About this project
GobboNet is a Windows-based AI chatbot front end designed for complete local privacy — no accounts, no monthly fees, and no data sent to the internet after initial setup. It provides a one-click installer (GobboNetSetup.exe) that bundles a launcher, downloads the llama.cpp AI engine (~300 MB), helps users select and download an AI model (.gguf files, typically 4–16 GB) based on detected hardware, and fetches a small retrieval model (~146 MB) for lorebook/RAG functionality. Once configured, everything runs entirely offline.
Supported working models include Llama, Gemma, DeepSeek, Qwen, gpt-oss, Mistral, Granite, Command R, and GLM. The chat interface opens in a web browser and supports streaming responses, built-in and custom characters (including .png character card import/export), lorebooks with retrieval-based context injection, conversation threads with folder organization, optional web search via Ollama API key, model switching from a dropdown, and connection to external llama.cpp or Ollama servers. Users can also add .gguf model files manually.
Optional LAN setup enables phone/tablet access on the same Wi-Fi network, protected by a user-set password (stored in scrambled form). The tool includes a hardware probe that recommends models fitting available VRAM with at least 2 GB spare. Troubleshooting covers common errors (500, 502, 404), blank chat screens, password resets, and connectivity issues. Known bugs include broken logit bias (banned words feature). Planned features include PDF handling, Mac/Linux/Android ports, group chat, TTS/STT, and a canvas side panel.
Comments
0 Rating appears after 10 ratings
Sign in to join the discussion.