Fighting Document Bloat
Retrieved documents often contain a lot of irrelevant text. **Contextual Compression** uses a model to extract only the sentences or paragraphs that directly answer the user's query. This reduces token costs and ensures the agent isn't distracted by "Noise" during its reasoning process.
The Compressor-Retriever Loop
LangChain provides utilities like