Adding migrations and lot of work. Starting work on the demo scope

server/ocr.ts

@@ -0,0 +1,78 @@
+import { spawn } from 'node:child_process'
+
+export type TextExtractionResult = {
+  extractor: string
+  extractorVersion: string
+  language: string
+  status: 'succeeded' | 'unsupported' | 'failed'
+  text: string
+  error?: string
+}
+
+export interface TextExtractor {
+  provider: string
+  extract(file: { buffer: Buffer; mimetype: string }): Promise<TextExtractionResult>
+}
+
+function plainText(file: { buffer: Buffer; mimetype: string }, maximumCharacters: number): TextExtractionResult | null {
+  if (!file.mimetype.startsWith('text/')) return null
+  return { extractor: 'plain-text', extractorVersion: '1', language: 'und', status: 'succeeded', text: file.buffer.toString('utf8').slice(0, maximumCharacters) }
+}
+
+function runTesseract(command: string, buffer: Buffer, languages: string, pageSegmentationMode: string, timeoutMs: number) {
+  return new Promise<string>((resolve, reject) => {
+    const child = spawn(command, ['stdin', 'stdout', '-l', languages, '--psm', pageSegmentationMode], { stdio: ['pipe', 'pipe', 'pipe'] })
+    const stdout: Buffer[] = []
+    const stderr: Buffer[] = []
+    let settled = false
+    const timer = setTimeout(() => {
+      if (settled) return
+      settled = true
+      child.kill('SIGKILL')
+      reject(new Error(`OCR timed out after ${timeoutMs}ms`))
+    }, timeoutMs)
+    child.stdout.on('data', chunk => stdout.push(Buffer.from(chunk)))
+    child.stderr.on('data', chunk => stderr.push(Buffer.from(chunk)))
+    child.once('error', error => {
+      if (settled) return
+      settled = true
+      clearTimeout(timer)
+      reject(error)
+    })
+    child.once('close', code => {
+      if (settled) return
+      settled = true
+      clearTimeout(timer)
+      if (code === 0) resolve(Buffer.concat(stdout).toString('utf8').trim())
+      else reject(new Error(Buffer.concat(stderr).toString('utf8').trim() || `OCR exited with status ${code}`))
+    })
+    child.stdin.end(buffer)
+  })
+}
+
+export function createTextExtractorFromEnv(): TextExtractor {
+  const enabled = process.env.OCR_ENABLED !== 'false'
+  const command = process.env.OCR_COMMAND || 'tesseract'
+  const languages = process.env.OCR_LANGUAGES || 'nor+eng'
+  const extractorVersion = process.env.OCR_ENGINE_VERSION || 'tesseract-cli-5'
+  const pageSegmentationMode = process.env.OCR_PAGE_SEGMENTATION_MODE || '3'
+  const timeoutMs = Math.max(1_000, Number(process.env.OCR_TIMEOUT_MS || 20_000))
+  const maximumBytes = Math.max(1, Number(process.env.MAX_OCR_BYTES || 15 * 1024 * 1024))
+  const maximumCharacters = Math.max(1_000, Number(process.env.MAX_EXTRACTED_TEXT_CHARACTERS || 200_000))
+  return {
+    provider: enabled ? 'tesseract' : 'disabled',
+    async extract(file) {
+      const direct = plainText(file, maximumCharacters)
+      if (direct) return direct
+      if (!file.mimetype.startsWith('image/')) return { extractor: 'none', extractorVersion: '1', language: 'und', status: 'unsupported', text: '' }
+      if (!enabled) return { extractor: 'tesseract', extractorVersion, language: languages, status: 'unsupported', text: '' }
+      if (file.buffer.byteLength > maximumBytes) return { extractor: 'tesseract', extractorVersion, language: languages, status: 'failed', text: '', error: `Image exceeds the ${maximumBytes}-byte OCR limit` }
+      try {
+        const text = (await runTesseract(command, file.buffer, languages, pageSegmentationMode, timeoutMs)).slice(0, maximumCharacters)
+        return { extractor: 'tesseract', extractorVersion, language: languages, status: 'succeeded', text }
+      } catch (error) {
+        return { extractor: 'tesseract', extractorVersion, language: languages, status: 'failed', text: '', error: error instanceof Error ? error.message : 'OCR failed' }
+      }
+    },
+  }
+}