Jurist uses multiple agents designed to automate tasks commonly done by humans, like reading, editing, analyzing, or synthesizing. Jurist uses a manager agent to delegate or sequence tasks through agents in order to achieve the best output.
When agents use Large Language Models (LLMs) to accomplish a task, they are limited by the LLM's context window, which is the amount of information that the LLM can either consume or output. These context windows vary by model and provider, and are typically measured in tokens.
Tokens can be mapped to words, paragraphs, and pages and are directionally accurate. The actual words, paragraphs, and pages a model can handle will vary based on the content itself (eg. paragraphs vs tables) as well as things like font and font size.
Use the following table to understand the token-limited operations Jurist can perform and associated limits.
| Operation | Sub Operation | Approximate Limits |
|---|---|---|
| Editing | Redlining | 70,000 words |
| Editing from a template | 70,000 words | |
| Translating | 70,000 words or 300 paragraphs, whatever is lesser | |
| Document Analysis | 70,000 words |