# pandas functions that answer questions about text

A pandas Series goes in, and a Series with the same index comes back in `Call.value`.

## Install

| Channel |
| --- |
| `pip install thinkthen pandas` |

Set `THINKTHEN_API_KEY` in the environment.

## A first call

```
import pandas as pd
import thinkthen as tt

question = "Does the customer ask for a refund?"
refund = tt.question(decide=question, threshold=(0.2, 0.8))

messages = pd.Series(
    [
        "Please refund my order. It arrived broken.",
        "I want to send this back.",
    ],
    name="message",
)
is_refund = tt.decide(refund, messages).value
assert is_refund.dtype == "boolean"
assert is_refund.tolist() == [True, pd.NA]

for_a_person = messages[is_refund.isna()]
assert for_a_person.tolist() == [
    "I want to send this back.",
]
```

Not sure comes back as `pd.NA`. Branch on it and send those cases to a person.

## Run facts

`Call.facts` counts this call.

## Errors

Every failure raises a `ThinkThenError`, as in Python. Its subclass names the kind.

## Settings and recording

`tt.Engine` takes each setting as a keyword, as in Python. `record=` writes a recording to a folder. `replay=` answers from that recording with no connection. The [Settings](/install/settings/) page lists every setting.

## Backends

pandas runs on the Python engine, so it reaches each backend the same way. The [Python page](/install/python/#backends) shows Jev, Liquid d1 and Ollama.

## What is different in pandas

- `decide`, `choose`, `score` and `tag` take a Series and keep its index and name.
- `decide` gives a `boolean` Series, `score` gives `Float64`, `choose` gives `string`, and `tag` gives one list of labels per row.
- `annotate` takes a DataFrame with `on=` and adds one column per question.

On GitHub: [github.com/botassembly/thinkthen](https://github.com/botassembly/thinkthen)