voxflow 1.15.0 → 1.15.2

package/lib/commands/asr/local-transcribe.js

@@ -0,0 +1,125 @@
+/**
+ * ASR command — Local Whisper pipeline.
+ */
+
+'use strict';
+
+const fs = require('fs');
+const path = require('path');
+const { ASR_DEFAULTS } = require('../../core/config');
+const { getMediaInfo, extractAudioForAsr } = require('../../core/audio-extract');
+const { transcribeLocal } = require('../../core/whisper-local');
+const { formatSrt, formatPlainText, formatJson } = require('../../core/srt');
+const { FORMAT_EXTENSIONS, formatDuration, formatSize } = require('./helpers');
+
+/**
+ * Run ASR using local Whisper engine.
+ */
+async function asrLocal(opts) {
+  const {
+    input,
+    format = ASR_DEFAULTS.format,
+    output: userOutput,
+    model = 'base',
+    lang = ASR_DEFAULTS.lang,
+  } = opts;
+
+  console.log('\n=== VoxFlow ASR (Local Whisper) ===');
+
+  // Validate: local engine requires --input (no URL or mic)
+  if (!input) {
+    throw new Error(
+      'Local whisper engine requires --input <file>.\n' +
+      'URL and microphone input are cloud-only features.\n' +
+      'Use: voxflow asr --input <file> --engine local'
+    );
+  }
+
+  const localFile = path.resolve(input);
+  if (!fs.existsSync(localFile)) {
+    throw new Error(`Input file not found: ${localFile}`);
+  }
+
+  // Probe media info
+  const info = await getMediaInfo(localFile);
+  console.log(`Input:    ${path.basename(localFile)}`);
+  console.log(`Duration: ${formatDuration(info.durationMs)}`);
+  console.log(`Engine:   whisper (local)`);
+  console.log(`Model:    ${model}`);
+
+  // Extract audio to 16kHz WAV
+  console.log(`\n[1/2] Extracting audio (16kHz mono WAV)...`);
+  const extracted = await extractAudioForAsr(localFile);
+  console.log(`    OK (${formatSize(fs.statSync(extracted.wavPath).size)})`);
+
+  // Transcribe with local whisper
+  console.log(`[2/2] Whisper local recognition...`);
+  const startTime = Date.now();
+  const captions = await transcribeLocal(extracted.wavPath, { model, lang });
+  const elapsed = ((Date.now() - startTime) / 1000).toFixed(1);
+  console.log(`    OK (${elapsed}s, ${captions.length} segments)`);
+
+  // Clean up temp WAV
+  try { fs.unlinkSync(extracted.wavPath); } catch { /* ignore */ }
+
+  // Format output
+  let outputContent;
+  switch (format) {
+    case 'srt':
+      outputContent = formatSrt(captions);
+      break;
+    case 'txt':
+      outputContent = formatPlainText(captions);
+      break;
+    case 'json':
+      outputContent = formatJson(captions);
+      break;
+    default:
+      throw new Error(`Unknown format: ${format}. Use: srt, txt, json`);
+  }
+
+  // Determine output path
+  const ext = FORMAT_EXTENSIONS[format] || '.txt';
+  let outputPath;
+  if (userOutput) {
+    outputPath = path.resolve(userOutput);
+  } else {
+    const fname = path.basename(input, path.extname(input)) + ext;
+    outputPath = path.resolve(path.dirname(input), fname);
+  }
+
+  fs.writeFileSync(outputPath, outputContent, 'utf8');
+
+  // Summary
+  console.log(`\n=== Done ===`);
+  console.log(`Output:     ${outputPath}`);
+  console.log(`Captions:   ${captions.length}`);
+  console.log(`Duration:   ${formatDuration(info.durationMs)}`);
+  console.log(`Engine:     whisper (local, no quota used)`);
+
+  // Preview
+  if (captions.length > 0 && format !== 'json') {
+    console.log(`\n--- Preview ---`);
+    const preview = captions.slice(0, 3);
+    for (const cap of preview) {
+      const ts = formatDuration(cap.startMs);
+      const textSnippet = cap.text.length > 60
+        ? cap.text.slice(0, 57) + '...'
+        : cap.text;
+      console.log(`  ${ts}  ${textSnippet}`);
+    }
+    if (captions.length > 3) {
+      console.log(`  ... (${captions.length - 3} more)`);
+    }
+  }
+
+  return {
+    outputPath,
+    mode: 'local',
+    duration: info.durationMs / 1000,
+    captionCount: captions.length,
+    quotaUsed: 0,
+  };
+}
+
+module.exports = { asrLocal };

package/lib/commands/asr-jobs.js

@@ -0,0 +1,257 @@
+/**
+ * VoxFlow CLI — `asr-jobs` subcommand surface.
