Nous Research has officially announced the release of Hermes Agent v0.11.0, codenamed "The Interface Release." This significant update introduces a fundamental architectural refACToring, a completely rewritten user interface, and expanded model support, including the highly anticipated integration of GPT-5.5. The most substantial change in v0.11.0 is the introduction of a Pluggable Transport Layer (Transport ABC). Previously, format conversion and HTTP transmission for various inference providers were tightly coupled within a single module. This version decouples them into distinct, independent implementations: This new abstraction allows for seamless integration of new providers simply by implementing the corresponding interface. Built upon this new architecture is native support for AWS Bedrock, leverAGIng the Converse API. 🚀 Expanded Model Support and Inference Paths
Hermes Agent v0.11.0 significantly broadens its compatibility with leading AI models and inference services: GPT-5.5 Integration: The Framework now supports OpenAI's latest gpt-5.5 reasoning model through Codex OAuth.
Dynamic Model Discovery: The model selector can now discover new models in real-time, eliminating the need for manual directory updates when new models are released.
New Inference Paths: Five new推理 paths have been added, including NVIDIA NIM, Arcee AI, Step Plan, Google Gemini CLI OAuth, and the Vercel AI Gateway, which features dynamic pricing discovery.
✨ Completely Rewritten Interactive Terminal (TUI)
The user interface has been completely rebuilt from the ground up using React and Ink, with a Python JSON-RPC backend. The new Terminal User Interface (TUI), launched via hermes --tui, offers a modern and robust experience with features such as:
A fixed input editor and real-time streaming ouTPUt.
A status bar displaying the Git branch and a per-turn timer.
OSC-52 clipboard support and a light theme preset.
A visual overlay for monitoring the activity of spawned sub-agents.
🧩 Enhanced plugin system and Agent Control
The plugin system has been greatly expanded, granting developers more control and flexibility:
Slash Commands: Plugins can now register custom slash commands.
Tool Interception: Plugins can directly schedule tools, intercept or veto tool execution, and rewrite tool results.
Dashboard Customization: Developers can add custom tabs to the agent's dashboard.
A new /steer command allows users to inject Prompts mid-task. The agent will see this instruction after its next tool call, enabling course correction without interrupting the current turn or breaking the prompt cache. Sub-agent capabilities have also been improved with a dedicated "orchestrator" role, the ability to configure maximum nesting depth, and a file-based coordination layer to prevent concurrent sub-agents from overwriting each other's work. 🌐 New Communication Platform: QQBot
Hermes Agent continues to expand its reach across communication platforms. QQBot is now supported as the 17th platform, built upon the official QQ API v2 adapter. It features QR code-based configuration, streaming cursors, and EMOji reactions.
Be the first to rate this article.
Comments & Questions (0)
No comments yet
Be the first to comment!