This project is currently under active development. You might encounter some issues during setup or usage. We appreciate your understanding and welcome any feedback to improve the experience.
Explore our plans and upcoming features for Agent Studio!
The full roadmap is available here.
Alternatively, you can view it directly on Vimeo.
Agent Studio is an AI agent application designed to handle real-time interactions through phone calls, web-based voice user interfaces (VUI), and SIP capabilities. This open-source platform allows you to create advanced AI-driven communication systems. If you need an on-premise solution, customization, or a secure app setup without third-party APIs, please contact us at [email protected]. We are available to assist with setup and provide tailored solutions to meet your requirements.
Welcome to the Agent Studio installation guide! This document will guide you through setting up the backend server, SIP server, and the user interface (UI). The Agent Studio leverages powerful technologies like Groq, a highly optimized LLM processing system, and LlamaIndex for retrieval-augmented generation (RAG) to provide real-time communication and data management.
The code repository for Agent Studio is available on GitHub: NidumAI-Inc/agent-studio.
- Prerequisites
- Step 1: Set Up Agent Studio Backend
- Step 2: LiveKit On-Premise Installation
- Step 3: SIP Server On-Premise Installation
- Step 4: Set Up the User Interface (UI)
- Step 5: Run Agent Studio and UI
- Step 6: Start Services in
tmuxSessions - Credits
- Support
Before you begin, ensure you have the following installed on your system:
- Python 3.8+
- MongoDB (or access to a MongoDB instance)
- Node.js and npm (for the UI)
- Git
- tmux (for managing multiple terminal sessions)
git clone https://github.com/NidumAI-Inc/agent-studio.git
cd aistudioIt’s recommended to use a virtual environment to manage dependencies.
python3 -m venv venv
source venv/bin/activate # For Linux and macOS
# For Windows:
# venv\Scripts\activatepip install --upgrade pip
pip install -r requirements.txtCreate a .env file in the app/env/ directory:
mkdir -p app/env
touch app/env/.envAdd the following content to app/env/.env:
# MongoDB settings
MONGO_USER=your_mongo_user
MONGO_PASSWORD=your_mongo_password
MONGO_HOST=your_mongo_host
# API Keys
OPENAI_API_KEY=your_openai_api_key
LIVEKIT_API_KEY=your_livekit_api_key
LIVEKIT_API_SECRET=your_livekit_api_secret
DEEPGRAM_API_KEY=your_deepgram_api_key
# LiveKit URL
LIVEKIT_URL=ws://localhost:7880Note: Replace placeholders (e.g., your_mongo_user, your_openai_api_key) with actual credentials.
pip install -e .For setting up the LiveKit server and enabling SIP voice communication features, refer to the guides below. These guides will walk you through downloading, configuring, and running LiveKit and the SIP server for real-time communication:
This section will guide you in setting up the Nidum Voice Agent and Nidum Bot projects. These two UI components are interlinked and must be configured correctly.
Clone both repositories to your local machine:
# Clone Nidum Voice Agent repository
git clone https://github.com/NidumAI-Inc/agent-studio-ui.git
cd agent-studio-ui # Nidum Voice Agent directory
# Clone Nidum VUI (Nidum Bot) repository
git clone https://github.com/NidumAI-Inc/agent-studio-vui-widget.git
cd agent-studio-vui-widget # Nidum Bot directoryInstall the dependencies for both projects separately.
Nidum Voice Agent:
cd aistudio_ui
npm install
# or
yarn install
# or
pnpm installNidum Bot:
cd aistudio_vui_widget
npm install
# or
yarn install
# or
pnpm installBoth projects require environment variables for configuration. Set up the environment variables for each project before running them.
