An open AI platform, in the making
A personal R&D project built alongside full-time work: RAG, real-time voice agents, multi-LLM routing and workflow automation on one event-driven core. Open-sourcing soon.
Project status — pre-release
- ✓ RAG pipeline · OCR · embeddings · Qdrant
- ✓ Real-time voice agents · TTS · dubbing
- ✓ Multi-LLM routing & agent fleet
- ◻ Public open-source release
- ◻ Plugin SDK & self-host guide
A look inside the platform
A few of the consoles it ships with so far.
One console for every AI service
Usage, recent jobs, quotas and live system status across RAG, voice, dubbing and agents — in a single pane of glass.
- Real-time job & queue status
- Per-tenant usage & quota tracking
- Service health at a glance
Live agents with Claude assist
Low-latency voice with streaming transcription, suggested replies, and live signals — so every call is handled with context.
- Streaming transcription & diarization
- Claude-powered agent assist
- Sentiment & purchase-intent signals
Dub video into any language
Translate and re-voice content into multiple languages with lip-sync, preserved speaker tone and background audio.
- Lip-synced, mouth-matched output
- Cloned-voice multilingual dubbing
- Per-language progress tracking
Visual workflow automation
Drag-and-drop pipelines that orchestrate AI steps, approvals and integrations over replayable, event-driven jobs.
- Extract → validate → route → notify
- Human-in-the-loop approvals
- Replayable, event-driven runs
What's inside
Eight capabilities, one event-driven core.
Building this in the open
A side project, developed alongside full-time work — open-sourcing soon. Star the repo to follow along.
☆ Watch the repo soon