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

# Running queries

> Compose a research question, tune its dimensions and metrics, and watch it run live.

Queries are how you ask Antlect questions, and the working surface of the **Discovery Terminal**. This guide covers composing a query, tuning it, and following its execution in the thread panel.

Questions are interpreted through the [Semantic Ontology Layer](/platform/semantic-ontology-layer), which is why plain language works and why results include records that never used your wording.

## Composing a query

Start a new query from the thread panel inside any project. Write the question the way you'd ask a colleague:

> Which research institutes in Australia have growing patent activity in battery chemistry, and who are their key people?

You don't need special syntax: Antlect interprets the entities, geography, and intent from natural language.

### What makes a good query

* **Name the domain**: "photonics", "battery chemistry", "mRNA delivery", rather than a broad field like "science".
* **Say what you want ranked**: people, institutions, funders, so the answer is organised the way you'll use it.
* **Include geography if it matters.** Antlect's coverage is deepest for New Zealand and Australia.

## Tuning a query

Before running, you can open the tuning controls to shape the analysis:

<AccordionGroup>
  <Accordion title="Dimension">
    The primary entity type the answer is organised around. Choose **People** for a who's-who, **Institutions** for a capability map, or **Funding** to follow the money.
  </Accordion>

  <Accordion title="Metrics">
    How results are ranked: publication output, citation impact, funding volume, collaboration breadth. You can combine metrics; the Summary view explains how each entity scored.
  </Accordion>
</AccordionGroup>

## Watching a query run

Queries execute inside the thread panel, in place: you never leave the workspace. As the analysis proceeds you'll see:

1. **Step trace**: each stage of the analysis appears as it starts, with elapsed timing.
2. **Expandable steps**: click a step to see what that stage examined.
3. **Completion**: the thread card links to the full results view when done.

Past runs stay in the thread, so a project accumulates a visible history of how its answers were produced.

## Re-running and iterating

Queries are cheap to iterate. If the Summary shows the question was too broad, refine the wording or narrow the dimension and run again: each run is kept as its own entry in the thread, and nothing is overwritten.

<Tip>
  Treat threads like a lab notebook: keep the exploratory runs. When a result is worth keeping, save its artifacts to the Library or a report rather than trying to reproduce it later.
</Tip>

## Next steps

<CardGroup cols={2}>
  <Card title="Reading results" icon="table-columns" href="/guides/results">
    Navigate the Summary, Publications, and Patents views.
  </Card>

  <Card title="Network graph" icon="circle-nodes" href="/guides/network-graph">
    See the same answer as a map of people and institutions.
  </Card>
</CardGroup>