At a Glance:
VT is a privacy-first, local-storage AI chat application that runs in the browser with zero server-side conversation storage, supports loading models locally through Ollama and LM Studio, and provides access to premium models for users who bring their own API keys.
Overview:
VT is a browser-based AI chat application that stores all conversation data in the browser’s IndexedDB and never sends chat history to a server. It consolidates access to multiple AI providers, including OpenAI, Anthropic, Google, Fireworks, Together AI, and xAI, while also allowing users to run models locally through Ollama or LM Studio. The application provides a set of built-in free tools such as document processing, chart visualization, web search, and a semantic tool router. A conversational image editor called Nano Banana is included for iterative image generation and editing. VT is deployed on Fly.io using Next.js 15, React 19, and a Turborepo-managed monorepo.
Key Decision Points:
Local-first chat storage: All conversations are stored in the browser using IndexedDB, and the application does not retain chat data on the server.
Model access model: Nine server-side models are available for free without an API key, while premium models require users to supply their own API keys.
Local model support: Users can run AI models entirely on their own machine through integrations with Ollama and LM Studio.
Authentication requirement: Login is required to access the application, and authentication is managed through Better Auth.
Deployment model: The application is a hosted web service; the README does not describe a self-hosted deployment or publishing an official Docker image for self-hosting.
Core Features:
Nano Banana conversational image editor: Generates images and applies iterative edits through natural language instructions such as resizing objects or changing backgrounds.
Multi-provider AI support: Connects to OpenAI, Anthropic, Google, Fireworks, Together AI, xAI, and local runtimes like Ollama and LM Studio.
Local-first storage via Dexie.js: Persists all chat data in the browser’s IndexedDB without transmitting conversations to a server.
Semantic tool router: Uses OpenAI embeddings and pattern matching to activate the appropriate built-in tool based on the user’s query.
Free built-in tools: Includes document processing for PDF, DOC, DOCX, TXT, and MD files, chart visualization, web search, thinking mode, and a mathematical calculator.
Structured output extraction: Extracts JSON data from uploaded documents using AI processing.
Use Cases:
Developers who want to use premium AI models without a monthly subscription by bringing their own API keys for models such as Claude Sonnet and GPT-4.1.
Privacy-conscious users who need an AI chat interface that does not store conversation history on external servers.
Users who want to run AI models locally through Ollama or LM Studio and switch between local and cloud-based models in one interface.
Open-Source Alternative Value:
VT provides a browser-based AI chat client with local-first storage, meaning conversation data remains on the user’s device rather than being stored on a server. It supports local model execution through Ollama and LM Studio, which can reduce reliance on cloud-based AI services for users who prefer or need to run models on their own hardware. The option to use premium models with a personal API key allows paid-tier access without linking usage to a platform-specific subscription. These privacy and access characteristics distinguish it from AI chat interfaces that require server-stored history or bundled subscription plans.
