h2oai/h2ogpt

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

11969 stars · 1303 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

h2oGPT is an Apache-2.0 open-source application for chatting with local, private LLMs and querying/summarizing your own documents (PDFs, Word, images, video, audio, code) with RAG. It provides a Gradio web UI and CLI, supports many models via HF, llama.cpp, GPT4All, and AutoGPTQ, and includes vision, image generation, STT/TTS, and authentication features.

Use cases

• chat with local private gpt offline
• ask questions about my pdf documents
• summarize documents with an llm
• run a private chatgpt alternative on my own machine
• build a document question answering chatbot
• self-host a private ai assistant with voice
• index and search my files with embeddings

When to choose

• you need 100% private, offline LLM chat and document Q&A
• you want a self-hosted RAG pipeline with a ready-made UI
• you need support for many model backends (llama.cpp, HF, GPT4All, AutoGPTQ) on CPU or GPU
• you want built-in OCR, vision, image generation, and speech features in one app

When to avoid

• you only need a lightweight library to embed LLM calls in your own app
• you want a fully managed hosted service rather than self-hosting
• you need a minimal single-purpose tool rather than a feature-rich platform
• you lack the hardware for local inference and prefer cloud APIs

Facets

application · maturity active

rag llm-inference chatbot chat-interface ocr speech-recognition tts image-processing search-engine vector-database auth cli ui-components large-language-models chatbots artificial-intelligence privacy self-hosted pdf python windows self-hosted cli privategpt document-qa local-llm llama-cpp gradio-ui embeddings vector-store openai-proxy voice-assistant stable-diffusion offline-ai retrieval-augmented-generation natural-language-processing linux macos docker gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="h2oai/h2ogpt")