Project Name: FreeAskInternet - Localized Dialogue System for Large Language Models
Github Open Source Address: https://github.com/nashsu/FreeAskInternet
FreeAskInternet is a free and open-source tool for large language models. It combines multiple advanced large language models to provide search and answer generation services.
We can install and run it directly on our local machine without the need for a GPU. It looks more like an aggregator of various large language models.
According to the project description, FreeAskInternet uses Large Language Models (LLM) to generate answers without relying on Graphics Processing Units (GPUs). After users ask questions, the system uses SearXNG for multi-engine search and combines the search results with the ChatGPT3.5 language model to generate answers based on these search results. All processes run locally and do not require a GPU or API keys from OpenAI or Google.
Note: SearXNG is a free internet metasearch engine that aggregates results from various search services and databases. Open source address: https://github.com/searxng/searxng.
In summary, the features of FreeAskInternet are:
- Free and open-source, we can install it locally ourselves.
- It uses APIs such as ChatGPT3.5, Qwen, Kimi, and ZhipuAI, allowing users to use them without needing API keys.
- Due to the use of free APIs, unlike large language models that usually require GPUs to run, FreeAskInternet can run without a GPU, reducing the hardware requirements for users.
- We can also customize large language models, such as ollama (download address: https://ollama.com/), which provides users with more flexibility and personalized options.
- Easy installation, using Docker Compose, a popular container orchestration tool, users can quickly deploy FreeAskInternet, simplifying the installation and configuration process.
Working Principle of FreeAskInternet: When users input questions in the interface, SearXNG performs multi-engine search. After the search is completed, the system automatically crawls and collects the content of the links in the search results. Then, this content is passed to the selected LLM (Large Language Models) such as ChatGPT3.5, Kimi, Qwen, ZhipuAI, or custom ollama. These models use the passed content as a reference to generate content and answer the user's questions.
Installation and Usage:
This project is still in its early stages, so there may be some issues, but it is completely sufficient for personal learning and usage.
Although the project is relatively new, it has already received 6.8k stars and its popularity is still increasing. It is worth paying more attention to.
Before installation, you need to have Docker installed on your local machine and support docker-compose. Once installed, you can open the link http://localhost:3000 to start using it.
Installation and usage are as follows:
git clone https://github.com/nashsu/FreeAskInternet.git
cd ./FreeAskInternet
docker-compose up -d