Context window

The context window is the amount of text a language model can consider at once, including your instructions, documents and its own output.

Written by Reflective Data experts · updated Oct 5, 2026

Bigger windows allow longer documents and more history, but cost more and can dilute focus. Good systems decide carefully what to put in the window, using retrieval rather than pasting everything.

Size is measured in tokens.

Large language model (LLM)Retrieval-augmented generation (RAG)Token

// how we can help

We build company brains, agent harnesses and AI bots on top of data you can trust. AI services → · The data platform behind it

Multi-agent system

A multi-agent system uses several AI agents with different roles that collaborate on a task.

chunking.md

Chunking

Chunking is splitting documents into smaller pieces so they can be indexed and retrieved for AI…

ai-evals.md

AI evaluations (evals)

AI evaluations are tests that measure how well an AI system performs on realistic tasks.

chatbot.md

Chatbot

A chatbot is a software assistant that converses with users in text or voice.

gemini.md

Gemini

Gemini is Google family of AI models and assistant, available across Google products and through APIs.

human-in-the-loop.md

Human in the loop

Human in the loop means a person reviews or approves an AI output or action before…

← full analytics dictionary