Row-Bot
Conversations, goals, memory, and scheduled workflows live together in one private AI assistant that runs on your own server and opens in any browser, on a laptop or a phone. Chat with local models through Ollama, provider API keys, ChatGPT, Claude, or Grok subscriptions, or any OpenAI-compatible endpoint, and every tool call shows up in the transcript. Risky actions pause for Approve or Deny in the chat, on the Home screen, or on a connected channel. Memory grows into a personal knowledge graph with recall, review, and an Obsidian-compatible wiki vault. Ask for a deck, a mockup, or an app and a design or code panel opens beside the chat, with Present, export to PDF, HTML, PNG, or PPTX, a changes view, Git, pull requests, checks, and an interactive terminal. Goals keep going until done, delegated agents get their own conversations, and workflows run on a schedule, a webhook, or a chain, with run history and delivery to channels. Telegram, WhatsApp, Discord, Slack, and SMS channels, voice dictation, and an app catalog for GitHub, Notion, Linear, and Stripe extend it, and health checks propose one fix per problem. Pairing a phone takes a QR code and a one-time invitation, and the server serves a single owner across many devices. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.
AMC WebUI
AMC WebUI is one chat window for every Gemini trick, with your conversations kept in your own browser instead of someone else's database. Pick a model from a picker that shows context window, output limit, and capabilities, then watch its reasoning stream with adjustable thinking levels. Voice chat and screen sharing run in realtime, and code blocks can turn into interactive HTML, SVG, Mermaid, or Graphviz previews. A Python sandbox executes snippets with numpy, pandas, and matplotlib inside the page. MCP settings list built-in tool servers beside any stdio, SSE, or HTTP server you add, and tool calls wait for your approval. Upload images, PDFs, audio, video, or whole ZIP folders, search past chats, and read per-request token and pricing logs. Deep search cites sources, text to speech offers thirty voices, and image generation takes aspect ratio and size options. Switch to any OpenAI-compatible endpoint with its own keys and models, choose among four themes and seven interface languages, and install it as a PWA. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.