Skip to main content

​
Monorepo Structure

Unpod is an NX monorepo with four main applications and a shared library layer. 
unpod/
├── apps/
│   ├── web/              # Next.js 16 frontend (React 19)
│   ├── backend-core/     # Django 5 REST API
│   ├── api-services/     # FastAPI microservices
│   ├── super/            # Voice AI engine (LiveKit + Pipecat)
│   └── unpod-tauri/      # Desktop app (Tauri 2)
├── libs/
│   └── nextjs/           # Shared React libraries (@unpod/*)
├── infrastructure/
│   └── docker/           # Dockerfiles & service configs
└── scripts/              # Setup, migration, and utility scripts

​
Core Services

​
Web Frontend (apps/web/)

Next.js 16 frontend with App Router, group-based layouts, styled-components, and Ant Design.
  • Key features: Auth, dashboard, AI Studio, spaces, knowledge bases, settings
  • Testing: Playwright for E2E tests
npx nx dev web              # Dev server at port 3000
npx nx build web            # Production build
npx nx e2e web              # Playwright E2E tests
Environment: copy apps/web/.env.local.example to apps/web/.env.local. Desktop app (Tauri): npm run desktop:dev / npm run desktop:build

​
Backend Core (apps/backend-core/)

Django 5 REST API with JWT auth, multi-tenant organizations, RBAC, and background tasks.
  • Auth: JWT authentication + Google OAuth
  • Storage: PostgreSQL (relational), MongoDB (documents), Redis (cache)
  • Endpoints: All under /api/v1/
cd apps/backend-core
source .venv/bin/activate
python manage.py runserver        # API at port 8000
pytest                            # Run tests

​
API Services (apps/api-services/)

FastAPI microservices for messaging, document store, AI search, and task management. MongoDB primary storage. 
cd apps/api-services
pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 9116 --reload
Interactive docs at http://localhost:9116/docs.
RouteServiceDescription
/api/v1/storestore_serviceDocument store & indexing
/api/v1/connectorstore_serviceData connectors
/api/v1/voicestore_serviceLiveKit voice/video
/api/v1/searchsearch_serviceAI-powered search
/api/v1/conversationmessaging_serviceChat conversations
/api/v1/agentmessaging_serviceAgent management
/api/v1/tasktask_serviceTask management
WebSocket: ws://localhost:9116/ws/v1/conversation/{thread_id}/

​
Voice AI Engine (apps/super/)

Voice AI engine built on LiveKit and Pipecat. Orchestrates real-time voice agents with LLM providers, TTS/STT engines, and workflow automation via Prefect. 
cd apps/super

# Install (uv recommended)
uv pip install -r requirements/super.txt -r requirements/super_services.txt

# Run voice executor
uv run super_services/orchestration/executors/voice_executor_v3.py start

# Run Prefect worker
uv run -m prefect worker start --pool call-work-pool

# Testing
pytest                      # All tests
pytest -m unit              # Unit tests only
pytest -m integration       # Integration tests

​
Desktop Application (apps/unpod-tauri/)

Native cross-platform desktop client built with Tauri 2 and Rust.

​
Tech Stack

LayerTechnology
FrontendNext.js 16 / React 19 / styled-components / Ant Design
MonorepoNX 22
DesktopTauri 2
BackendDjango 5 + DRF / FastAPI
Voice AILiveKit + Pipecat + LangChain
DatabasesPostgreSQL 16, MongoDB 7, Redis 7
MessagingKafka (KRaft), Centrifugo v5

​
Data Flow

How calls flow through the system: The Bridge routes incoming calls to the appropriate Agent, which processes speech through the AI pipeline (STT → LLM → TTS) and streams audio back to the caller via the configured Voice Provider.

​
Next Steps

Quickstart

Get Unpod running on your machine.

Core Components

Numbers, Providers, Bridges, and Agents explained.

Dev Platform

Configure telephony and agents via the dashboard.

API Reference

Full API documentation.