LlamaIndex

How to use it

Keep document metadata when splitting source documents into nodes. Inspect retrieved nodes independently of the generated answer to spot misses caused by chunking or filters.

Build a baseline with a small source collection and one retriever, then test any changes against the same questions. Select and configure local models explicitly if the workflow needs to run offline.

Compatibility

Python, Local inference, Hosted integrations

Components

Document ingestion, RAG pipeline, Retrieval orchestration

Access

Local Python framework. Provider integrations can require credentials.