Gemini LLM¶
Runs a prompt against Google’s Gemini models via the LiteLLM service. Builds a chat-style message array from a system prompt and a user prompt (with optional context variables), then returns the model’s text response. The node auto-fetches available Gemini models from the service and falls back to a curated list if the service is unavailable.
Usage¶
Use this node whenever you need a Gemini-family LLM to generate, analyze, summarize, or transform text. Choose a model, set a system prompt to control tone and behavior, and write a user prompt. Optionally inject up to four context strings into the prompt using {{input_1}}–{{input_4}} placeholders. Temperature and max_tokens control creativity and response length. Connect this node after upstream nodes that produce context or knowledge and pass those into input_1–input_4.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| model | True | CHOICE | Select the Gemini model to use. Options are fetched from the LiteLLM service; if unavailable, a fallback list is shown. | gemini-2.5-pro |
| system_prompt | True | STRING | Instructions that set behavior, style, or constraints for the model. If left blank, only the user prompt is sent. | Respond concisely and use bullet points where helpful. |
| prompt | True | DYNAMIC_STRING | The user message. Supports placeholders {{input_1}}–{{input_4}} that are replaced with the optional inputs. | Summarize the following notes: {{input_1}} |
| temperature | True | FLOAT | Controls randomness/creativity. Higher values produce more varied outputs. | 0.7 |
| max_tokens | True | INT | Upper bound on response length in tokens. | 1024 |
| input_1 | False | STRING | Optional context string. Use via {{input_1}} in the prompt. | Meeting notes from 2025-10-05 ... |
| input_2 | False | STRING | Optional context string. Use via {{input_2}} in the prompt. | Product specs v3.2 ... |
| input_3 | False | STRING | Optional context string. Use via {{input_3}} in the prompt. | Customer email thread ... |
| input_4 | False | STRING | Optional context string. Use via {{input_4}} in the prompt. | Support ticket #12345 details ... |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| Output | STRING | The text response returned by the selected Gemini model. | Here is a concise summary of your notes... |
Important Notes¶
- Model options are cached per provider for approximately 1 hour; if live fetching fails, a built-in fallback model list is used.
- Use {{input_1}}–{{input_4}} exactly in the prompt to inject the optional inputs; unmatched placeholders remain unchanged.
- Temperature and max_tokens are passed to the model as generation parameters and directly influence style and length.
- A sensible default system prompt is provided; clear, specific prompts with relevant context produce the best results.
- If a selected model is not in the available list, the node will raise an error prompting you to choose a valid option.
Troubleshooting¶
- Model not found: Ensure the selected model name matches an available option. Refresh the model list or pick from the dropdown.
- Service request failed: Network or service issues can cause failures. Retry later or use a different model. The node will fall back to a predefined model list for selection, but requests still require service availability.
- Empty or overly short output: Increase max_tokens or reduce prompt constraints. Provide more context via input_1–input_4.
- Incoherent or too-random responses: Lower the temperature to make outputs more deterministic.
- Placeholders not replaced: Verify you used {{input_1}}–{{input_4}} syntax in the prompt and provided corresponding input values.