Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

10.8 KB
Raw Blame
import {
  fileNameFromUri,
  joinUri,
  normalizeDirUri,
  normalizeFileUri,
} from '#/lib/viking-uri'
import type { VikingFileType, VikingFsEntry } from '../-types/viking-fm'

export {
  fileNameFromUri,
  joinUri,
  normalizeDirUri,
  normalizeFileUri,
  parentUri,
} from '#/lib/viking-uri'

function isRecord(value: unknown): value is Record<string, unknown> {
  return value !== null && typeof value === 'object' && !Array.isArray(value)
}

const SIZE_REGEX = /^([\d.]+)\s*([kmgtp]?i?b?)$/

const SIZE_MULTIPLIERS: Record<string, number> = {
  b: 1,
  kb: 1024,
  kib: 1024,
  mb: 1024 ** 2,
  mib: 1024 ** 2,
  gb: 1024 ** 3,
  gib: 1024 ** 3,
  tb: 1024 ** 4,
  tib: 1024 ** 4,
  pb: 1024 ** 5,
  pib: 1024 ** 5,
}

const IMAGE_EXTS = [
  'jpg',
  'jpeg',
  'png',
  'gif',
  'webp',
  'svg',
  'ico',
  'bmp',
  'avif',
]
const MARKDOWN_EXTS = ['md', 'markdown', 'mdx']
const JSONL_EXTS = ['jsonl', 'ndjson']
const CODE_EXTS = [
  'js',
  'ts',
  'jsx',
  'tsx',
  'mjs',
  'cjs',
  'py',
  'go',
  'rs',
  'java',
  'kt',
  'c',
  'h',
  'hpp',
  'cpp',
  'json',
  'yaml',
  'yml',
  'toml',
  'xml',
  'html',
  'css',
  'scss',
  'less',
  'sh',
  'bash',
  'zsh',
  'fish',
  'sql',
  'graphql',
  'proto',
]
const BINARY_EXTS = [
  'pdf',
  'zip',
  'gz',
  'tgz',
  'tar',
  '7z',
  'rar',
  'mp3',
  'wav',
  'mp4',
  'mov',
  'avi',
  'mkv',
  'woff',
  'woff2',
  'ttf',
  'otf',
  'exe',
  'dll',
  'so',
  'dylib',
  'bin',
]
const TEXT_FILES = ['readme', 'license', 'dockerfile', 'makefile']
const HIDDEN_SUMMARY_FILENAMES = new Set([
  '.abstract',
  '.abstract.md',
  '.overview',
  '.overview.md',
])

function pickFirstNonEmpty(values: Array<unknown>): unknown {
  for (const value of values) {
    if (value !== undefined && value !== null && String(value).trim() !== '') {
      return value
    }
  }
  return ''
}

export function parseSizeToBytes(value: unknown): number | null {
  if (typeof value === 'number') {
    return Number.isFinite(value) ? value : null
  }

  const text = String(value ?? '').trim()
  if (!text) {
    return null
  }

  if (/^\d+$/.test(text)) {
    const direct = Number(text)
    return Number.isFinite(direct) ? direct : null
  }

  const normalized = text.replace(/,/g, '').toLowerCase()
  const match = normalized.match(SIZE_REGEX)
  if (!match) {
    return null
  }

  const amount = Number(match[1])
  const unit = match[2]
  if (!Number.isFinite(amount)) {
    return null
  }

  const multiplier = SIZE_MULTIPLIERS[unit || 'b']
  if (!multiplier) {
    return null
  }

  return Math.round(amount * multiplier)
}

const TIME_ONLY_RE = /^\d{1,2}:\d{2}(:\d{2})?(\s*[apAP][mM])?$/
const DATE_ONLY_RE = /^\d{4}-\d{2}-\d{2}$/

function todayUtcDateKey(): string {
  // Backend simplifies modTime against UTC's "today", not the user's
  // local today. We need to match that reference so time-only values
  // bind to the right UTC moment regardless of the viewer's tz.
  const d = new Date()
  const pad = (n: number) => String(n).padStart(2, '0')
  return `${d.getUTCFullYear()}-${pad(d.getUTCMonth() + 1)}-${pad(d.getUTCDate())}`
}

export function parseModTimeToTs(value: unknown): number | null {
  if (typeof value === 'number') {
    return Number.isFinite(value) ? value : null
  }

  const text = String(value ?? '').trim()
  if (!text) {
    return null
  }

  // Time-only ("HH:MM:SS", "12:34am" etc) — backend's format_simplified
  // returns this only when the entry's UTC date matches UTC's today.
  // Combine with UTC today + explicit Z so the moment is anchored in UTC,
  // and downstream Date methods naturally render it in the viewer's local
  // tz (no double-shift, no day mis-attribution).
  if (TIME_ONLY_RE.test(text)) {
    const normalized = text.replace(/\s+/g, '').toUpperCase()
    const ts = Date.parse(`${todayUtcDateKey()}T${normalized}Z`)
    if (Number.isFinite(ts)) {
      return ts
    }
  }

