Reusable skills, chat templates, bundles, and utilities for practical local AI agent workflows.
Hermes Local AI Stack is an independent community collection built around HermesAgent, local model runtimes such as LM Studio, and reusable agent workflows. Each product is packaged so people can adopt it on its own or combine it with other local-stack components.
Catalog · Contributing · License · HermesAgent · LM Studio
| Asset type | Purpose |
|---|---|
| Skills | Reusable HermesAgent-compatible procedures and supporting assets. |
| Chat templates | Model-specific prompt rendering for conversation history, reasoning, and structured tool use. |
| Bundles | Coordinated collections for complete, repeatable workflows. |
| Utilities | Focused helpers for local-stack setup, operation, and maintenance. |
Products are self-contained and may be installed, adapted, and used independently.
Caduceus is the first shipped product in the repository.
| Product | Asset type | Model family | Purpose | Files |
|---|---|---|---|---|
| Caduceus v1.8 — Recommended | Chat template | Qwen3.6 | Maintains current-task continuity, emits structured tool calls, groups independent actions, sequences dependencies, and grounds completion in returned results. | Template · Guide |
| Caduceus v1.7 — Compatibility | Chat template | Qwen3.6 | Retains the earlier Caduceus thinking and history defaults for compatible setups. | Template · Guide |
The Caduceus architecture guide explains how prompt rendering, model output, structured-call parsing, and tool execution fit together.
- Choose an item from the catalog.
- Read the item’s own README or product guide.
- Copy or install it into the relevant part of your local stack.
Each product documents its dependencies, controls, setup, and integration points close to the files you will use.
Clone the collection:
git clone https://github.com/Stacey2911/hermes-local-ai-stack.git
cd hermes-local-ai-stackThen configure Caduceus for a Qwen3.6 model:
- Load the intended Qwen3.6 model in LM Studio.
- Open the model under My Models.
- Expose or select the Prompt Template field.
- Select the Jinja template option.
- Paste the contents of
qwen3.6-caduceus-v1.8.jinja.
Start the LM Studio server from the app or run:
lms server start --port 1234Configure HermesAgent:
hermes setupFor an existing HermesAgent installation, choose the model provider again:
hermes modelSelect LM Studio and the Qwen3.6 model loaded there. The Caduceus for Qwen3.6 guide covers version selection, controls, and workflow behavior.
Official references:
chat-templates/<model-family>/ Model-specific chat templates
skills/<category>/<name>/ Reusable skill packages
bundles/<name>/ Coordinated workflow bundles
tools/<name>/ Local-stack utilities
docs/ Product and architecture guides
Category directories are added when a product ships in that category, keeping the collection focused on usable assets.
Focused contributions are welcome when they add or improve a reusable asset, its documentation, or its integration instructions. See CONTRIBUTING.md for directory conventions, product documentation expectations, publication safety, and the contribution flow.
Hermes Local AI Stack is independently maintained for the HermesAgent and local-AI community. Products in the catalog are designed to be understandable, adaptable, and useful in individual local setups.
The repository is licensed under Apache-2.0. Its templates, skills, bundles, and utilities are provided as-is for users to inspect, adapt, and integrate. See THIRD_PARTY_NOTICES.md for upstream attribution.