Hermes AgentInstall

Summary

Send each input item’s prompt to a Hermes API Server using the Responses API or Chat Completions endpoint, then return the assistant text and available response metadata.

Inputs

• Prompt (required) — Prompt sent to Hermes. Defaults to the first truthy value among $json.prompt, $json.chatInput, and $json.text; the value must be non-empty after trimming.
• System / Instructions — Optional instructions included as the Responses API instructions field or as a system message for Chat Completions.
• Endpoint Mode — Select Responses API (/responses, default) or Chat Completions (/chat/completions).
• Model — Model name sent in the request; defaults to hermes-agent.
• Session Mode — Stateless by default. Stateful mode sends X-Hermes-Session-Key; for Responses API it also sets store: true and can chain from a previous response ID.
• Session Key — Required in stateful mode. Sent as the X-Hermes-Session-Key header.
• Previous Response ID — For stateful Responses API requests, optionally supplies previous_response_id to continue a response chain. Defaults to $json.responseId when present.
• Include Raw Response — When enabled, adds the unmodified API response as raw in the output.
• Include Intermediate Steps — For Responses API only, adds function_call and function_call_output entries as intermediateSteps.
• Max Retries — Retry count for HTTP 429, 500, 502, or 503 responses; defaults to 2 and is limited to 0–3.

Output shape

One output item per input item, each containing a JSON object.

Successful output includes text, responseId when supplied by the server, sessionKey in stateful mode, and usage when available. Optional fields are raw and, for Responses API, intermediateSteps. If continue-on-fail is enabled, a failed item instead contains its original JSON plus an error message; otherwise the node raises an item-scoped error.

Links

• n8n community nodes installation guide