Transcription
And again, I am glad to welcome esteemed subscribers and guests of the channel. Today's topic will be especially interesting for those who have already heard about Perplexity, a popular search engine that has largely changed the approach to finding information on the internet. Perplexity AI is a smart search engine that works based on large language models. It doesn't just provide links, but forms connected answers, cites sources, and allows for dialogue, clarifying queries. Its main strength is that it saves time. Instead of dozens of open tabs, we immediately get a ready explanation or a selection of materials. In addition, Perplexity can handle different types of queries: from searching for scientific articles to analyzing news and social networks. But like any commercial technology, Perplexity has its limitations and disadvantages. Firstly, it is a closed commercial platform, and access to advanced features costs from 20 to 200 dollars per month. We should not forget that it is a cloud service, and your queries and data always pass through external servers, which can both read them and impose censorship. For those who value privacy, this can be a critical drawback. And finally, the search itself is limited by Perplexity's built-in infrastructure. We cannot change models, manage configuration, or run the system locally. And this is where its open alternative, Perplexica, comes onto the scene. And today we will analyze in detail what this tool is, how it works, how it differs from the original, and most importantly, how it can be installed and used in practice. The Perplexica project began back in 2024 and has already gathered over 22,000 stars on GitHub. Unlike the commercial version, Perplexica is a completely free solution and is distributed under an open-source MIT license. This also means that we can deploy the system on our own computer or server, fully control our data, and even adapt the architecture to our own tasks. Perplexica is built on a modular architecture and uses several key components. Firstly, there is, of course, the interface – it is a modern web application on NextJS with support for dialog mode and different types of search. When we pose our query, agents come into play. This is an intelligent layer that decides how exactly to process queries. There are separate agents for searching scientific publications, YouTube, Reddit, and other social networks. And a separate AI also works for text generation. Metasearch in Perplexica is carried out through CRXNG. You may be familiar with this name, as in previous videos I talked about CRX. I highly recommend watching that video at least for general development. So, CRXNG is a continuation of the CRX project. A self-hosted search engine that immediately searches across multiple sources, providing broader coverage than traditional search. Within Perplexica, support for different AI providers is also implemented: OpenAI, Anthropic, Grok, Gemini, Deepseek, and, importantly, local models through Ollama. That is, you decide for yourself whether to use expensive cloud APIs or local artificial intelligence models on your own computer. In addition to all the listed modules, Perplexica implements semantic search. These are built-in Bed1 models that help sort CRXNG results by semantic proximity. As you can see, Perplexica is a rather complex system consisting of many components. However, it is very easy to implement it on your own computer, which we will now see. The easiest way to install Perplex is through Docker. First, install Docker and Docker Compose. Instructions for this process are widely available on the project page for all supported operating systems, so there should be no difficulties. Next, clone the Perplexica repository to our computer. After that, navigate to the project folder and copy the example configuration, like config.toml. In the created file, enter your API keys or parameters, depending on which artificial intelligence model we want to use. As the simplest example, I will enter my API key from OpenAI. Well, and finally, all that remains is to launch Perplexica with the corresponding command. After that, Perplexica will be available in the browser at localhost on port 3000. We can go there by opening a browser. And, as you can see on your screens, everything works wonderfully. What advantages and disadvantages does Perplexica give us? We have full control over data and privacy, especially if we use local models through Ollama, which I talked about in my previous videos. We don't need any subscriptions, and we are not dependent on any services, only our own resources. But at the same time, not every user can handle Docker or manual configuration, and performance depends on the chosen models and the power of your hardware. Nevertheless, if you want to take your place in the new world, are interested in earning with AI, or simply want automation, then Perplexica is not just an alternative to Perplexity, but a tool to gain more opportunities. All links to the project and instructions can be found in the description under the video. And thank you, dear friends, for watching this video to the end. [music]