| import CRC32 from 'crc-32'; |
|
|
| export const parseFile = async (file) => { |
| if (file.type === 'application/json') { |
| return await parseJsonFile(file); |
| } else if (file.type === 'image/png') { |
| return await parsePngFile(file); |
| } else { |
| throw new Error('Unsupported file type'); |
| } |
| }; |
|
|
| const parseJsonFile = async (file) => { |
| const text = await file.text(); |
| const json = JSON.parse(text); |
|
|
| const character = extractCharacter(json); |
|
|
| return { |
| file, |
| json, |
| formats: detectFormats(json), |
| character |
| }; |
| }; |
|
|
| const parsePngFile = async (file) => { |
| const arrayBuffer = await file.arrayBuffer(); |
| const text = parsePngText(arrayBuffer); |
| const json = JSON.parse(text); |
|
|
| const image = URL.createObjectURL(file); |
| const character = extractCharacter(json); |
|
|
| return { |
| file, |
| json, |
| image, |
| formats: detectFormats(json), |
| character |
| }; |
| }; |
|
|
| const parsePngText = (arrayBuffer) => { |
| const textChunkKeyword = 'chara'; |
| const chunks = readPngChunks(new Uint8Array(arrayBuffer)); |
|
|
| const textChunk = chunks |
| .filter((chunk) => chunk.type === 'tEXt') |
| .map((chunk) => decodeTextChunk(chunk.data)) |
| .find((entry) => entry.keyword === textChunkKeyword); |
|
|
| if (!textChunk) { |
| throw new Error(`No PNG text chunk named "${textChunkKeyword}" found`); |
| } |
|
|
| try { |
| return new TextDecoder().decode(Uint8Array.from(atob(textChunk.text), (c) => c.charCodeAt(0))); |
| } catch (e) { |
| throw new Error('Unable to parse "chara" field as base64', e); |
| } |
| }; |
|
|
| const readPngChunks = (data) => { |
| const isValidPng = |
| data[0] === 0x89 && |
| data[1] === 0x50 && |
| data[2] === 0x4e && |
| data[3] === 0x47 && |
| data[4] === 0x0d && |
| data[5] === 0x0a && |
| data[6] === 0x1a && |
| data[7] === 0x0a; |
|
|
| if (!isValidPng) throw new Error('Invalid PNG file'); |
|
|
| let chunks = []; |
| let offset = 8; |
|
|
| while (offset < data.length) { |
| let length = |
| (data[offset] << 24) | (data[offset + 1] << 16) | (data[offset + 2] << 8) | data[offset + 3]; |
| let type = String.fromCharCode.apply(null, data.slice(offset + 4, offset + 8)); |
| let chunkData = data.slice(offset + 8, offset + 8 + length); |
| let crc = |
| (data[offset + 8 + length] << 24) | |
| (data[offset + 8 + length + 1] << 16) | |
| (data[offset + 8 + length + 2] << 8) | |
| data[offset + 8 + length + 3]; |
|
|
| if (CRC32.buf(chunkData, CRC32.str(type)) !== crc) { |
| throw new Error(`Invalid CRC for chunk type "${type}"`); |
| } |
|
|
| chunks.push({ type, data: chunkData, crc }); |
| offset += 12 + length; |
| } |
|
|
| return chunks; |
| }; |
|
|
| const decodeTextChunk = (data) => { |
| let i = 0; |
| const keyword = []; |
| const text = []; |
|
|
| for (; i < data.length && data[i] !== 0; i++) { |
| keyword.push(String.fromCharCode(data[i])); |
| } |
|
|
| for (i++; i < data.length; i++) { |
| text.push(String.fromCharCode(data[i])); |
| } |
|
|
| return { keyword: keyword.join(''), text: text.join('') }; |
| }; |
|
|
| const extractCharacter = (json) => { |
| function getTrimmedValue(json, keys) { |
| return keys |
| .map((key) => { |
| const keyParts = key.split('.'); |
| let value = json; |
| for (const part of keyParts) { |
| if (value && value[part] != null) { |
| value = value[part]; |
| } else { |
| value = null; |
| break; |
| } |
| } |
| return value && value.trim(); |
| }) |
| .find((value) => value); |
| } |
|
|
| const name = getTrimmedValue(json, ['char_name', 'name', 'data.name']); |
| const summary = getTrimmedValue(json, ['personality', 'title', 'data.description']); |
| const personality = getTrimmedValue(json, ['char_persona', 'description', 'data.personality']); |
| const scenario = getTrimmedValue(json, ['world_scenario', 'scenario', 'data.scenario']); |
| const greeting = getTrimmedValue(json, [ |
| 'char_greeting', |
| 'greeting', |
| 'first_mes', |
| 'data.first_mes' |
| ]); |
| const examples = getTrimmedValue(json, [ |
| 'example_dialogue', |
| 'mes_example', |
| 'definition', |
| 'data.mes_example' |
| ]); |
|
|
| return { name, summary, personality, scenario, greeting, examples }; |
| }; |
|
|
| const detectFormats = (json) => { |
| const formats = []; |
|
|
| if ( |
| json.char_name && |
| json.char_persona && |
| json.world_scenario && |
| json.char_greeting && |
| json.example_dialogue |
| ) |
| formats.push('Text Generation Character'); |
| if ( |
| json.name && |
| json.personality && |
| json.description && |
| json.scenario && |
| json.first_mes && |
| json.mes_example |
| ) |
| formats.push('TavernAI Character'); |
| if ( |
| json.character && |
| json.character.name && |
| json.character.title && |
| json.character.description && |
| json.character.greeting && |
| json.character.definition |
| ) |
| formats.push('CharacterAI Character'); |
| if ( |
| json.info && |
| json.info.character && |
| json.info.character.name && |
| json.info.character.title && |
| json.info.character.description && |
| json.info.character.greeting |
| ) |
| formats.push('CharacterAI History'); |
|
|
| return formats; |
| }; |
|
|