arc53/docsgpt

DocsGPT is an open-source AI platform for building chatbots and search tools that work with your own documents and data sources. Rather than querying a general-purpose AI, you connect it to your company's files, and it answers questions using those specific documents, with citations back to the source, to reduce the chance of the AI inventing facts. The platform supports a wide range of document formats: PDFs, Word documents, spreadsheets, presentations, Markdown, HTML, JSON, audio files (which it transcribes), and web content fetched from URLs, sitemaps, Reddit, or GitHub. You can ingest all of these into a shared knowledge base and then deploy a chat interface that searches and reasons over that content. Beyond simple question-answering, DocsGPT includes an agent builder with a visual workflow editor that lets you create more complex automated processes, for instance, looking up information from a document, calling an external API, and returning a combined answer. It supports multiple AI model providers (OpenAI, Google, Anthropic) as well as locally-run models, giving you control over privacy and cost. Pre-built integrations include embeddable chat widgets for websites, Discord and Telegram bots, and connectors for services like SharePoint and Confluence. The tech stack uses Python and Flask for the backend and a JavaScript-based frontend. Deployment is handled via Docker. You would use DocsGPT if you are a developer or business that wants to build a private, document-aware AI assistant or internal search tool without sending sensitive data to third-party AI providers.