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

# Advanced config

> Build a custom JSON request body using Mustache templates.

<Warning>
  The endpoint must still return the final assistant text in
  `choices[0].message.content`.
</Warning>

Advanced mode lets you edit the full request body that Coreply sends. Use it when the built-in provider forms are not enough.

## Enable advanced mode

<Steps>
  <Step title="Choose Advanced Mode">
    In Coreply, open **API Provider** and choose **Advanced Mode**.
  </Step>

  <Step title="Fill in the endpoint">
    Enter the full **Request URL**. If the endpoint uses bearer auth, fill in
    **Authorization Bearer**.
  </Step>

  <Step title="Edit the templates">
    Update the **Body Template** and **Suggestion Template** fields.
  </Step>
</Steps>

The body template uses [Mustache](https://mustache.github.io/). Coreply supplies the following context to the template and the resulting string is the request body.

## Current context format

Templates should start from `contexts`. It contains the structured contexts that Coreply has collected for the current request.

| Field          | Type   | Description                                                                                               |
| -------------- | ------ | --------------------------------------------------------------------------------------------------------- |
| `contexts`     | list   | Structured context objects. Use this when your backend understands Coreply context objects.               |
| `contextsJson` | string | The same `contexts` value pre-serialized as JSON. Use this when your request body needs a raw JSON array. |

Strings inside `contexts` are plain strings. If you need JSON-safe interpolation inside a string value, use the compatibility fields below or send `contexts` as structured JSON.

### Context objects

Each item in `contexts` has these top-level fields:

| Field       | Type   | Description                           |
| ----------- | ------ | ------------------------------------- |
| `type`      | string | `chat` or `screen`                    |
| `profileId` | string | The profile that produced the context |
| `label`     | string | Optional label such as `messages`     |
| `dropRule`  | object | Context retention rule metadata       |
| `data`      | object | The actual context payload            |

### Chat contexts

When `type` is `chat`, `data` has this shape:

| Field        | Type   | Description                 |
| ------------ | ------ | --------------------------- |
| `data.id`    | string | Optional chat identifier    |
| `data.title` | string | Optional chat title         |
| `data.turns` | list   | Conversation turns in order |

Each item in `data.turns`:

| Field      | Type    | Description                 |
| ---------- | ------- | --------------------------- |
| `sender`   | string  | Optional sender name        |
| `userSent` | boolean | `true` for the user's turns |
| `messages` | list    | Messages in that turn       |

Each item in `messages`:

| Field   | Type   | Description          |
| ------- | ------ | -------------------- |
| `body`  | string | Message text         |
| `time`  | string | Optional timestamp   |
| `quote` | string | Optional quoted text |

### Screen contexts

When `type` is `screen`, `data` is a text tree:

| Field           | Type   | Description                              |
| --------------- | ------ | ---------------------------------------- |
| `data.text`     | string | Optional text at this node               |
| `data.children` | list   | Optional child nodes with the same shape |

## Suggestion template

After the API call completes, Coreply reads `choices[0].message.content` from the JSON response and exposes it to the suggestion template.

| Field                                          | Type   | Description                                                                    |
| ---------------------------------------------- | ------ | ------------------------------------------------------------------------------ |
| `assistantMessage`                             | string | The raw text returned by the model                                             |
| `assistantMessageAutoTrimCurrentTyping`        | string | `assistantMessage` with `currentTyping` trimmed from the start when it matches |
| `assistantMessageAutoTrimCurrentTypingTrimmed` | string | Same as above, but trims `currentTypingTrimmed` instead                        |

### Show the response as-is

```text theme={null}
{{assistantMessage}}
```

### Prepend the user's current typing

```text theme={null}
{{currentTyping.raw}}{{assistantMessage}}
```

## Compatibility fields

Coreply still exposes the older `TypingInfo.contextMap` fields for compatibility. Use them when you want to build a plain chat prompt inside the JSON body. They only cover chat history and typing helpers. They do not include screen contexts.

### String map fields

Compatibility `string` fields are exposed as a string map with four variants:

| Variant                         | Description                           |
| ------------------------------- | ------------------------------------- |
| `{{field.raw}}`                 | Original unescaped string             |
| `{{field.jsonEscaped}}`         | JSON-escaped string                   |
| `{{field.regexLiteral}}`        | Regex-literal form of the string      |
| `{{field.regexLiteralEscaped}}` | Regex-literal form, then JSON-escaped |

In JSON string values, use `.jsonEscaped`.

### Compatibility top-level fields

| Field                            | Type       | Description                                         |
| -------------------------------- | ---------- | --------------------------------------------------- |
| `currentTyping`                  | string map | What the user is currently typing                   |
| `currentTypingTrimmed`           | string map | `currentTyping` without the last incomplete token   |
| `currentTypingLastToken`         | string map | The last incomplete token the user is typing        |
| `currentTypingEndsWithSeparator` | boolean    | `true` when typing ends with a space or punctuation |
| `pastMessages`                   | list       | Chat history only, grouped into turns               |

`pastMessages` does not include screen contexts. Use `contexts` if you need them.

### The `pastMessages` list

`{{#pastMessages}}...{{/pastMessages}}` iterates over chat turns from oldest to newest.

| Field      | Type       | Description                             |
| ---------- | ---------- | --------------------------------------- |
| `sent`     | boolean    | `true` if the turn was sent by the user |
| `received` | boolean    | `true` if the turn was received         |
| `sender`   | string map | Sender name or role                     |
| `messages` | list       | Messages inside the turn                |

Each item in `messages`:

| Field      | Type       | Description                |
| ---------- | ---------- | -------------------------- |
| `content`  | string map | Message text               |
| `sent`     | boolean    | `true` if sent by the user |
| `received` | boolean    | `true` if received         |
| `sender`   | string map | Sender name or role        |

### Compatibility example

```json theme={null}
{
  "model": "gpt-5.6-luna",
  "messages": [
    {
      "role": "system",
      "content": "You are an AI texting assistant. Generate a suggested reply based on the conversation history and current typing. Output only the suggested text without quotation marks or extra formatting."
    },
    {
      "role": "user",
      "content": "Chat history:\n{{#pastMessages}}{{#messages}}{{#sent}}Me: {{/sent}}{{#received}}Them: {{/received}}{{content.jsonEscaped}}\n{{/messages}}{{/pastMessages}}{{#currentTyping}}Current typing: {{currentTyping.jsonEscaped}}{{/currentTyping}}{{^currentTyping}}Suggest a reply.{{/currentTyping}}"
    }
  ],
  "max_tokens": 50,
  "stream": false
}
```

### Compatibility example with assistant prefill

```json theme={null}
{
  "model": "<your-model-name>",
  "messages": [
    {
      "role": "system",
      "content": "You are a texting assistant. Output only the suggested message text."
    },
    {
      "role": "user",
      "content": "{{#pastMessages}}{{#messages}}{{#sent}}Me: {{/sent}}{{#received}}Them: {{/received}}{{content.jsonEscaped}}\n{{/messages}}{{/pastMessages}}"
    },
    {
      "role": "assistant",
      "content": "{{#currentTyping}}{{currentTyping.jsonEscaped}}{{/currentTyping}}"
    }
  ],
  "max_tokens": 60,
  "stream": false
}
```

<Tip>
  Enable **Show errors** under [Configuration](/configuration/settings) while
  you are debugging templates.
</Tip>