Work with pipeline

Use pipeline when you need to orchestrate retrieval, prompt assembly, and llm generation in a single ragpipeline.run() call — returns ragresult with the answer and citationrecord provenance.

Prerequisites

• Access to the project source code
• Familiarity with the files under src/attune_rag/pipeline.py

Steps

1. Understand the class hierarchy. Read the interfaces to see how pipeline is structured before extending or modifying. The key classes are:

   • RagResult in src/attune_rag/pipeline.py — Output of RagPipeline.run.
   • RagPipeline in src/attune_rag/pipeline.py — LLM-agnostic RAG pipeline.

2. Decide whether to extend or modify. If the class has subclasses, extend with a new one rather than changing the base. If it stands alone, modify directly.

3. Make your change. Follow existing patterns — naming, error handling, and logging style.

4. Run the related tests. Target with pytest -k "pipeline".

Key files

• src/attune_rag/pipeline.py
• src/attune_rag/__init__.py

Common modifications

Classes you are most likely to extend:

• RagResult in src/attune_rag/pipeline.py
• RagPipeline in src/attune_rag/pipeline.py