BasicRunPipeline

BasicRunPipeline Class Documentation

Overview

The BasicRunPipeline class handles the run process of document question answering. It is simple pipeline for beginners. It retrieves passages from a retrieval module, runs a Language Model (LLM) module to get an answer, and returns the answer and the retrieved passages.

Usage

Initialize

The BasicRunPipeline class is initialized with the following parameters:

• retrieval: Retrieval module to retrieve passages.

• llm: LLM module to get answer. If not provided, a default BasicLLM module is used.

retrieval = BM25Retrieval(save_path="./bm25.pkl")
pipeline = BasicRunPipeline(retrieval=retrieval, llm=OpenAI())

Run

The run variable is runnable of Langchain LCEL. So you can use all method for running pipeline, like invoke, stream, batch, and so on.

answer = pipeline.run.invoke({"question": "Where is the capital of Korea?"})

If you want to get used passages or relevance scores of retrieved passages, you can use get_passages_and_run method.

answers, passages, scores = pipeline.get_passages_and_run(["Where is the capital of Korea?"])