@intrvls/n8n-nodes-google-gemini

An n8n community node that exposes a Google Gemini Chat Model as an AI Language Model sub-node, built on
@intrvls/langchain-google-genai — the standalone port of
@langchain/google-genai re-based onto the new @google/genai SDK and pinned to the @langchain/core line n8n ships
(1.1.x).

It is a feature-complete equivalent of n8n's built-in LmChatGoogleGemini node, packaged for standalone/community
installation and wired to the port instead of the upstream LangChain package.

Features

• Outputs an ai_languageModel connection — drop it onto any AI Agent or Chain node.
• Model selector populated live from the Gemini API (GET /v1beta/models), filtering out embedding/imagen models
  and anything older than Gemini 2.5 (1.x, 2.0, and unversioned aliases), sorted alphabetically.
• Node versions 1 and 1.1 with version-appropriate default models
  (models/gemini-2.5-flash for v1, models/gemini-3-flash-preview for v1.1+).
• Thinking / reasoning controls:
  • Thinking Level (Minimal / Low / Medium / High) — the reasoning-effort knob for Gemini 3+.
  • Thinking Budget (token integer; -1 dynamic, 0 off) — the equivalent control for Gemini 2.5.
  • Include Thoughts — return the model's thought summaries when available.
  • These are sent as thinkingConfig and omitted entirely when unset (the API rejects them on non-thinking models).
• Options: maximum output tokens, sampling temperature, top-K, top-P, stop sequences (up to 5),
  maximum retries, and JSON-only output.
• Safety settings (fixedCollection): per-HarmCategory block thresholds, mapped to @google/genai's
  HarmCategory / HarmBlockThreshold enums.
• Uses the standard googlePalmApi credential (API key + host).

Installation

In n8n: Settings → Community Nodes → Install, then enter @intrvls/n8n-nodes-google-gemini.

Compatibility

• n8n: requires a version on the @langchain/core@1.1.x line (the node and its underlying port are pinned to it).
• Node.js: >=20.
• Underlying library: @intrvls/langchain-google-genai@^3.0.0-alpha.0.

Local development

npm install
npm run build   # tsc && gulp build:icons — compiles to dist/ and copies the SVG icon
npm run dev     # tsc --watch
npm run lint    # tsc --noEmit (type-check only)

The compiled dist/credentials and dist/nodes entries are referenced from the n8n block in package.json.
dist/ is gitignored and rebuilt automatically on publish via the prepublishOnly script.

Credentials

Create a Google Gemini(PaLM) Api (googlePalmApi) credential:

The credential authenticates by appending ?key=<apiKey> and is tested against /v1beta/models.

Links

• Repository: https://github.com/intrvls/n8n-nodes-google-gemini
• Issues: https://github.com/intrvls/n8n-nodes-google-gemini/issues
• Underlying library: https://www.npmjs.com/package/@intrvls/langchain-google-genai

License

MIT