About this project

FlexiGPT is a desktop application that provides a provider-independent interface for interacting with multiple LLM providers. It emphasizes user control and privacy by storing conversation history, configurations, and workspace context locally using JSON and SQLite. Key capabilities include: - **Multi-Provider Support**: Built-in presets for OpenAI, Anthropic, Google Gemini, xAI, Mistral, Hugging Face, and OpenRouter, as well as local runtimes like Ollama, LocalAI, and vLLM. - **Repeatable Workflows**: Users can create assistant presets that bundle specific models, tools, skills, and starting prompts to standardize recurring tasks. - **Agentic Controls**: Supports tools and MCP servers with configurable execution modes, allowing users to choose between automatic execution or manual human-in-the-loop review. - **Rich Rendering**: Integrated support for Markdown, KaTeX math rendering, and Mermaid diagrams. - **Specialized Workflows**: Includes pre-built assistant presets for software development (feature implementation, code review, bug investigation) and professional writing (PRDs, technical documentation). - **Local-First Privacy**: Requests are sent directly to the configured provider without proxying through FlexiGPT servers, and API keys are stored securely via the OS keyring.