e2e

Documentation

Overview

Command e2e is an end-to-end tutorial for the recall library, covering the full lifecycle of a RAG application:

1. ingest a local Markdown corpus (directory loader -> ingest pipeline -> SQLite store with dedup and progress reporting),
2. search it (vector similarity and hybrid BM25+vector),
3. answer a question with the RAG pipeline and a mock LLM backend,
4. evaluate retrieval quality against a ground-truth dataset (Precision/Recall/MRR/NDCG@K), and
5. extract a knowledge graph and run multi-hop reasoning over it.

The example is fully deterministic and offline: it uses embedder.NewMockEmbedder and llm.NewMockBackend, so it needs no API keys, network access, or configuration.

Run it with:

go run ./example/e2e