  // Date-only ("YYYY-MM-DD") — backend's format_simplified returns this
  // when the entry is older than UTC's today. Treat as local midnight on
  // that date for sort purposes (we have no sub-day precision anyway).
  if (DATE_ONLY_RE.test(text)) {
    const [year, month, day] = text.split('-').map(Number)
    if (year && month && day) {
      return new Date(year, month - 1, day).getTime()
    }
  }

  const direct = Date.parse(text)
  if (Number.isFinite(direct)) {
    return direct
  }

  const fallback = Date.parse(text.replace(' ', 'T'))
  if (Number.isFinite(fallback)) {
    return fallback
  }

  return null
}

export function formatModTime(value: unknown): string {
  const text = String(value ?? '').trim()
  if (!text) return ''
  // Date-only inputs (older entries) — the backend has already dropped
  // sub-day precision, so we render the date as-is without faking an
  // 00:00 time component.
  if (DATE_ONLY_RE.test(text)) {
    return text
  }
  const ts = parseModTimeToTs(text)
  if (ts === null) return text
  const d = new Date(ts)
  const pad = (n: number) => String(n).padStart(2, '0')
  return `${d.getFullYear()}-${pad(d.getMonth() + 1)}-${pad(d.getDate())} ${pad(d.getHours())}:${pad(d.getMinutes())}`
}

export function detectFileType(uri: string): VikingFileType {
  const name = fileNameFromUri(uri).toLowerCase()
  const ext = name.includes('.') ? name.split('.').pop() || '' : ''

  if (uri.endsWith('/')) {
    return 'directory'
  }

  if (IMAGE_EXTS.includes(ext)) {
    return 'image'
  }
  if (MARKDOWN_EXTS.includes(ext)) {
    return 'markdown'
  }
  if (JSONL_EXTS.includes(ext)) {
    return 'jsonl'
  }
  if (CODE_EXTS.includes(ext)) {
    return 'code'
  }

  if (!ext && TEXT_FILES.includes(name)) {
    return 'text'
  }

  if (isLikelyBinary(uri)) {
    return 'binary'
  }

  return 'text'
}

export function isLikelyBinary(uri: string): boolean {
  const name = fileNameFromUri(uri).toLowerCase()
  const ext = name.includes('.') ? name.split('.').pop() || '' : ''

  return BINARY_EXTS.includes(ext)
}

export function shouldAutoRead(
  entry: Pick<VikingFsEntry, 'isDir' | 'uri' | 'sizeBytes'>,
  maxAutoReadBytes = 2 * 1024 * 1024,
  requireKnownSize = false,
): {
  shouldRead: boolean
  reason?: 'binary' | 'too-large' | 'unknown-size'
} {
  if (entry.isDir) {
    return { shouldRead: false }
  }

  const fileType = detectFileType(entry.uri)
  if (fileType === 'image' || isLikelyBinary(entry.uri)) {
    return { shouldRead: false, reason: 'binary' }
  }

  if (requireKnownSize && entry.sizeBytes === null) {
    return { shouldRead: false, reason: 'unknown-size' }
  }

  if (entry.sizeBytes !== null && entry.sizeBytes > maxAutoReadBytes) {
    return { shouldRead: false, reason: 'too-large' }
  }

  return { shouldRead: true }
}

export function normalizeFsEntry(
  item: unknown,
  currentUri: string,
): VikingFsEntry {
  if (typeof item === 'string') {
    const isDir = item.endsWith('/')
    const uri = joinUri(currentUri, item)
    const normalizedUri = isDir ? normalizeDirUri(uri) : normalizeFileUri(uri)

    return {
      uri: normalizedUri,
      name: fileNameFromUri(normalizedUri),
      isDir,
      size: '',
      sizeBytes: null,
      modTime: '',
      modTimestamp: null,
      abstract: '',
      overview: '',
    }
  }