+ *
+ * Browse, inspect, cancel, and download long-running Azure ASR jobs.
+ * The transcription itself is started by `voxflow asr --engine azure`;
+ * this command is the "dashboard" — it talks to /api/asr/jobs/* and the
+ * ~/.config/voxflow/jobs/ pointer store and lets users pick up jobs they
+ * started in any other terminal session.
+ *
+ * Subcommands:
+ *   list                          List recent jobs (server-side, paginated)
+ *   show <jobId>                  Show one job (full transcript if succeeded)
+ *   cancel <jobId>                Cancel a running job (refunds quota)
+ *   download <jobId> [--output] [--format]   Re-emit the transcript locally
+ */
+
+const fs = require('fs');
+const path = require('path');
+const { API_BASE } = require('../core/config');
+const { getToken, getTokenInfo } = require('../core/auth');
+const { parseFlag, parseIntFlag, runWithRetry } = require('../core/args');
+const azureClient = require('../core/asr-jobs-client');
+const azureStore = require('../core/asr-jobs-store');
+const { formatSrt, formatPlainText, formatJson } = require('../core/srt');
+
+const FORMAT_EXT = { srt: '.srt', txt: '.txt', json: '.json' };
+
+// ─── Subcommand: list ────────────────────────────────────────────────────
+async function listJobs({ apiBase, token, limit, jsonOut }) {
+  const remote = await azureClient.listJobs(apiBase, token, { limit });
+  if (jsonOut) {
+    process.stdout.write(`${JSON.stringify(remote, null, 2)}\n`);
+    return { count: remote.length };
+  }
+  if (remote.length === 0) {
+    console.log('No ASR jobs found.');
+    return { count: 0 };
+  }
+  console.log(`\n=== ASR Jobs (${remote.length}) ===\n`);
+  console.log('  ID                                    Status     Duration   Created                Filename');
+  console.log('  ────────────────────────────────────  ─────────  ─────────  ─────────────────────  ────────');
+  for (const j of remote) {
+    const id = (j.id || '').padEnd(38);
+    const status = String(j.status || '?').padEnd(9);
+    const dur = j.result_duration_sec
+      ? `${Math.round(j.result_duration_sec)}s`
+      : (j.source_duration_sec ? `~${j.source_duration_sec}s` : '?');
+    const created = (j.created_at || '').slice(0, 19).replace('T', ' ').padEnd(21);
+    const file = j.source_filename || '';
+    console.log(`  ${id}  ${status}  ${dur.padEnd(9)}  ${created}  ${file}`);
+  }
+  return { count: remote.length };
+}
+
+// ─── Subcommand: show ────────────────────────────────────────────────────
+async function showJob({ apiBase, token, jobId, jsonOut }) {
+  const job = await azureClient.getJob(apiBase, token, jobId);
+  if (jsonOut) {
+    process.stdout.write(`${JSON.stringify(job, null, 2)}\n`);
+    return job;
+  }
+  console.log(`\n=== ASR Job ${job.id} ===`);
+  console.log(`Status:     ${job.status}${job.progress != null ? ` (${job.progress}%)` : ''}`);
+  console.log(`Provider:   ${job.provider}`);
+  console.log(`Source:     ${job.source_filename || '(unknown)'}${job.source_size ? ` (${formatSize(job.source_size)})` : ''}`);
+  console.log(`Language:   ${job.result_language || job.language || '?'}`);
+  console.log(`Duration:   ${job.result_duration_sec ? Math.round(job.result_duration_sec) + 's' : (job.source_duration_sec ? '~' + job.source_duration_sec + 's' : '?')}`);
+  console.log(`Quota:      ${job.quota_charged}`);
+  console.log(`Created:    ${job.created_at}`);
+  if (job.completed_at) console.log(`Completed:  ${job.completed_at}`);
+  if (job.error_code) console.log(`Error:      ${job.error_code} — ${job.error_message || ''}`);
+  if (job.status === 'succeeded' && job.result_json?.text) {
+    const preview = job.result_json.text.slice(0, 500);
+    console.log(`\n--- Transcript (${(job.result_segments_count || 0)} segments) ---`);
+    console.log(preview + (job.result_json.text.length > 500 ? '...' : ''));
+    console.log(`\nRun \`voxflow asr-jobs download ${job.id}\` to write the full transcript.`);
+  }
+  return job;
