclasses may reuse.

Parameters:

• model (str)

• api_key (str | None)

generate(question, context)

Return a natural-language answer to question grounded in context.

Parameters:

• question (str)

• context (list[Retrieved])

Return type:

str