# Configuration

Ragit is highly configurable. The config files can be found at `.ragit/configs`, but I don't recommend you modifying it manually. If you have modified it manually and have trouble accessing a knowledge-base, try `rag check --recover`.

## Global Configuration

You can set global configuration defaults by placing configuration files in `~/.config/ragit/`. When initializing a new ragit repository, it will check for the following files:

- `~/.config/ragit/api.json` - For API configuration
- `~/.config/ragit/build.json` - For build configuration
- `~/.config/ragit/query.json` - For query configuration

These files can contain a subset of the configuration fields that you want to override. You don't need to include all fields - any fields not specified will use the default values. For example, if you only want to override the `model` and `dump_log` fields in api.json, your file might look like:

```json
{
  "model": "gpt-4o",
  "dump_log": true
}
```

Any values found in these files will override the default values when creating a new repository. This allows you to have consistent configuration across all your ragit repositories.

## `config` command

A recommended way of reading/writing config is `rag config` command.

`rag config --get <KEY>` shows you a value. For example, `rag config --get model` tells you which model you're using.

`rag config --get-all` shows you all the configs.

`rag config --set <KEY> <VALUE>` allows you to set a value.

## Reference

- chunk_size: int (number of characters)
    - default: 4000
    - Ragit tries its best to make each chunk smaller than this.
    - `chunk_size` and `slide_len` isn't always perfect because ragit can handle images. It cannot divide an image into 2 pieces, so an image at the end might make a chunk bigger than `chunk_size`.
- slide_len: int (number of characters)
    - default: 1000
    - There's a sliding window between 2 chunks. Each sliding window has this length.
    - `chunk_size` and `slide_len` isn't always perfect because ragit can handle images. It cannot divide an image into 2 pieces, so an image at the end might make a chunk bigger than `chunk_size`.
- image_size: int
    - default: 2000
    - If it's 2000, ragit treats an image as 2000 characters (when calculating `chunk_size` and `slide_len`).
- min_summary_len: int (number of characters)
    - default: 200
    - Ragit uses pdl schema to force LLMs generate summaries longer than this.