Libra: a client for generating local AI intelligences with dialog (internal test)

General Introduction

Libra is an innovative tool from Greenbit.ai whose core function is to generate AI intelligences that can run locally through natural language conversations. Called the "Vibe Agent", it allows users to quickly create proprietary intelligences that can perform tasks such as web search, data analysis, visualization reports, and more by describing their needs in simple words. Libra uses low-bit quantization technology to compress large models to run on ordinary computers (e.g. Mac), reducing memory footprint by more than 751 TP3T and performance loss by less than 11 TP3T. It also enables intelligences to work efficiently even with limited resources through adaptive context management and responsive scheduling.

Function List

• Dialog-generating IntelligentsiaLibra automatically generates AI intelligences that are capable of self-planning.
• Local runtime support: Compress large models down to 3/4 bit to run smoothly on consumer-grade hardware.
• Web Search: Intelligent bodies can automatically search the web and extract useful information.
• data analysis: Processing data, calculating statistical indicators and generating results.
• Visualization reports: Automatically create charts and graphs to visualize the results of your analysis.
• context management (computing): Prioritize retention of important information to improve comprehension in long conversations.
• Tool Integration: Connect to external tool chains to support collaboration on complex tasks.

Using Help

Registration and Application

1. Open the official website https://www.greenbit.ai/, find the "Apply for Experience Libra" button and click on it.
2. Fill in your name, email address and other information to submit your request. The team will contact you within 1-2 business days to provide access.
3. After receiving the confirmation email, follow the instructions to download the Libra client (currently supported on Mac systems).

Installation process

• Download & Setup
  Once the download is complete, double-click the installation package and follow the prompts to complete the installation. The installation process takes about 5 minutes and takes up less than 1GB of space. after installation, open the Libra client and enter your registered e-mail address to log in.
• hardware requirement
  Libra runs on Apple's M-series chips (e.g. M1, M2, M3) and recommends at least 16GB of RAM. No additional configuration is required and it is ready to use after installation.

Creating the first intelligence

• input requirement
  After logging in, enter the main interface and click "New Intelligence". In the dialog box, enter a requirement, such as: "Help me analyze the trend of EV sales over the past 5 years and generate a chart." Libra will automatically parse this sentence.
• generation process
  After clicking "Generate", Libra uses AI to create an intelligence. This process usually takes 10-30 seconds. After generating, the interface will show the work plan of the intelligent body, such as searching data, cleaning data, and generating charts.
• Operation and adjustment
  Click "Run" and the smart body starts working. You can see it searching the web, processing data and outputting results. If you're not satisfied with the results, you can continue the conversation and make adjustments, for example, "Add a forecast for the next three years." The intelligence will replan and execute.

Featured Function Operation

• local operation
  Libra processes tasks locally by default, without the need for a cloud API, and thanks to low-bit quantization, a large model with 7 billion parameters takes up a few gigabytes of memory, making it a breeze to run on Mac Studio. You can check memory usage in Settings.
• Web Search
  The smart body can open a browser and search for relevant information. For example, if you type in "find the latest smartphone reviews", it will automatically visit tech websites and extract key content.
• Data Analysis and Visualization
  If you upload an Excel file and say, "Calculate the annual growth rate and draw a line graph." The smart body will process the data in Python, generate the chart, and the results will be displayed directly in the interface.
• Export and Deployment
  When you're satisfied, click Export and select Deploy Locally, and Libra will generate a standalone executable file for use on other Macs.

Viewing Operational Status

• On the "Task Management" page, you can see the real-time status of the smart body, such as whether it is currently searching or analyzing data. Each task has a progress bar, and the result will be saved automatically after completion.

Getting Help

• If you have a problem, click the "Contact" button at the bottom of the interface and fill in a description of the problem. The team will respond within the next business day.
• The official website also provides a video tutorial, at the path https://www.greenbit.ai/contact, which shows the complete process of generating an intelligent body.

caveat

• When using it for the first time, it is recommended to test it with simple requirements, such as "search for today's weather". Try complex tasks after you are familiar with them.
• A stable power supply is needed for local operation and to avoid hibernation of the computer when handling large tasks.
• If you need cloud support, you can turn on "End-to-End Cloud Collaboration" in "Settings", but this will add a small amount of time. Token Costs.

With these steps, you can quickly get started with Libra, whose conversational operation makes it easy for non-technical users to create intelligences that really "speak and work.

application scenario

1. Market Research
   Analysts can use Libra to generate an intelligence, type in "analyze EV market trends," and get a report with sales data and charts in 10 minutes, eliminating the need for manual searches and programming.
2. content creation
   Digital bloggers can say, "Filter the latest cell phone information and write a microblog", and the intelligent body will automatically search for information and generate content to enhance creative efficiency.
3. Personal Assistant
   Users can enter "Schedule today's tasks according to my schedule" to generate a local assistant to help you plan your time and remind important things.

QA

1. Is Libra free?
   Currently you need to apply for a trial, no specific pricing has been announced on the official website. There may be a free trial period after application, and the official version may be charged on a subscription basis.
2. What devices are supported?
   It supports Apple Mac series (M1/M2/M3 chips) at this stage and may be extended to other platforms in the future.
3. What's better than cloud-based AI?
   Libra runs locally, at low cost (Token fees down to 90%), with more guaranteed data privacy and no need to be connected all the time.