+}
+
+// ─── Subcommand: cancel ──────────────────────────────────────────────────
+async function cancelJob({ apiBase, token, jobId }) {
+  const job = await azureClient.cancelJob(apiBase, token, jobId);
+  if (job.status === 'cancelled') {
+    console.log(`Cancelled job ${job.id}. Quota refunded: ${job.quota_charged}.`);
+  } else {
+    console.log(`Job ${job.id} is already ${job.status} (no-op).`);
+  }
+  azureStore.deleteJob(jobId);
+  return job;
+}
+
+// ─── Subcommand: download ────────────────────────────────────────────────
+async function downloadJob({ apiBase, token, jobId, format, outputPath }) {
+  const job = await azureClient.getJob(apiBase, token, jobId);
+  if (job.status !== 'succeeded') {
+    throw new Error(`Job is ${job.status}, no transcript available yet. Status:\n  voxflow asr-jobs show ${jobId}`);
+  }
+  const segments = job.result_json?.segments || [];
+  const captions = segments.map((s, i) => ({
+    id: i + 1,
+    startMs: Math.round((s.start || 0) * 1000),
+    endMs: Math.round((s.end || 0) * 1000),
+    text: s.text || '',
+    speakerId: s.speaker || null
+  }));
+  let content;
+  switch (format) {
+    case 'srt':  content = formatSrt(captions); break;
+    case 'txt':  content = formatPlainText(captions, { includeSpeakers: true }); break;
+    case 'json': content = formatJson(captions); break;
+    default:     throw new Error(`Unknown format: ${format}`);
+  }
+  const ext = FORMAT_EXT[format] || '.txt';
+  // Build filename string before path.resolve — passing a template literal
+  // directly triggers ncc's asset-rewrite guard (see scripts/check-no-asset-rewrite.js).
+  const defaultFilename = 'asr-' + jobId + ext;
+  const target = outputPath
+    ? path.resolve(outputPath)
+    : path.resolve(defaultFilename);
+  fs.writeFileSync(target, content, 'utf8');
+  console.log(`Wrote ${captions.length} segments to ${target}`);
+  return { outputPath: target, captionCount: captions.length };
+}
+
+// ─── Helpers ─────────────────────────────────────────────────────────────
+function formatSize(bytes) {
+  if (!Number.isFinite(bytes)) return '?';
+  if (bytes < 1024) return `${bytes} B`;
+  if (bytes < 1024 * 1024) return `${(bytes / 1024).toFixed(1)} KB`;
+  return `${(bytes / 1024 / 1024).toFixed(1)} MB`;
+}
+
+// ─── Entry point ─────────────────────────────────────────────────────────
+async function handle(args) {
+  const subcommand = args[0];
+  if (!subcommand || subcommand.startsWith('-')) {
+    printHelp();
+    process.exit(subcommand ? 0 : 1);
+  }
+  if (!['list', 'show', 'cancel', 'download'].includes(subcommand)) {
+    console.error(`Error: unknown subcommand "${subcommand}". Use one of: list | show | cancel | download`);
+    process.exit(1);
+  }
+
+  const rest = args.slice(1);
+  const apiBase = parseFlag(rest, '--api') || API_BASE;
+  const explicitToken = parseFlag(rest, '--token');
+  const jsonOut = rest.includes('--json');
+  const limit = parseIntFlag(rest, '--limit') || 20;
+  const format = parseFlag(rest, '--format') || 'srt';
+  const outputPath = parseFlag(rest, '--output', '-o');
+  if (format && !['srt', 'txt', 'json'].includes(format)) {
+    console.error(`Error: --format must be one of: srt, txt, json (got: "${format}")`);
+    process.exit(1);
+  }
+
+  // First non-flag positional after the subcommand is the jobId (for show/cancel/download).
+  const positional = rest.filter((a) => !a.startsWith('-'));
+  // Skip values that follow consumed flags by reusing parseFlag — easier: positionals
+  // are the args that aren't flags AND aren't immediately after a known flag with value.
+  const jobId = pickJobId(rest);
+
+  const token = explicitToken || await getToken({ api: apiBase });
+  const info = getTokenInfo();
+  if (info && !jsonOut) console.log(`\x1b[32mLogged in as ${info.email}\x1b[0m`);
+
+  // opts is what runWithRetry mutates on 401 — it overwrites opts.token with
+  // a freshly-refreshed JWT before retrying the inner fn once.
