Deepseek-V3 Computer Use - This FULLY FREE AI Agent is INSANE & CAN DO ANYTHING!
AI Summary
Video Summary
- The video introduces Deep Seek V3, a cost-effective tool comparable to Sonet.
- A GitHub repository is mentioned, providing a web UI for Browser Use, making it easy to control web browsers.
- The web UI supports various providers and is simple to configure.
- The Browser Use library is the foundation of this web UI, offering a browser interface to control Browser Use agents.
- The web UI has expanded LLM support, including compatibility with Gemini, OpenAI, Azure, OpenAI Anthropic, Deep Seek, and Al Llama.
- Custom browser support is available, allowing users to connect to their own browsers for tasks that require login.
- The system prompts have been improved for better task performance.
Detailed Instructions
- Clone the GitHub repo to your computer.
- Install the Browser Use Python package with
pip install browseruse
.- Install Playwright and related components with
playwright install
.- Install required packages with
pip install -r requirements.txt
.- Create an environment variables file from the example provided, adding necessary API keys.
- Run the server with a specified command to start the app on localhost.
- Use the web UI to select agent type, set maximum run steps, and choose whether to use vision.
- Select the LLM provider and enter the model name.
- Modify the LLM base URL and API key directly in the web UI if not set in the environment file.
- Adjust browser settings, including using your own browser, running in headless mode, and setting window dimensions.
- The web UI records the browser session for later viewing, with an option to change the recording path.
- Enter the task for the agent to perform, along with additional info, and execute it.
Usage Examples
- The video demonstrates using the web UI to search for “AI code King” using Deep Seek.
- It also shows how to check the stock price of Nvidia using the Gemini 2.0 flash model.
Conclusion
- The Browser Use web UI simplifies the process of controlling browser agents and is compatible with various models.
- The video encourages viewers to share their thoughts and subscribe to the channel.