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

# Engines

> engine vs engine_account: which AI assistant ran a prompt vs which scraper produced the row.

Two related but distinct identifiers travel with every prompt and
every history row.

## `engine` — which AI assistant

This is the AI you queried. Lowercase. One of:

| Value     | Meaning          |
| --------- | ---------------- |
| `chatgpt` | OpenAI ChatGPT   |
| `gemini`  | Google Gemini    |
| `grok`    | xAI Grok         |
| `claude`  | Anthropic Claude |

**This is the value you pass to `?engine=` query parameters** on every
endpoint. Match exactly — values are case-sensitive.

## `engine_account` — which scraper produced the row

Only on the **prompts** table (not history). Tells you which pipeline
captured the data:

| Value           | Pipeline                                          |
| --------------- | ------------------------------------------------- |
| `BD_chatgpt`    | Bright Data scraper for ChatGPT (current default) |
| `BD_gemini`     | Bright Data scraper for Gemini                    |
| `BD_grok`       | Bright Data scraper for Grok                      |
| `BD_perplexity` | Bright Data scraper for Perplexity                |
| empty / null    | Legacy in-process Cloro consumer                  |

You'd typically only look at `engine_account` if you're debugging a
specific scrape (e.g. cross-referencing a `snapshot_id` in the Bright
Data console).

## Filtering examples

```bash theme={null}
# Just ChatGPT data
GET /universes/<id>/history?engine=chatgpt

# Compare Gemini vs Claude on the same day
GET /universes/<id>/metrics/ai_avg_trust_position?day=2026-04-27&engine=gemini
GET /universes/<id>/metrics/ai_avg_trust_position?day=2026-04-27&engine=claude
```

## What if you don't pass `?engine=`?

The endpoint returns rows for **every engine** the universe has data
for, ordered by `day DESC, engine`. If you only run ChatGPT, you'll
only see ChatGPT rows — there's no implicit join with engines that
have no data.