Player not loading? Watch on YouTube
Vane, formerly Perplexica, is presented as a research and answering engine that runs on your computer. The video demonstrates questions that produce answers with clickable numbered sources, then explains how to add documents and choose a model. Installation requires Docker, Ollama and terminal commands; the speaker does not present it as a desktop app you simply download and open.
The walkthrough compares Speed, Balanced and Quality modes using a question about charging a phone overnight. In this example, they take four, seven and 32 steps respectively. These are demonstration results, not fixed counts. The speaker describes Quality as slower and more demanding on the CPU.
Vane can use attached PDFs, Markdown and text files alongside web research. Its Web, Academic and Social filters can be combined. According to the speaker, turning all three off lets it answer from the model and attached files without searching the internet. That is the offline workflow described here; live web research still requires a connection.
The model section shows Qwen through Ollama and discusses GPT, Gemini and Claude API options. Hardware needs depend on the model, with larger models described as more demanding. Comparisons with Perplexity and NotebookLM focus on citations and source selection. Vane lacks the dedicated single-source control shown in NotebookLM, though the speaker demonstrates requesting one source in a prompt. The closing assessment calls Vane slower and less capable than paid services.
