```bash
pip install -e git+https://github.com/Tawlk/synt/#egg=synt
```

**Note: On your first run of any cli command a config will be copied into
~/.synt/config.py that you should configure. It uses sane defaults. The values should be
self-explanatory. This will only happen on the first run of synt.**

```bash
synt fetch --db_name "mysamples.db"
```

By default (with no db_name) it will be stored as 'samples.db'.

If you'd prefer to build a fresh sample db and have the time, just run collect with
the desired amount.

**Note: In order to collect you will require [kral](http://github.com/tawlk/kral), which is our
"social data gatherer" built in Python.**

```bash
synt collect --max_collect 10000 --db_name 'awesome.db'
```

**Note: You can also collect incrementally by providing the same db_name.**

A basic example of training

```bash
synt train 'samples.db' 20000
```

Train takes two required arguments: a training database (name), and the amount of
samples to train on.

At this point you might want to see the classifiers accuracy on the
training data.

```bash
synt accuracy
```

Accuracy takes a number of testing samples. By default 25% of your training
sample count will be used as the testing set. You can over-ride this by
providing the --test_samples argument.

The database used for these testing samples will be the same as the database
used to train. The testing samples will be new samples and can be
guaranteed to be samples the classifier hasn't already seen.

You should now have a trained classifier and its time to see
some classification of text.

```bash
synt guess
```

This will drop you into a synt prompt where you can write text and see
the score between -1 and 1.

You can alternativley also just classify text without having to drop into
a prompt:

```bash
synt guess --text "i like ponies and rainbows"
```