+  const opts = { apiBase, token, jobId, limit, jsonOut, format, outputPath };
+
+  async function dispatch(o) {
+    if (subcommand === 'list') {
+      return await listJobs({ apiBase: o.apiBase, token: o.token, limit: o.limit, jsonOut: o.jsonOut });
+    }
+    if (!o.jobId) {
+      console.error(`Error: <jobId> is required for ${subcommand}.\nUsage: voxflow asr-jobs ${subcommand} <jobId>`);
+      process.exit(1);
+    }
+    if (subcommand === 'show')     return await showJob({ apiBase: o.apiBase, token: o.token, jobId: o.jobId, jsonOut: o.jsonOut });
+    if (subcommand === 'cancel')   return await cancelJob({ apiBase: o.apiBase, token: o.token, jobId: o.jobId });
+    if (subcommand === 'download') return await downloadJob({ apiBase: o.apiBase, token: o.token, jobId: o.jobId, format: o.format, outputPath: o.outputPath });
+    return null;
+  }
+
+  try {
+    return await runWithRetry(dispatch, opts, apiBase, explicitToken);
+  } catch (err) {
+    console.error(`Error: ${err.message}`);
+    process.exit(1);
+  }
+
+  // Unreachable — silences lint for "code path without return"
+  return positional;
+}
+
+// Walk argv and return the first positional that doesn't follow a value-taking flag.
+function pickJobId(args) {
+  const valueFlags = new Set(['--api', '--token', '--limit', '--format', '--output', '-o']);
+  let skip = false;
+  for (const a of args) {
+    if (skip) { skip = false; continue; }
+    if (valueFlags.has(a)) { skip = true; continue; }
+    if (a.startsWith('-')) continue;
+    return a;
+  }
+  return null;
+}
+
+function printHelp() {
+  console.log(`Usage: voxflow asr-jobs <subcommand> [options]
+
+Subcommands:
+  list                              List recent ASR jobs (server-side)
+  show <jobId>                      Show one job (with transcript preview)
+  cancel <jobId>                    Cancel a running job (refunds quota)
+  download <jobId>                  Write the transcript to a file
+
+Options:
+  --api <url>                       API base URL (default: ${API_BASE})
+  --token <jwt>                     Bypass token cache for one-off calls
+  --limit <n>                       (list) Number of jobs to fetch (default 20)
+  --format srt|txt|json             (download) Output format (default srt)
+  --output, -o <path>               (download) Output file path
+  --json                            (list/show) Emit raw JSON to stdout
+
+Examples:
+  voxflow asr-jobs list
+  voxflow asr-jobs show 6f3c2798-87bf-4367-bb4c-08b872e12bef
+  voxflow asr-jobs download 6f3c2798-… --format srt -o meeting.srt
+  voxflow asr-jobs cancel 6f3c2798-…`);
+}
+
+const meta = {
+  'asr-jobs': {
+    usage: '<list|show|cancel|download> [opts]',
+    description: 'Browse, inspect, cancel, and download long-running Azure ASR jobs',
+    options: [
+      `list                              List recent jobs`,
+      `show <jobId>                      Show one job (with preview)`,
+      `cancel <jobId>                    Cancel a running job`,
+      `download <jobId>                  Write the transcript locally`,
+      `--limit <n>                       (list) default 20`,
+      `--format srt|txt|json             (download) default srt`,
+      `--output, -o <path>               (download) target file path`,
+      `--json                            (list/show) raw JSON output`
+    ],
+    examples: [
+      'voxflow asr-jobs list',
+      'voxflow asr-jobs show 6f3c2798-87bf-4367-bb4c-08b872e12bef',
+      'voxflow asr-jobs download 6f3c2798-… --format srt -o meeting.srt',
+      'voxflow asr-jobs cancel 6f3c2798-…'
+    ]
+  }
+};
+
+module.exports = { handle, meta, listJobs, showJob, cancelJob, downloadJob };

package/lib/commands/asr.js

@@ -0,0 +1,11 @@
+/**
+ * VoxFlow CLI — ASR (Automatic Speech Recognition) command
+ *
+ * Thin dispatcher — delegates to engine-specific submodules in ./asr/.
+ * See asr/index.js for the orchestrator, asr/cloud-transcribe.js,
+ * asr/azure-transcribe.js, asr/local-transcribe.js for pipelines.
+ */
+
+'use strict';
+
+module.exports = require('./asr/index');