Skip to content

Instant Voice Agent Generator

Auto-generate AI voice agents from any website URL. Crawls site → builds knowledge base → creates Dograh Cloud voice agent → generates split-view companion page → auto-generates AI promo video.

  • Status: 🟢 Active
  • Started: August 17, 2026
  • URL: https://demo.converso.work
  • Server: 76.13.246.122, port 8936
  • Voice Platform: Dograh Cloud (app.dograh.com)
  • Video Platform: MoneyPrinterTurbo (port 8937)
  • Details: [[D-insights Voice Agent Platform]]

Architecture

User enters URL → URL normalization → Playwright crawler (SPA-aware) → KB content
  → Dograh Cloud API: Create workflow + Upload KB + Attach to workflow
  → Create draft → Publish → Get embed token
  → Generate companion page (split-view: site iframe + voice widget)
  → Phase 7: MoneyPrinterTurbo auto-generates 9:16 promo video from KB content
  → Companion page shows "Download Video" button when ready

Components

Component Technology Port/URL
Web App FastAPI + Jinja2 8936 (demo.converso.work)
Crawler Playwright + Chromium (SPA scrolling, hash anchors) —
Voice Platform Dograh Cloud api.dograh.com
Video Generator MoneyPrinterTurbo 8081 (internal), 8937 (SSL)
Video LLM Kimi K2.7 Code (Ollama Cloud) ollama.com/v1
Video TTS Edge TTS (en-US-ChristopherNeural) —
Video Footage Pexels stock videos —
TURN Metered TURN (ports 80/443) global.relay.metered.ca
Web Server Nginx (SSL via Let's Encrypt) —
Systemd voice-agent-generator.service, moneyprinter.service —

URLs

  • Landing: https://demo.converso.work/
  • Companion: https://demo.converso.work/companion/{workflow_id}?site={url}
  • Video Download: https://76.13.246.122.sslip.io:8937/tasks/{task_id}/final-1.mp4
  • Video API (same-origin): /api/video-file/{mpt_task_id}

Active Agents

Agent Workflow URL Voice Language Promo Video
Kiara (Technofire) 10705 technofire.com Janvi (Dograh) English ✅ 15MB generated
Smartline IT 10431 smartlineit.com Janvi (Dograh) English Pending
GSB Temple Trust 10457 gsbtempletrust.com Janvi (Dograh) Marathi/Hindi/English Pending
Cathedral School 10463 cathedral-school.com Janvi (Dograh) English Pending

Key Features

  • Playwright crawler — JS-rendered SPAs, hash-anchor scrolling, priority path probing
  • Per-site workflows — each website gets its own Dograh workflow + KB
  • Split-view companion — live site in iframe (left) + voice widget (right)
  • Headless mode — custom call controls, no floating widget button
  • Near-live transcripts — polls Dograh API every 3s, shows text ~10s after call ends
  • Barge-in — caller can interrupt agent mid-sentence (provisional_vad, 0.5s pause)
  • 2-minute max calls — auto-disconnect to conserve credits
  • LifeOS theme — Inter font, dark #0d2745, primary #1a5276, accent #3498db
  • D-insights branding — logo, contact footer (+91 9820370923, www.d-insights.global)
  • 8-hour demo expiry — demos auto-expire
  • URL auto-normalization — accepts any format, normalizes to https://www.domain.tld/
  • Animated progress bar — creating-v2.html with shimmer gradient, expandable steps
  • AI Promo Video — auto-generates 9:16 portrait video after agent creation (Kimi K2.7 Code script + Edge TTS voiceover + Pexels footage + FFmpeg render)
  • Regenerate Video — companion page "Generate Video" button for existing agents via POST /api/regenerate-video/{job_id}
  • Background KB updater — continuous_scraper.py with CDP-based SPA crawling

Promo Video Pipeline (Phase 7)

Agent created → POST /api/v1/videos to MoneyPrinterTurbo
  → Kimi K2.7 Code writes script from site content
  → Pexels API downloads stock footage (9:16 portrait)
  → Edge TTS generates voiceover (en-US-ChristopherNeural)
  → Subtitle generation with timestamps
  → FFmpeg merges clips + audio + subtitles → final-1.mp4
  → Companion page polls /api/promo-video/{job_id} every 30s
  → "Download Video" button appears when ready
  • Generation time: ~5-7 min per video on CPU
  • Output: 1080×1920 portrait, 15-47MB
  • API: POST /api/regenerate-video/{job_id} to trigger for existing agents
  • Same-origin proxy: /api/video-file/{mpt_task_id} (avoids CORS/popup issues)

Dograh Cloud Details

  • API Key: [REDACTED]
  • Service Key: [REDACTED] (mps_sk_...)
  • Pricing: $0.08/min (₹6.72/min)
  • Credits: 664 (~83 min, ~41 calls)
  • TTS config: Org-level via /api/v1/organizations/model-configurations/v2
  • Org mode: dograh (managed)
  • Available voices: 892 Dograh voices, 31 ElevenLabs voices
  • ElevenLabs: Can use BYOK mode (needs non-Dograh LLM for full BYOK)

MoneyPrinterTurbo Details

  • Path: /root/MoneyPrinterTurbo
  • Systemd: moneyprinter.service (port 8081)
  • Nginx SSL: port 8937
  • Config: config.toml
  • LLM: OpenAI provider → Ollama Cloud (kimi-k2.7-code:cloud)
  • TTS: Edge TTS (free, no API key)
  • Footage: Pexels (free API key)
  • Subtitles: Edge timestamp mode
  • ffmpeg: System-installed (v6.1.1)

Files

  • /root/voice-agent-generator/main.py — FastAPI app + pipeline + video endpoints
  • /root/voice-agent-generator/api/crawler.py — Playwright crawler with SPA scrolling
  • /root/voice-agent-generator/api/crawler_adapter.py — sync wrapper for main.py
  • /root/voice-agent-generator/api/web_crawler.py — legacy requests crawler + build_knowledge_base
  • /root/voice-agent-generator/api/dograh_cloud.py — Dograh Cloud API integration
  • /root/voice-agent-generator/templates/index.html — landing page with URL normalization
  • /root/voice-agent-generator/templates/companion.html — split-view + transcript + promo video button
  • /root/voice-agent-generator/templates/creating-v2.html — animated progress bar template
  • /root/voice-agent-generator/templates/creating.html — original template (preserved)
  • /root/voice-agent-generator/data/jobs.json — Job state tracking
  • /root/.hermes/scripts/continuous_scraper.py — CDP-based background KB scraper
  • /root/.hermes/scripts/dograh_kb_client.py — Dograh KB API client
  • /root/MoneyPrinterTurbo/config.toml — Video generator config
  • /etc/systemd/system/voice-agent-generator.service — Systemd service
  • /etc/systemd/system/moneyprinter.service — MoneyPrinterTurbo systemd service
  • /etc/nginx/sites-available/voice-agent-generator — Nginx config (ports 8936/8937)

Backups

  • /root/voice-agent-generator-backup-2026-08-24.tar.gz — Full backup (main.py, templates, crawler, jobs.json)
  • /root/MoneyPrinterTurbo/config.toml.bak-2026-08-24 — MPT config backup
  • /root/nginx-voice-agent-generator.bak-2026-08-24 — Nginx config backup
  • [[Projects Dashboard]]
  • [[D-insights Voice Agent Platform]]