  if (isRecord(item)) {
    const label = String(
      pickFirstNonEmpty([
        item.name,
        item.path,
        item.relative_path,
        item.uri,
        item.id,
        'unknown',
      ]),
    )

    const isDir =
      Boolean(item.is_dir) ||
      Boolean(item.isDir) ||
      item.type === 'dir' ||
      item.type === 'directory' ||
      label.endsWith('/')

    const rawUri = String(
      pickFirstNonEmpty([item.uri, item.path, item.relative_path, label]),
    )
    const joinedUri = joinUri(currentUri, rawUri)
    const normalizedUri = isDir
      ? normalizeDirUri(joinedUri)
      : normalizeFileUri(joinedUri)

    const sizeRaw = pickFirstNonEmpty([
      item.size,
      item.size_bytes,
      item.content_length,
      item.contentLength,
    ])
    const modRaw = pickFirstNonEmpty([
      item.modTime,
      item.mod_time,
      item.modified_at,
      item.modifiedAt,
      item.updated_at,
      item.updatedAt,
    ])

    return {
      uri: normalizedUri,
      name: String(
        pickFirstNonEmpty([item.name, fileNameFromUri(normalizedUri)]),
      ),
      isDir,
      size: String(sizeRaw ?? ''),
      sizeBytes: parseSizeToBytes(sizeRaw),
      modTime: formatModTime(modRaw),
      modTimestamp: parseModTimeToTs(modRaw),
      abstract: String(
        pickFirstNonEmpty([item.abstract, item.summary, item.description]),
      ),
      overview: String(pickFirstNonEmpty([item.overview, item.l1, item.L1])),
    }
  }

  const fallbackUri = normalizeFileUri(joinUri(currentUri, String(item ?? '')))
  return {
    uri: fallbackUri,
    name: fileNameFromUri(fallbackUri),
    isDir: false,
    size: '',
    sizeBytes: null,
    modTime: '',
    modTimestamp: null,
    abstract: '',
    overview: '',
  }
}

function isHiddenSummaryEntry(entry: VikingFsEntry): boolean {
  return !entry.isDir && HIDDEN_SUMMARY_FILENAMES.has(entry.name)
}

export function normalizeFsEntries(
  result: unknown,
  currentUri: string,
): Array<VikingFsEntry> {
  const normalizedCurrentUri = normalizeDirUri(currentUri)

  if (Array.isArray(result)) {
    return result
      .map((item) => normalizeFsEntry(item, normalizedCurrentUri))
      .filter((entry) => !isHiddenSummaryEntry(entry))
  }

  if (isRecord(result)) {
    const buckets = [
      result.entries,
      result.items,
      result.children,
      result.results,
      result.nodes,
    ]
    for (const bucket of buckets) {
      if (Array.isArray(bucket)) {
        return bucket
          .map((item) => normalizeFsEntry(item, normalizedCurrentUri))
          .filter((entry) => !isHiddenSummaryEntry(entry))
      }
    }
  }

  return []
}

export function formatSize(
  value: unknown,
  options?: { maximumFractionDigits?: number; fallback?: string },
): string {
  const maximumFractionDigits = options?.maximumFractionDigits ?? 1
  const fallback = options?.fallback ?? '-'
  const bytes = parseSizeToBytes(value)

  if (bytes === null || bytes < 0) {
    return fallback
  }

  if (bytes < 1024) {
    return `${bytes} B`
  }

  const units = ['KB', 'MB', 'GB', 'TB', 'PB']
  let scaled = bytes
  let unitIndex = -1

  while (scaled >= 1024 && unitIndex < units.length - 1) {
    scaled /= 1024
    unitIndex += 1
  }

  return `${scaled.toFixed(maximumFractionDigits)} ${units[unitIndex]}`
}

export function normalizeReadContent(result: unknown): string {
  if (typeof result === 'string') {
    return result
  }

  if (Array.isArray(result)) {
    return result.map((item) => String(item)).join('\n')
  }

  if (isRecord(result)) {
    const content = pickFirstNonEmpty([
      result.content,
      result.text,
      result.body,
      result.value,
      result.data,
    ])
    if (typeof content === 'string') {
      return content
    }
  }

  if (result === undefined) {
    return ''
  }

  return JSON.stringify(result, null, 2)
}