> ## Documentation Index
> Fetch the complete documentation index at: https://docs.parallel.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Decisions Best Practices

> Shared evidence, citation fields, and API limits for Decisions

## Group Related Questions

All questions in a request share one set of retrieved evidence. Group questions about
the same entity or topic; use separate requests for unrelated entities.

## Citations

Each object in the `citations` array has these fields:

| Field | Type | Meaning |
| - | - | - |
| `url` | `string` | Source page URL. |
| `title` | `string`, optional | Page title; omitted when unavailable. |
| `excerpts` | `string[]` | Passages included in the decision context, possibly shortened. Can be empty. |

Excerpts are included by default. Set `include_excerpts: false` to return
`"excerpts": []` for each citation. URLs and available titles are still returned;
the decision still uses the excerpts as evidence.

## Input Limits

| Setting | Limit |
| - | - |
| Mode | `fast` (default) |
| Questions per request | 1–100 |
| Options per choice question | 2–255 |
| Levels per score question | 2–10 |
| Input per question | 32,000 estimated tokens, including shared state |
| Rendered question | 5,000 characters, including its name, instructions, criteria, and formatting |

See the [API Reference](https://docs.parallel.ai/api-reference/decisions/make-web-grounded-typed-decisions.md)
for the complete request and response schemas.