n8n-nodes-prenzllmgateway

A custom n8n community node that exposes a private LLM gateway as a Chat Model. It plugs into the AI Agent (Tools Agent) and Basic LLM Chain Chat Model slot, just like the built-in OpenAI / Anthropic chat-model nodes — including function/tool calling.

It talks to a backend endpoint, POST /api/agent/messages, which:

• takes a native OpenAI-style messages array (system / user / assistant / tool),
• selects the assistant by persona id or name (which governs model + system prompt),
• passes tools / tool_choice straight through and returns tool_calls — so the n8n Tools Agent works end-to-end.

Requirements

• A gateway instance exposing /api/agent/messages and /api/agent/assistants.
• An API key (sent as Authorization: Bearer <key> by default).
• n8n with community nodes enabled. For use inside an AI Agent, also set N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true.

Credentials — "Prenz LLM Gateway API"

| Field | Default | Notes | |---|---|---| | Base URL | – (required) | Root URL of your gateway, no trailing slash | | API Key | – | API key for the gateway | | Auth Header Name | Authorization | Change if your gateway expects e.g. X-API-Key | | Add "Bearer " Prefix | true | Standard Bearer-token auth |

The credential Test button uses the Authorization header. The node's own requests honour a custom Auth Header Name if you set one.

Node properties

• Persona (Assistant) — dropdown loaded live from /api/agent/assistants. Selects the assistant by id. Leave on Default LLM to use the instance default model with no persona.
• Options → Timeout (ms) — request timeout (default 60000).
• Options → Temperature — only applied when using the default LLM provider.
• Options → Max Tokens — -1 uses the provider default.

Build & publish

npm install --ignore-scripts
npm run build               # tsc + copies icon and codex metadata into dist/
npm pack --dry-run          # inspect the exact files that will be published
npm publish --access public # publish to npm

Then install in n8n via Settings → Community Nodes → Install → n8n-nodes-prenzllmgateway, with N8N_COMMUNITY_PACKAGES_ENABLED=true and N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true.

Usage

1. Add an AI Agent (or Basic LLM Chain) node.
2. Click the Chat Model + → choose Prenz LLM Gateway Chat Model.
3. Select/create the Prenz LLM Gateway API credential and pick a persona.
4. (Optional) connect tools to the agent — the node forwards them and handles tool_calls.

Notes / limitations

• Tool calling depends on the underlying LLM provider/model actually emitting tool calls.
• No streaming in v1 (single JSON response). SSE can be added later via _streamResponseChunks.
• Token usage is not reported (the backend endpoint does not return usage in v1).
• @langchain/core is an optional peer dependency on purpose — n8n provides it, and a second copy would break LangChain's instanceof checks inside the Agent.