49 lines
1.4 KiB
TypeScript
49 lines
1.4 KiB
TypeScript
/**
|
|
* This utility is to extract cortexso model description from README.md file
|
|
* @returns
|
|
*/
|
|
export const extractDescription = (text?: string) => {
|
|
if (!text) return text
|
|
const normalizedText = removeYamlFrontMatter(text)
|
|
const overviewPattern = /(?:##\s*Overview\s*\n)([\s\S]*?)(?=\n\s*##|$)/
|
|
const matches = normalizedText?.match(overviewPattern)
|
|
let extractedText =
|
|
matches && matches[1]
|
|
? matches[1].trim()
|
|
: normalizedText?.slice(0, 500).trim()
|
|
|
|
// Remove image markdown syntax 
|
|
extractedText = extractedText?.replace(/!\[.*?\]\(.*?\)/g, '')
|
|
|
|
// Remove <img> HTML tags
|
|
extractedText = extractedText?.replace(/<img[^>]*>/g, '')
|
|
|
|
return extractedText
|
|
}
|
|
/**
|
|
* Remove YAML (HF metadata) front matter from content
|
|
* @param content
|
|
* @returns
|
|
*/
|
|
export const removeYamlFrontMatter = (content: string): string => {
|
|
return content.replace(/^---\n([\s\S]*?)\n---\n/, '')
|
|
}
|
|
|
|
/**
|
|
* Extract model name from repo path, e.g. cortexso/tinyllama -> tinyllama
|
|
* @param modelId
|
|
* @returns
|
|
*/
|
|
export const extractModelName = (model?: string) => {
|
|
return model?.split('/')[1] ?? model
|
|
}
|
|
|
|
/**
|
|
* Extract model name from repo path, e.g. https://huggingface.co/cortexso/tinyllama -> cortexso/tinyllama
|
|
* @param modelId
|
|
* @returns
|
|
*/
|
|
export const extractModelRepo = (model?: string) => {
|
|
return model?.replace('https://huggingface.co/', '')
|
|
}
|