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

# Ask questions

> How to phrase questions and how Fella chooses deterministic tools.

Ask in plain language. The model coordinates a fixed set of read-only tools; the tools, not the model, perform data access and computation.

## Ask a question

<Steps>
  <Step title="State the result you need">
    Ask about the data as you would ask a person who knows the folder.

    ```text filename="question" theme={null}
    Is my spending increase seasonal, or was it a few large purchases?
    ```

    You do not need to name a file, write SQL, or choose a tool.
  </Step>

  <Step title="Add constraints">
    Include the period, exclusions, grouping, or definition that matters. For example: `Exclude transfers and compare monthly spending in 2025 with 2024.`
  </Step>

  <Step title="Inspect and refine">
    Open **Evidence** to confirm the source and computation. If the interpretation is wrong, state the correction in a follow-up.
  </Step>
</Steps>

## Match the question to the data

| Question shape                                                   | Usual path                 |
| ---------------------------------------------------------------- | -------------------------- |
| Total, average, count, filter, trend, or join                    | `run_sql`                  |
| Targeted word, phrase, or regular-expression search in documents | `grep_files`               |
| Broad summary or comparison across documents                     | `read_file`                |
| Median, standard deviation, correlation, or regression           | `run_python`               |
| A breakdown or time series worth seeing                          | `make_chart` after a query |

SQL is preferred because its result is compact and can often be re-run during verification. Python is an analysis escape hatch, not a file-management shell, and requires Python 3 on the computer.

## Ask for the shape, not just the headline

For a simple lookup, ask directly:

```text filename="question" theme={null}
How much did I spend on rent in 2024?
```

For a change, comparison, or correlation, ask for the explanation too:

```text filename="question" theme={null}
Did my spending increase because of a few outliers, or did most months move together?
```

The harness prompt tells the model to inspect whether a trend is broad-based, whether a correlation has enough points, and whether a difference is meaningful rather than reporting only one number.

## Follow-up questions

Follow-ups can reuse a bounded digest of the last three useful turns in that conversation, including recent headlines and successful SQL. Narrow a question when the first answer used the wrong interpretation:

```text filename="conversation" theme={null}
You included transfers. Exclude them and group the remaining spending by month.
```

If a correction represents a durable vocabulary rule, Fella can record it in folder memory. See [context and memory](/developer-platform/using-fella/context-and-memory).

## When Fella refuses

Past files cannot establish a future value. Questions such as "How many books will I finish next year?" should be declined rather than turned into a forecast. If the folder does not contain enough information, Fella says so.

Definitions and general explanations can be answered without a tool. A leading `Background:` line marks general knowledge, not workspace evidence, and the prompt bars it from supplying specific figures.

## Stop or retry

Press `Esc` or click the stop button beside the composer to stop. The run keeps evidence already collected and returns `Stopped.` While an answer is running, submitting a plain-text correction cancels it and re-asks with that correction. Use `/retry` after a transient provider error or after changing models.

<CardGroup cols={2}>
  <Card title="Inspect the evidence" icon="search-check" href="/developer-platform/using-fella/evidence">
    See the exact computation behind a response.
  </Card>

  <Card title="Learn the harness" icon="workflow" href="/concepts/harness">
    Read the loop and its limits in source-backed detail.
  </Card>
</CardGroup>