Question

A developer is working in a large monorepo and notices that Copilot's inline completions become noticeably more relevant when they keep a few related files open in adjacent editor tabs. A teammate asks how Copilot decides what code to include when constructing the prompt sent to the model for a completion. Which statement best describes how Copilot builds the prompt for a code completion?

Accepted Answer

Copilot assembles a prompt on the client from local context — including the code around the cursor and relevant snippets from open/neighboring files — then sends it to the model, without indexing or uploading the entire repository.

Answer

Copilot uploads and indexes the full repository on GitHub's servers so the model can search every file before generating each completion.

Answer

Copilot only ever considers the single line directly above the cursor and ignores any other files or surrounding code entirely.

Answer

Copilot sends the raw model output directly to the editor without any client-side prompt construction, relying solely on the model's training data for context.