Use the Hermes Agent Adapter

The NVIDIA NeMo Fabric nvidia.fabric.hermes adapter runs Hermes Agent through its Python SDK. The adapter maps normalized NeMo Fabric configuration into Hermes Agent models, tools, skills, Model Context Protocol (MCP) servers, sessions, and telemetry.

Install the Adapter

Hermes Agent and the adapter support Python 3.11 through 3.13. Hermes Agent 0.20 and later is no longer installable from PyPI. End users must first follow the Hermes Agent installation guide.

Install the NeMo Fabric runtime and bare Hermes adapter into the Python environment that runs Hermes Agent:

$ pip install "nemo-fabric[hermes-agent]"

These packages do not install Hermes Agent. If a different environment runs NeMo Fabric, install only the bare adapter in the Hermes Agent environment:

$ python -m pip install nemo-fabric-adapters-hermes

The adapter package also provides relay and full extras. Both install the NeMo Relay Python package, but neither installs Hermes Agent.

For local NeMo Fabric development, run the following command from the repository root. The recipe checks out the pinned Hermes Agent source and synchronizes it into the project environment:

$ just install-hermes-agent

For separate environments, set ADAPTER_PYTHON in the runtime environment to the adapter environment’s Python interpreter. Use matching NeMo Fabric release versions for the runtime and adapter package unless a different pairing has been explicitly validated.

Configure the Adapter

Select the Hermes Agent harness integration in HarnessConfig:

1 from nemo_fabric import HarnessConfig
2 
3 harness = HarnessConfig(adapter_id="nvidia.fabric.hermes")

Use normalized FabricConfig fields to configure the model, workspace, skills, MCP servers, instructions, turn limit, native tool selectors, and telemetry.

The Hermes descriptor validates the following harness.settings fields:

Setting	Type	Default	Description
`reasoning_config`	object	`{"effort": "none"}`	Configures Hermes model reasoning. The closed object accepts an optional `enabled` boolean and an optional `effort` value of `none`, `minimal`, `low`, `medium`, `high`, or `xhigh`.
`plugins_enabled`	array of nonempty strings	`[]`	Enables Hermes plugins by identifier. NeMo Fabric adds `observability/nemo_relay` when Relay telemetry is enabled.
`save_trajectories`	boolean	`false`	Enables Hermes-native JSONL conversation trajectory saving. This is separate from normalized NeMo Fabric telemetry.
`max_tokens`	positive integer	`512`	Limits the number of tokens in each Hermes model response.
`terminal_timeout`	positive number	`60`	Limits a Hermes terminal operation in seconds.

For example:

1 from nemo_fabric import HarnessConfig
2 
3 harness = HarnessConfig(
4     adapter_id="nvidia.fabric.hermes",
5     settings={
6         "reasoning_config": {"enabled": True, "effort": "medium"},
7         "plugins_enabled": ["disk-cleanup"],
8         "save_trajectories": True,
9         "max_tokens": 1024,
10         "terminal_timeout": 90,
11     },
12 )

Use runtime.max_turns to set the Hermes agent-loop budget. The adapter maps that normalized field to AIAgent.max_iterations; max_iterations is not a Hermes harness setting.

Understand the Runtime Lifecycle

Each NeMo Fabric runtime creates one Hermes Agent AIAgent and one SessionDB. Ordered invocations reuse those objects and the conversation history until the runtime stops. When NeMo Relay telemetry is enabled, the adapter finalizes Agent Trajectory Observability Format (ATOF) and Agent Trajectory Interchange Format (ATIF) artifacts after each invocation.

For the complete configuration and lifecycle reference, refer to the Hermes Agent adapter guide.