BlockBeats News, April 9, Nous Research, an open-source AI model agency, released the cross-platform AI Agent framework Hermes Agent v0.8.0, with the GitHub repository now having 37.9k stars. This framework can connect AI Agents to multiple instant messaging platforms such as Telegram, Discord, Slack, WhatsApp, Lark, and WeCom, supporting both CLI and Gateway operation modes. Key new features in this update:
· Background Process Auto Notification: Background tasks can now automatically notify the agent upon completion. After starting a time-consuming process, the agent will receive a notification upon completion without the need for polling. The agent can continue to handle other tasks and directly obtain the results when available.
· Nous Portal Free Support for Xiaomi MiMo v2 Pro: Nous Portal now supports the free version of the Xiaomi MiMo v2 Pro model for handling auxiliary tasks (compression, vision, summarization) and has added a free tier gating and price display to the model selection interface.
· Real-Time Model Switching: Support for switching models and providers mid-conversation on CLI, Telegram, Discord, Slack, or any gateway platform. It has aggregator-aware resolution, aims to stay on OpenRouter/Nous as much as possible, and automatically falls back to cross-provider when necessary. Telegram and Discord support interactive model selectors with inline buttons.
· Self-Optimizing GPT/Codex Tool Usage Guide: Through automated behavioral benchmarking, the agent diagnosed and fixed five failure modes of GPT and Codex tool invocation, significantly improving the reliability of OpenAI models. This includes execution best practices and a "pure thinking" prefix continuation for structured reasoning.
· Google AI Studio (Gemini) Native Provider: Directly access the Gemini model via the Google AI Studio API. Integrated with automated models.dev registry for real-time context length detection across providers.
