Message
A Message is a piece of content and Annotations that are relevant to that Message.
Messages represent the content that is sent to / returned from LLMs. They contain the message text, any associated files with the content, the author of the content, and the role in which the content applies.
Annotations at the Message level directly apply to this specific message. For example, the annotation for “response formatting” will be at this level.
The role of the sender in the conversation (e.g., user, assistant).
system
, user
, assistant
, function
The content of the message, including text and any attachments.
The textual content of the message.
Any files or attachments included with the message.
The base64-encoded content of the file.
The MIME type of the file, indicating its format.
"text/plain"
A URL pointing to the location of the file in Scale's storage.
A URL pointing to the source location of the file.
The name of the file.
Chunks specific to this message.
The type of chunked data.
Annotations for a chunk of the message.
Represents a generic annotation.
Unique identifier for the annotation.
"an_abc123efg456"
Key for the annotation.
"formatting"
The type of the value and the possible_values, if they exist.
Title of the annotation.
"Response Formatting"
Further details about the question.
"Does the response contain issues with formatting?"
String representation of the possible options.
A string representation of the annotation.
["Major Issues", "Minor Issues", "No Issues"]
Unique identifier for an annotation.
"overall_quality"
Integer type annotation value.
3
The possible values for this annotation.
Integer type annotation value.
A plain text field.
The ID of the source system or user that sent the message.
"source_abc123"
Annotations specific to this message.
Represents a generic annotation.
Unique identifier for the annotation.
"an_abc123efg456"
Key for the annotation.
"formatting"
The type of the value and the possible_values, if they exist.
Title of the annotation.
"Response Formatting"
Further details about the question.
"Does the response contain issues with formatting?"
String representation of the possible options.
A string representation of the annotation.
["Major Issues", "Minor Issues", "No Issues"]
Unique identifier for an annotation.
"overall_quality"
Integer type annotation value.
3
The possible values for this annotation.
Integer type annotation value.
The name of the model that generated the message.
"my-model-123"
The temperature of the model that generated the message.
0.9
The maximum number of tokens the model can generate.
1000
The top-p value of the model that generated the message.
0.9
The top-k value of the model that generated the message.
4
{
"role": "system",
"content": {
"text": "<string>",
"reference_texts": [
{
"content": "<string>",
"category": "<string>",
"url": "<string>"
}
],
"attachments": [
{
"content": "aSDinaTvuI8gbWludGxpZnk=",
"mime_type": "text/plain",
"scale_url": "<string>",
"url": "<string>",
"name": "<string>"
}
],
"chunks": [
{
"type": "<string>",
"annotations": [
{
"id": "an_abc123efg456",
"key": "formatting",
"type": "<string>",
"title": "Response Formatting",
"description": "Does the response contain issues with formatting?",
"labels": [
"Major Issues",
"Minor Issues",
"No Issues"
],
"metadata": {
"criteria": "overall_quality"
},
"value": 3,
"possible_values": [
[
1,
2,
3
]
]
}
],
"text": "<string>"
}
]
},
"source_id": "source_abc123",
"model_parameters": {
"model": "my-model-123",
"temperature": 0.9,
"max_completion_tokens": 1000,
"top_p": 0.9,
"top_k": 4
},
"annotations": [
{
"id": "an_abc123efg456",
"key": "formatting",
"type": "<string>",
"title": "Response Formatting",
"description": "Does the response contain issues with formatting?",
"labels": [
"Major Issues",
"Minor Issues",
"No Issues"
],
"metadata": {
"criteria": "overall_quality"
},
"value": 3,
"possible_values": [
[
1,
2,
3
]
]
}
]
}