Flashcards

A worked example: 12 CSV files, 35 columns, ~72,000 rows, 30 MB. AI in SharePoint reads, analyzes, and produces an interactive HTML report in about seven minutes.

A worker agent that writes and runs code in a sandbox against SharePoint files. Different from Copilot, which is a chat/RAG experience grounded in search.

The read-tool flag that tells SharePoint to keep file content in tool storage instead of returning it to the model. The mechanism that prevents context overflow.

The token budget the LLM can see at once. Even modern windows cannot fit 30 MB of data — and even if they could, the model still gets confused.

Final tool call in the pipeline. Writes the sandbox-generated artifact into the SharePoint document library where the user can open it.

First tool call in a typical run. Returns file metadata matching the prompt — names, sizes, columns — without transferring content.

Retrieval-Augmented Generation. The LLM searches a corpus, retrieves chunks, and reasons over them. Powers Copilot; cannot do precise math at large scale.

Executes the JavaScript the model has generated against data in tool storage. Produces a new artifact — JSON analytics, an image, or an HTML file.

A secure execution boundary where agent-generated code runs against data the user supplied. Cannot call external services or mutate other content.

An HTML file that includes all CSS, JS, and data inline. No CDN dependencies, no live data source. Portable, shareable, snapshot in time.

The classic "how many Rs in strawberry" failure. Demonstrates why letting the LLM write code, not the answer, produces reliable results for any countable task.

A single step the agent invokes — find files, read files, run code, create file. The model chains tool calls to build a workflow.

Server-side storage that holds file content during an agent run. The model references the data by handle; the bytes never enter the context window.

The strict edge around the sandbox. Data and code-execution stay inside. No outbound network, no mutation of other content, no escape — only the produced artifact crosses out.

An AI agent that delivers work artifacts — files, reports — by writing and executing code, rather than predicting answers from a trained model.