GeminiRequest

Defined in: adapters/gemini.ts:90

Google Gemini API request structure.

Defines parameters for making a request to Google’s Gemini API. Key differences from other providers: uses contents (plural) for messages, systemInstruction for system prompts, and generation parameters in a nested generationConfig object. Temperature range is 0-1 (not 0-2 like some other providers).

See

• GeminiContent
• GeminiResponse
• https://ai.google.dev/api/rest/v1/models/generateContent

Example

const request: GeminiRequest = {
  contents: [
    {
      role: 'user',
      parts: [{ text: 'Explain quantum computing' }]
    }
  ],
  systemInstruction: {
    parts: [{ text: 'You are a helpful physics teacher.' }]
  },
  generationConfig: {
    temperature: 0.7,
    maxOutputTokens: 1024
  }
};

Properties

contents

contents: GeminiContent[]

Defined in: adapters/gemini.ts:92

Array of conversation contents (user and model messages)

---

generationConfig?

optional generationConfig: object

Defined in: adapters/gemini.ts:98

Generation parameters nested in config object

maxOutputTokens?

optional maxOutputTokens: number

Maximum tokens to generate

stopSequences?

optional stopSequences: string[]

Stop sequences - generation stops when encountered

temperature?

optional temperature: number

Sampling temperature 0-1 (Gemini uses 0-1, not 0-2)

topK?

optional topK: number

Top-K sampling parameter

topP?

optional topP: number

Nucleus sampling parameter 0-1

---

systemInstruction?

optional systemInstruction: object

Defined in: adapters/gemini.ts:95

System instruction/prompt (separate from contents)

parts

parts: object[]