Create a .env file in the aistudio_ui directory and add the following variables:
MONGODB_URI=your_mongodb_uri
NEXTAUTH_SECRET=your_nextauth_secret
LIVEKIT_API_KEY=your_livekit_api_key
LIVEKIT_API_SECRET=your_livekit_api_secret
NEXT_PUBLIC_LIVEKIT_URL=ws://localhost:7880
GMAIL_ID=your_gmail_id
GMAIL_PASS=your_gmail_password
NEXT_PUBLIC_ML_BACKEND_URL=http://localhost:8000
ALLOWED_BOT_ORIGINS=[]
NEXT_PUBLIC_BOT_LIVE_URL=http://localhost:5000 # Adjust the port if necessaryNote: Refer to the .env.example file in the repository for details on each variable.
Create a .env file in the aistudio_vui_widget directory and add the following variables:
VITE_API_LIVEKIT_URL=ws://localhost:7880
VITE_API_NEXT_backend=http://localhost:3000 # Nidum Voice Agent backend URL
VITE_API_ML_Backend=http://localhost:8000 # Agent Studio backend URLNote: Check the .env.example file in the repository for more details.
Navigate to the aistudio_ui directory and run:
npm run dev
# or
yarn dev
# or
pnpm devOpen http://localhost:3000 in your browser to access the Nidum Voice Agent.
Navigate to the aistudio_vui_widget directory and run:
npm run build
# or
yarn build
# or
pnpm build
# Install serve globally if not already installed
npm install -g serve
# or
yarn global add serve
# or
pnpm add -g serve
# Serve the build
serve -s distNow, Nidum Bot will be accessible and can be integrated with the Nidum Voice Agent.
Both projects are interdependent and must be configured correctly to interact:
- Nidum Voice Agent relies on Nidum Bot for bot communication capabilities. Ensure that
NEXT_PUBLIC_BOT_LIVE_URLin Nidum Voice Agent's.envfile points to the correct URL of the Nidum Bot (e.g.,http://localhost:5000). - Nidum Bot needs to communicate with the Nidum Voice Agent backend. Ensure that the
VITE_API_NEXT_backendenvironment variable is correctly set to point to the Nidum Voice Agent’s backend URL (http://localhost:3000).
In one terminal window, activate your virtual environment and run:
aistudio start api
In another terminal window, activate your virtual environment and run:
aistudio start agentThe vision-agent feature provides additional capabilities for visual data processing, though it currently has limited functionality. Start it in a new terminal window:
aistudio start vision-agentNote: The vision-agent is in its initial stages and currently offers limited functionality. Further updates will expand its capabilities.
- Agent Studio Backend API Docs: http://localhost:8000/docs
- Nidum Voice Agent UI: http://localhost:3000
- Nidum Bot UI: Accessible through the Nidum Voice Agent interface.
For continuous logging and management, we’ll start the following scripts in separate tmux sessions:
- Chat Log Creation Service
- SIP Log Creation Service
- Agent Studio App
To do this, follow the instructions below.
Start chat_log_creation.py in a new tmux session:
tmux new-session -d -s chat_log "python /app/log_service/chat_log_creation.py"This will create a detached tmux session named chat_log running chat_log_creation.py. You can attach to this session using:
tmux attach -t chat_logStart sip_log_creation.py in a new tmux session:
tmux new-session -d -s sip_log "python /Users/kesavan/aistudio/app/log_service/sip_log_creation.py"This will create a detached tmux session named sip_log running sip_log_creation.py. You can attach to this session using:
tmux attach -t sip_logTo start the main aistudio app in a tmux session, replace the incorrect command with:
tmux new-session -d -s aistudio "aistudio start api"This will create a detached tmux session named aistudio. You can attach to this session using:
tmux attach -t aistudioTo view or manage any of these sessions, list them using:
tmux list-sessionsAgent Studio integrates several powerful technologies:
- Groq: The fastest large language model (LLM) processing system, offering unparalleled speed for AI model deployment.
- LlamaIndex: Optimized for retrieval-augmented generation (RAG), LlamaIndex enhances the application’s ability to access and generate relevant, contextually informed responses.
- LiveKit: For real-time communication capabilities.
For inquiries or support, please contact us at [email protected].
Thank you for trying out Agent Studio! Your feedback is invaluable in helping us improve the platform.