Framework to build video agents that can reason through complex video tasks like search, editing, compilation, generation etc & instantly stream the results.
⭐️ Built on top of the cutting edge 'Video-as-Data' infrastructure, VideoDB
⚡️Watch Agent Demos
✨Try Hosted Version
📖 Documentation
👩💻New Agent Request
Think of Director as ChatGPT for videos. It is a framework to build video agents that can reason through complex video tasks like search, editing, compilation, generation etc & instantly stream the results.
For example, a simple natural language command like: Upload this video and send the highlights to my Slack
, sets everything in motion - Director’s reasoning will orchestrate the different agents intelligently to complete the task for you.
Built on top of VideoDB’s ‘video-as-data’ infrastructure, Director enables you to:
- Summarize videos in seconds.
- Search for specific moments.
- Create clips instantly.
- Integrate top GenAI projects and APIs and create and edit content instantly.
- Add overlays, generate thumbnails, and much more.
Built with flexibility in mind, Director is perfect for developers, creators, and teams looking to harness AI to simplify media workflows and unlock new possibilities. 📺 Watch: Intro video
Demos.Compilation-Compressed.mp4
- Summarize videos in seconds.
- Generate full movies with voiceovers from a script.
- Search and index your media library.
- Organize and clip your content effortlessly.
- Dub and edit your audio and video with ease.
- Translate and add subtitle in any languge.
- ....and a whole lot more >>
Experience a sleek, chat-based interface with built-in video playback and intuitive controls. It’s like having a personal assistant for your media.
Connect seamlessly with powerful AI tools like LLMs, databases, and GenAI APIs, while VideoDB ensures your video infrastructure is reliable and scalable for cloud storage, indexing and streaming your content effortlessly.
Easily add new agents and tools to your workflow. Whether you want to run it locally or on your cloud, The Director adapts to your needs.
Director's architecture brings together:
-
Backend Reasoning Engine: Handles workflows and decision-making. Checkout the backend folder in director codebase.
-
Chat-Based UI: Engage with your media library conversationally. Check videodb-chat for the source code.
-
Video Player: Advanced playback and interaction tools. Check videodb-player for the details about the multi platform video player.
-
Collection View: Organize and browse your media effortlessly.
At the heart of The Director is its Reasoning Engine, a powerful core that drives intelligent decision-making and dynamic workflows. It acts as the brain behind the agents, enabling them to process commands, interact with data, and deliver meaningful outputs.
- Contextual Understanding: The engine analyzes user inputs and maintains context, ensuring smooth and coherent interactions with agents.
- Dynamic Agent Orchestration: Based on the user’s needs, it identifies and activates the right agents to complete tasks efficiently.
- Modular Processing: Tasks are broken into smaller steps, allowing agents to collaborate and deliver accurate results in real time.
- Multi-Agent Coordination: Seamlessly integrates multiple agents to handle complex workflows, such as summarizing, editing, and searching videos.
- Real-Time Updates: Provides live progress and feedback as tasks are being completed.
- Extensible Design: Easily adaptable to include custom logic or connect to external APIs for more advanced capabilities.
The Reasoning Engine works in tandem with the chat-based UI, making video interaction intuitive and efficient. For example:
- Input: "Create a clip of the funniest scene in this video and share it on Slack."
- Output: The engine orchestrates upload, scene detection, clipping, and sharing agents to deliver results seamlessly. Watch the video here
For a closer look, check out the detailed architecture diagram below:
- Python 3.9 or higher
- Node.js 22.8.0 or higher
- npm
1. Clone the repository:
git clone https://github.com/video-db/Director.git
cd Director
2. Run the setup script:
./setup.sh
This script will:
- Install Node.js 22.8.0 using nvm
- Install Python and pip
- Set up virtual environments for both frontend and backend.
3. Configure the environment variables:
Edit the .env
files to add your API keys and other configuration options.
- Mac
- Linux
- Windows (WSL)
To start both the backend and frontend servers:
make run
-
Backend:
http://127.0.0.1:8000
-
Frontend:
http://127.0.0.1:8080
For specific tasks:
-
Backend only:
make run-be
-
Frontend only:
make run-fe
Checkout hosted documentation at https://docs.director.videodb.io
To create a new agent in Director, follow these steps:
-
Copy the template: Duplicate
sample_agent.py
inDirector/backend/director/agents/
and rename it. -
Update class details:
- Rename the class.
- Update
agent_name
anddescription
-
Implement logic:
- Update parameters and
docstring
- Implement your agent's logic
- Update the run() method.
- Update parameters and
-
Handle output and status updates:
- Use appropriate content types (TextContent, VideoContent, ImageContent, SearchResultContent)
- Update
self.output_message.actions
for progress indicators - Use
push_update()
to emit progress events - Set content status (progress, success, error) and messages
-
Implement error handling:
- Set error status and messages if issues occur
-
Finalize the response:
- Call
self.output_message.publish()
to emit final state and persist session - Return an
AgentResponse
with result, message, and data
- Call
-
Register the agent:
- Import your new agent class in
Director/backend/director/handler.py
- Add it to the
self.agents
list inChatHandler
- Import your new agent class in
Remember to consider creating reusable tools if your agent's functionality could be shared across multiple agents.
Checkout hosted documentation at https://docs.director.videodb.io
To serve the documentation on port 9000:
source backend/venv/bin/activate
make install-be
mkdocs serve -a localhost:9000
To build the documentation:
mkdocs build
We welcome integrations from projects that can make video workflows easy and increase capabilities of the projects. Please check issues and discussions for details.
Any contributions you make are greatly appreciated. Here's the process:
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature
) - Commit your Changes (
git commit -m 'Add some AmazingFeature'
) - Push to the Branch (
git push origin feature/AmazingFeature
) - Open a Pull Request